Split (4)

long_tts_eval_en · 360 rows

id stringlengths 9 11	category stringclasses 6 values	word_count int64 62 2.45k	language stringclasses 1 value	text stringlengths 330 16.4k	text_norm stringlengths 330 17.5k
long_en_316	wiki_en	847	en	Travel beyond Earth's atmosphere and magnetosphere and exposure to microgravity can harm human health, and understanding these effects is essential for successful crewed spaceflight. Potential impacts on the central nervous system (CNS) are particularly important. Ground-based cellular and animal research will help quantify CNS risk from space radiation on long-duration missions and support development of optimized countermeasures. Acute and late CNS risks from galactic cosmic rays (GCRs) and solar proton events (SPEs) are a documented concern for human exploration. Historically, CNS effects in adults exposed to low-to-moderate doses of ionizing radiation (0–2 Gy; 1 Gy = 100 rad) were not a major focus, but the heavy-ion component of space radiation poses distinct biophysical challenges compared with terrestrial radiation. Soon after cosmic rays were discovered, concern about CNS effects arose with the prediction that single high-charge, high-energy (HZE) nuclei traversing the retina would produce light flashes—a phenomenon confirmed by Apollo astronauts in 1970 and 1973. HZE nuclei can create a column of severely damaged cells, or a microlesion, along their path through tissue, raising concern about serious impacts on the CNS. In recent years, concerns have arisen about neurogenesis and its impairment by HZE nuclei observed in experimental CNS models. Human epidemiology informs risk estimates for cancer, acute radiation effects, and cataracts, but it is not viable for estimating CNS risks from space radiation. At doses above a few gray, patients treated with radiation (e.g., gamma rays and protons) for cancer experience detrimental CNS changes; typical treatment doses are about 50 Gy, well above exposures expected in space even during a large solar particle event. Thus, of the four categories of space radiation risks—cancer, CNS, degenerative, and acute radiation syndromes—the CNS risk depends most heavily on animal experimental data. Understanding and mitigating CNS risks requires a vigorous research program drawing on cellular and animal models and on methods to extrapolate risks and evaluate potential countermeasures for astronauts. Several studies using heavy-ion beams that simulate space radiation provide evidence of CNS effects: exposure to HZE nuclei at low doses (<50 cGy) significantly induces neurocognitive deficits in mice and rats, including learning and behavioral changes and altered operant responses. Exposures to equal or higher doses of low-LET radiation (e.g., gamma or X-rays) do not produce similar effects. The threshold for performance deficits following HZE exposure depends on particle characteristics, such as linear energy transfer (LET), and the animal’s age at exposure. Performance deficits have been observed at doses comparable to those expected on a Mars mission (<0.5 Gy). Neurocognitive deficits affecting the dopaminergic system resemble aging and appear unique to space radiation. HZE exposure also disrupts neurogenesis in mice at low doses (<1 Gy), causing a dose-related reduction in new neurons and oligodendrocytes in the subgranular zone (SGZ) of the hippocampal dentate gyrus. Exposure to HZE nuclei and protons at low doses induces reactive oxygen species (ROS) in neuronal precursor cells that can persist for months; antioxidants and anti-inflammatory agents may mitigate these effects. HZE and proton exposure also elicit neuroinflammation, and age-related genetic changes increase CNS radiosensitivity. Animal studies using HZE irradiation show significant CNS changes at dose levels relevant to NASA, but the implications for astronaut morbidity remain unclear. One model of late tissue effects suggests that significant effects occur at lower doses but with increased latency. Studies to date have used relatively small numbers of animals (fewer than 10 per dose group), so testing of dose-threshold effects at lower doses (less than 0.5 Gy) has been insufficient. Extrapolating animal data to humans is a major challenge and may be limited by small animal study sizes. The role of dose protraction has not been studied. No reliable method exists to extrapolate existing observations to possible cognitive changes, performance degradation, or late central nervous system (CNS) effects in astronauts. New systems-biology approaches offer promising tools to address this challenge. Recently, eight gaps were identified in projecting CNS risks; research on new risk-assessment approaches is needed to provide the data and knowledge required to develop CNS risk-projection models for space radiation. Both galactic cosmic rays (GCRs) and solar particle events (SPEs) are of concern for CNS risks. Major GCR components include protons, α-particles, and high charge and energy (HZE) nuclei, with a broad energy spectrum ranging from a few tens to more than 10,000 MeV per nucleon (MeV/u). In interplanetary space, galactic cosmic rays (GCRs) are expected to deliver an organ absorbed dose of more than 0.2 Gy per year and a dose equivalent exceeding 0.6 Sv per year. The high energies of GCRs allow them to penetrate hundreds of centimeters of most materials, making shielding an impractical mitigation for GCR risks to the central nervous system. Solar particle events (SPEs) can produce absorbed doses above 1 Gy for crew in thinly shielded spacecraft or during extravehicular activity. Although SPE particle energies (tens to hundreds of MeV) can be attenuated by shielding, protecting against the largest events would be costly.	Travel beyond Earth's atmosphere and magnetosphere and exposure to microgravity can harm human health, and understanding these effects is essential for successful crewed spaceflight. Potential impacts on the central nervous system (CNS) are particularly important. Ground-based cellular and animal research will help quantify CNS risk from space radiation on long-duration missions and support development of optimized countermeasures. Acute and late CNS risks from galactic cosmic rays (GCRs) and solar proton events (SPEs) are a documented concern for human exploration. Historically, CNS effects in adults exposed to low-to-moderate doses of ionizing radiation (zero to two Gy; one Gy equals one hundred rad) were not a major focus, but the heavy-ion component of space radiation poses distinct biophysical challenges compared with terrestrial radiation. Soon after cosmic rays were discovered, concern about CNS effects arose with the prediction that single high-charge, high-energy (HZE) nuclei traversing the retina would produce light flashes—a phenomenon confirmed by Apollo astronauts in nineteen seventy and nineteen seventy-three. HZE nuclei can create a column of severely damaged cells, or a microlesion, along their path through tissue, raising concern about serious impacts on the CNS. In recent years, concerns have arisen about neurogenesis and its impairment by HZE nuclei observed in experimental CNS models. Human epidemiology informs risk estimates for cancer, acute radiation effects, and cataracts, but it is not viable for estimating CNS risks from space radiation. At doses above a few gray, patients treated with radiation (e.g., gamma rays and protons) for cancer experience detrimental CNS changes; typical treatment doses are about fifty Gy, well above exposures expected in space even during a large solar particle event. Thus, of the four categories of space radiation risks—cancer, CNS, degenerative, and acute radiation syndromes—the CNS risk depends most heavily on animal experimental data. Understanding and mitigating CNS risks requires a vigorous research program drawing on cellular and animal models and on methods to extrapolate risks and evaluate potential countermeasures for astronauts. Several studies using heavy-ion beams that simulate space radiation provide evidence of CNS effects: exposure to HZE nuclei at low doses (less than fifty cGy) significantly induces neurocognitive deficits in mice and rats, including learning and behavioral changes and altered operant responses. Exposures to equal or higher doses of low-LET radiation (e.g., gamma or X-rays) do not produce similar effects. The threshold for performance deficits following HZE exposure depends on particle characteristics, such as linear energy transfer (LET), and the animal’s age at exposure. Performance deficits have been observed at doses comparable to those expected on a Mars mission (less than zero point five Gy). Neurocognitive deficits affecting the dopaminergic system resemble aging and appear unique to space radiation. HZE exposure also disrupts neurogenesis in mice at low doses (less than one Gy), causing a dose-related reduction in new neurons and oligodendrocytes in the subgranular zone (SGZ) of the hippocampal dentate gyrus. Exposure to HZE nuclei and protons at low doses induces reactive oxygen species (ROS) in neuronal precursor cells that can persist for months; antioxidants and anti-inflammatory agents may mitigate these effects. HZE and proton exposure also elicit neuroinflammation, and age-related genetic changes increase CNS radiosensitivity. Animal studies using HZE irradiation show significant CNS changes at dose levels relevant to NASA, but the implications for astronaut morbidity remain unclear. One model of late tissue effects suggests that significant effects occur at lower doses but with increased latency. Studies to date have used relatively small numbers of animals (fewer than ten per dose group), so testing of dose-threshold effects at lower doses (less than zero point five Gy) has been insufficient. Extrapolating animal data to humans is a major challenge and may be limited by small animal study sizes. The role of dose protraction has not been studied. No reliable method exists to extrapolate existing observations to possible cognitive changes, performance degradation, or late central nervous system (CNS) effects in astronauts. New systems-biology approaches offer promising tools to address this challenge. Recently, eight gaps were identified in projecting CNS risks; research on new risk-assessment approaches is needed to provide the data and knowledge required to develop CNS risk-projection models for space radiation. Both galactic cosmic rays (GCRs) and solar particle events (SPEs) are of concern for CNS risks. Major GCR components include protons, alpha-particles, and high charge and energy (HZE) nuclei, with a broad energy spectrum ranging from a few tens to more than ten thousand MeV per nucleon (MeV slash u). In interplanetary space, galactic cosmic rays (GCRs) are expected to deliver an organ absorbed dose of more than zero point two Gy per year and a dose equivalent exceeding zero point six Sv per year. The high energies of GCRs allow them to penetrate hundreds of centimeters of most materials, making shielding an impractical mitigation for GCR risks to the central nervous system. Solar particle events (SPEs) can produce absorbed doses above one Gy for crew in thinly shielded spacecraft or during extravehicular activity. Although SPE particle energies (tens to hundreds of MeV) can be attenuated by shielding, protecting against the largest events would be costly.
long_en_306	wiki_en	836	en	Spacetoon is a pan-Arab free-to-air television channel that specializes in animation and children's programs. It began broadcasting on 15 March 2000 and is headquartered in Dubai, United Arab Emirates, with offices in Riyadh. The channel targets children aged four and up, while its late-night block, Space Power, is aimed at teenagers and young adults. The Spacetoon company also maintains a video-on-demand app called Spacetoon Go and has had two now-defunct channels in the Arab world, Space Power TV and Spacetoon English. The main Indonesian channel began airing on 24 March 2005 in Jakarta. It later became NET. and continues to broadcast via satellite. Currently, there are three Spacetoon channels in Indonesia: Spacetoon, Space Shopping, and Spacetoon Plus. In India, Spacetoon India exists as a licensing company rather than a separate TV channel. Spacetoon launched in South Korea in 2005 but has since closed down. Spacetoon is currently broadcast in 22 countries and has an audience of over 130 million viewers. History — Arab world: In 1999, Bahrain Radio and Television Corporation signed an agreement to broadcast a children's cartoon channel. On 15 March 2000, Spacetoon officially launched as a seven-hour block. According to the Ministry, it continued to air in this format until the contract ended on 12 January 2002. Later, in 2004, Spacetoon moved its headquarters to Dubai and was established as an independent channel on Nilesat, broadcasting 24/7. In the Arab world, most programs are dubbed into Modern Standard Arabic. Spacetoon is closely affiliated with Venus Center, a Syrian dubbing company that has historically provided the Arabic dubbed versions of its programming. Using Modern Standard Arabic in dubbing played a crucial role in preserving the language among children, especially given the emergence of regional spoken Arabic dialects. In Indonesia, Spacetoon officially launched on 24 March 2005. The network was founded by H. Sukoyo, co-founder of TV7. At launch, Spacetoon broadcast from 6:00 a.m. to 9:30 p.m. WIB; later the schedule extended from 5:00 a.m. to 11:00 p.m. When programming ended, a 10-minute segment of animation, songs, or messages for children was shown. In mid-2011, due to financial problems, Spacetoon began airing some home-shopping and alternative medicine programs. In March 2013, Net Visi Media acquired 95% of Spacetoon's ownership stakes. On 18 May 2013, Spacetoon ceased terrestrial broadcasts to make way for NET on the terrestrial network, while it continued to broadcast in Indonesia via satellite television. In September 2014, Spacetoon split into two channels: Spacetoon and Spacetoon 2. Spacetoon 2 aired more cartoons and animated series than Spacetoon, although it also carried some home-shopping programs. In May 2016, Spacetoon added another channel, Spacetoon 3. It had clearer audio than Spacetoon and Spacetoon 2, but was closed down in October of the same year. In November 2016, Spacetoon 2 was renamed Space Shopping because home-shopping programs contributed the most revenue to the channel, which otherwise had little income. Currently, Spacetoon has three channels in Indonesia: Spacetoon, Space Shopping, and Spacetoon Plus. Programming planets Each program was divided into blocks called "planets", one for each genre: - Action Planet (أکشن) — planet of excitement and mystery; for action series such as Dragon Ball and Iron Kid. - Sport Planet (رياضة) — planet of challenge and strength; for sports series and programs such as Inazuma Eleven. - Adventures Planet (مغامرات) — planet of imagination and thrill; for adventure series such as Future Boy Conan. - Comedy Planet (كوميديا) — planet of laughter; for comedy series such as Woody Woodpecker, Tom & Jerry Kids, and The Pink Panther. - Movies Planet (أفلام) — planet of all colors; for animated movies. - Abjad Planet (أبجد) — planet of numbers and letters; for educational programs such as Pappyland. - Bon Bon Planet (بون بون) — planet of heroes and adults; for preschool programs such as Thomas & Friends. History Planet (تاريخ; "planet of time immemorial") — formerly used from 2000 to 2013 — features historical series such as Liberty's Kids. Science Planet (علوم; "planet of discovery and knowledge") airs educational science programs such as Inspector Gadget's Field Trip. Zomoroda Planet (زمردة; "emerald") is a planet just for girls, featuring series and programs like Magical Princess Minky Momo. Programs aired on Spacetoon are sometimes censored from their source material. Scenes may be cropped or truncated to avoid showing excessive violence, as seen in Detective Conan, Romeo's Blue Skies, and Hunter × Hunter. Hunter × Hunter was also altered to remove depictions of Zen Buddhism and Taoism related to characters' powers; the censored version presents those powers instead as science and martial art. Selective cropping and editing are also used to hide cleavages and remove innuendo. See also: - Animation International - Game Power 7 (game publisher owned by Spacetoon) - Spacetoon Go References / Sources / External links: - Spacetoon Arabic website (Spacetoon Arabic page) - Spacetoon Indonesia website (Spacetoon Indonesia page) Categories: - Mass media companies of the United Arab Emirates - Arabic-language television stations - Children's television networks - Television channels and stations established in 2000 - 2003 establishments in the United Arab Emirates - Anime companies - Fictional planets - Preschool education television networks	Spacetoon is a pan-Arab free-to-air television channel that specializes in animation and children's programs. It began broadcasting on fifteen March two thousand and is headquartered in Dubai, United Arab Emirates, with offices in Riyadh. The channel targets children aged four and up, while its late-night block, Space Power, is aimed at teenagers and young adults. The Spacetoon company also maintains a video-on-demand app called Spacetoon Go and has had two now-defunct channels in the Arab world, Space Power TV and Spacetoon English. The main Indonesian channel began airing on twenty four March two thousand five in Jakarta. It later became NET. and continues to broadcast via satellite. Currently, there are three Spacetoon channels in Indonesia: Spacetoon, Space Shopping, and Spacetoon Plus. In India, Spacetoon India exists as a licensing company rather than a separate TV channel. Spacetoon launched in South Korea in two thousand five but has since closed down. Spacetoon is currently broadcast in twenty two countries and has an audience of over one hundred thirty million viewers. History — Arab world: In nineteen ninety nine, Bahrain Radio and Television Corporation signed an agreement to broadcast a children's cartoon channel. On fifteen March two thousand, Spacetoon officially launched as a seven-hour block. According to the Ministry, it continued to air in this format until the contract ended on twelve January two thousand two. Later, in two thousand four, Spacetoon moved its headquarters to Dubai and was established as an independent channel on Nilesat, broadcasting twenty four seven. In the Arab world, most programs are dubbed into Modern Standard Arabic. Spacetoon is closely affiliated with Venus Center, a Syrian dubbing company that has historically provided the Arabic dubbed versions of its programming. Using Modern Standard Arabic in dubbing played a crucial role in preserving the language among children, especially given the emergence of regional spoken Arabic dialects. In Indonesia, Spacetoon officially launched on twenty four March two thousand five. The network was founded by H. Sukoyo, co-founder of TV seven. At launch, Spacetoon broadcast from six a.m. to nine thirty p.m. WIB; later the schedule extended from five a.m. to eleven p.m. When programming ended, a ten-minute segment of animation, songs, or messages for children was shown. In mid-two thousand eleven, due to financial problems, Spacetoon began airing some home-shopping and alternative medicine programs. In March two thousand thirteen, Net Visi Media acquired ninety five percent of Spacetoon's ownership stakes. On eighteen May two thousand thirteen, Spacetoon ceased terrestrial broadcasts to make way for NET on the terrestrial network, while it continued to broadcast in Indonesia via satellite television. In September two thousand fourteen, Spacetoon split into two channels: Spacetoon and Spacetoon two. Spacetoon two aired more cartoons and animated series than Spacetoon, although it also carried some home-shopping programs. In May two thousand sixteen, Spacetoon added another channel, Spacetoon three. It had clearer audio than Spacetoon and Spacetoon two, but was closed down in October of the same year. In November two thousand sixteen, Spacetoon two was renamed Space Shopping because home-shopping programs contributed the most revenue to the channel, which otherwise had little income. Currently, Spacetoon has three channels in Indonesia: Spacetoon, Space Shopping, and Spacetoon Plus. Programming planets Each program was divided into blocks called "planets", one for each genre: - Action Planet (أکشن) — planet of excitement and mystery; for action series such as Dragon Ball and Iron Kid. - Sport Planet (رياضة) — planet of challenge and strength; for sports series and programs such as Inazuma Eleven. - Adventures Planet (مغامرات) — planet of imagination and thrill; for adventure series such as Future Boy Conan. - Comedy Planet (كوميديا) — planet of laughter; for comedy series such as Woody Woodpecker, Tom and Jerry Kids, and The Pink Panther. - Movies Planet (أفلام) — planet of all colors; for animated movies. - Abjad Planet (أبجد) — planet of numbers and letters; for educational programs such as Pappyland. - Bon Bon Planet (بون بون) — planet of heroes and adults; for preschool programs such as Thomas and Friends. History Planet (تاريخ; "planet of time immemorial") — formerly used from two thousand to two thousand thirteen — features historical series such as Liberty's Kids. Science Planet (علوم; "planet of discovery and knowledge") airs educational science programs such as Inspector Gadget's Field Trip. Zomoroda Planet (زمردة; "emerald") is a planet just for girls, featuring series and programs like Magical Princess Minky Momo. Programs aired on Spacetoon are sometimes censored from their source material. Scenes may be cropped or truncated to avoid showing excessive violence, as seen in Detective Conan, Romeo's Blue Skies, and Hunter times Hunter. Hunter times Hunter was also altered to remove depictions of Zen Buddhism and Taoism related to characters' powers; the censored version presents those powers instead as science and martial art. Selective cropping and editing are also used to hide cleavages and remove innuendo. See also: - Animation International - Game Power seven (game publisher owned by Spacetoon) - Spacetoon Go References / Sources / External links: - Spacetoon Arabic website (Spacetoon Arabic page) - Spacetoon Indonesia website (Spacetoon Indonesia page) Categories: - Mass media companies of the United Arab Emirates - Arabic-language television stations - Children's television networks - Television channels and stations established in two thousand - two thousand three establishments in the United Arab Emirates - Anime companies - Fictional planets - Preschool education television networks.
long_en_303	wiki_en	924	en	Strain engineering is a strategy used in semiconductor manufacturing to enhance device performance. Benefits are achieved by modulating strain in the transistor channel, which can improve electron or hole mobility and thus channel conductivity. Another example is semiconductor photocatalysts that are strain-engineered for more effective use of sunlight. The use of various strain engineering techniques has been reported by prominent microprocessor manufacturers, including AMD, IBM, and Intel, primarily in sub-130 nm technologies. One key consideration is that PMOS and NMOS respond differently to types of strain: PMOS benefits from compressive strain, while NMOS benefits from tensile strain. Many approaches induce strain locally, allowing n-channel and p-channel strain to be modulated independently. One common method uses a strain-inducing capping layer; CVD silicon nitride is often chosen because the magnitude and type of strain (tensile versus compressive) can be adjusted by changing deposition conditions, especially temperature. Standard lithography and patterning techniques can be used to selectively deposit strain-inducing capping layers—for example, to deposit a compressive film only over the PMOS. Capping layers are central to the Dual Stress Liner (DSL) approach reported by IBM and AMD: standard patterning is used to deposit a tensile silicon nitride film over the NMOS and a compressive silicon nitride film over the PMOS. Another approach uses silicon-rich solid solutions, especially silicon–germanium (SiGe), to modulate channel strain. One method involves epitaxial growth of silicon on a relaxed SiGe underlayer; tensile strain is induced in the silicon as its lattice is stretched to match the larger SiGe lattice. Conversely, compressive strain can be induced by using a solid solution with a smaller lattice constant, such as silicon–carbon. See, e.g., U.S. Patent No. 7,023,018. A related method replaces the source and drain regions of a MOSFET with silicon–germanium. In thin films, strain can be induced by epitaxial growth or, more recently, topological growth; epitaxial strain generally arises from lattice mismatch between the film and substrate or from thermal expansion mismatch. Tuning epitaxial strain can be used to modify the properties of thin films and induce phase transitions. The misfit parameter f is given by f = (a_f − a_s)/a_s, where a_f is the lattice parameter of the epitaxial film and a_s is the lattice parameter of the substrate. After some critical film thickness, it becomes energetically favorable to relieve mismatch strain through the formation of misfit dislocations or microtwins. Misfit dislocations can be interpreted as dangling bonds at an interface between layers with different lattice constants. The critical thickness h_c was computed by Matthews and Blakeslee; it depends on the Burgers vector length b, the Poisson ratio ν, the angle λ between the Burgers vector and the misfit dislocation line, and the angle α between the Burgers vector and the vector normal to the dislocation’s glide plane. The equilibrium in-plane strain for a thin film with thickness t that exceeds h_c is reduced relative to the original misfit strain and can be expressed in terms of the dislocation density. Strain relaxation at thin-film interfaces via misfit dislocation nucleation and multiplication occurs in three stages, distinguishable by the relaxation rate. The first stage is dominated by glide of pre-existing dislocations and is characterized by a slow relaxation rate. The second stage has a faster relaxation rate, which depends on the mechanisms for dislocation nucleation in the material. Finally, the last stage represents a saturation in strain relaxation due to strain hardening. Strain engineering is well studied in complex oxide systems: epitaxial strain can strongly influence coupling among spin, charge, and orbital degrees of freedom, thereby altering electrical and magnetic properties. Epitaxial strain has been shown to induce metal–insulator transitions and to shift the Curie temperature of the antiferromagnetic-to-ferromagnetic transition in La1-xSrxMnO3. In alloy thin films, epitaxial strain affects spinodal instability and thus the driving force for phase separation. This effect is explained by coupling between the imposed epitaxial strain and the composition-dependent elastic properties of the system. More recently, researchers have achieved strains in thick oxide films larger than those possible with epitaxial growth by incorporating nanostructured topologies (Guerra and Vezenov, 2002) and by embedding nanorods or nanopillars within an oxide film matrix. Following this work, researchers worldwide have produced self-organized, phase-separated nanorod and nanopillar structures in numerous oxide films. Thulin and Guerra (2008) calculated strain-modified band structures of anatase TiO2 and found that hole mobility increases with strain. Additionally, in two-dimensional materials, strain has been shown to convert an indirect semiconductor into a direct one, allowing up to a hundred-fold increase in the light emission rate. In normal bulk materials, the maximum elastic strain typically ranges from 0.1% to 1%, which limits our ability to reversibly and quantitatively modify material properties with strain. Recent research on nanoscale materials, however, shows a much broader elastic strain range; even diamond, the hardest natural material, can exhibit up to 9.0% uniform elastic strain at the nanoscale. Following Moore's law, semiconductor devices continue to shrink to the nanoscale, and because of the "smaller is stronger" effect, elastic strain engineering can be fully exploited there. Crystallographic direction plays a crucial role: most materials are anisotropic, so applying strain along different directions can significantly affect properties. For example, density functional theory (DFT) simulations for diamond show that strain along <110> causes a faster bandgap reduction, whereas strain along <111> yields a slower reduction but induces a transition from an indirect to a direct bandgap.	Strain engineering is a strategy used in semiconductor manufacturing to enhance device performance. Benefits are achieved by modulating strain in the transistor channel, which can improve electron or hole mobility and thus channel conductivity. Another example is semiconductor photocatalysts that are strain-engineered for more effective use of sunlight. The use of various strain engineering techniques has been reported by prominent microprocessor manufacturers, including AMD, IBM, and Intel, primarily in sub-one hundred thirty nanometer technologies. One key consideration is that PMOS and NMOS respond differently to types of strain: PMOS benefits from compressive strain, while NMOS benefits from tensile strain. Many approaches induce strain locally, allowing n-channel and p-channel strain to be modulated independently. One common method uses a strain-inducing capping layer; CVD silicon nitride is often chosen because the magnitude and type of strain (tensile versus compressive) can be adjusted by changing deposition conditions, especially temperature. Standard lithography and patterning techniques can be used to selectively deposit strain-inducing capping layers—for example, to deposit a compressive film only over the PMOS. Capping layers are central to the Dual Stress Liner (DSL) approach reported by IBM and AMD: standard patterning is used to deposit a tensile silicon nitride film over the NMOS and a compressive silicon nitride film over the PMOS. Another approach uses silicon-rich solid solutions, especially silicon–germanium (SiGe), to modulate channel strain. One method involves epitaxial growth of silicon on a relaxed SiGe underlayer; tensile strain is induced in the silicon as its lattice is stretched to match the larger SiGe lattice. Conversely, compressive strain can be induced by using a solid solution with a smaller lattice constant, such as silicon–carbon. See, e.g., U.S. Patent No. seven million twenty three thousand eighteen. A related method replaces the source and drain regions of a MOSFET with silicon–germanium. In thin films, strain can be induced by epitaxial growth or, more recently, topological growth; epitaxial strain generally arises from lattice mismatch between the film and substrate or from thermal expansion mismatch. Tuning epitaxial strain can be used to modify the properties of thin films and induce phase transitions. The misfit parameter f is given by f equals (a f minus a s) divided by a s, where a f is the lattice parameter of the epitaxial film and a s is the lattice parameter of the substrate. After some critical film thickness, it becomes energetically favorable to relieve mismatch strain through the formation of misfit dislocations or microtwins. Misfit dislocations can be interpreted as dangling bonds at an interface between layers with different lattice constants. The critical thickness h c was computed by Matthews and Blakeslee; it depends on the Burgers vector length b, the Poisson ratio nu, the angle lambda between the Burgers vector and the misfit dislocation line, and the angle alpha between the Burgers vector and the vector normal to the dislocation’s glide plane. The equilibrium in-plane strain for a thin film with thickness t that exceeds h c is reduced relative to the original misfit strain and can be expressed in terms of the dislocation density. Strain relaxation at thin-film interfaces via misfit dislocation nucleation and multiplication occurs in three stages, distinguishable by the relaxation rate. The first stage is dominated by glide of pre-existing dislocations and is characterized by a slow relaxation rate. The second stage has a faster relaxation rate, which depends on the mechanisms for dislocation nucleation in the material. Finally, the last stage represents a saturation in strain relaxation due to strain hardening. Strain engineering is well studied in complex oxide systems: epitaxial strain can strongly influence coupling among spin, charge, and orbital degrees of freedom, thereby altering electrical and magnetic properties. Epitaxial strain has been shown to induce metal–insulator transitions and to shift the Curie temperature of the antiferromagnetic-to-ferromagnetic transition in La1-xSrxMnO3. In alloy thin films, epitaxial strain affects spinodal instability and thus the driving force for phase separation. This effect is explained by coupling between the imposed epitaxial strain and the composition-dependent elastic properties of the system. More recently, researchers have achieved strains in thick oxide films larger than those possible with epitaxial growth by incorporating nanostructured topologies (Guerra and Vezenov, two thousand two) and by embedding nanorods or nanopillars within an oxide film matrix. Following this work, researchers worldwide have produced self-organized, phase-separated nanorod and nanopillar structures in numerous oxide films. Thulin and Guerra (two thousand eight) calculated strain-modified band structures of anatase Ti O two and found that hole mobility increases with strain. Additionally, in two-dimensional materials, strain has been shown to convert an indirect semiconductor into a direct one, allowing up to a hundred-fold increase in the light emission rate. In normal bulk materials, the maximum elastic strain typically ranges from zero point one percent to one percent, which limits our ability to reversibly and quantitatively modify material properties with strain. Recent research on nanoscale materials, however, shows a much broader elastic strain range; even diamond, the hardest natural material, can exhibit up to nine point zero percent uniform elastic strain at the nanoscale. Following Moore's law, semiconductor devices continue to shrink to the nanoscale, and because of the "smaller is stronger" effect, elastic strain engineering can be fully exploited there. Crystallographic direction plays a crucial role: most materials are anisotropic, so applying strain along different directions can significantly affect properties. For example, density functional theory (DFT) simulations for diamond show that strain along one hundred ten causes a faster bandgap reduction, whereas strain along one hundred eleven yields a slower reduction but induces a transition from an indirect to a direct bandgap.
long_en_297	wiki_en	907	en	The Changshu Institute of Technology (CIT) is a provincial public university characterized by polytechnic education, applied technology, and teacher training. It is located in the center of the Yangtze River Delta in Changshu, a satellite city of Suzhou, Jiangsu, China, and is across a lake from Shajiabang. The university is divided into the East Lake Campus and the Southeast Campus. It grew out of Suzhou Normal College and officially opened on 12 May 2004. The school motto is "Breed integrity, seek truth—keep at it every day, and you'll go far along the way." History The Changshu Institute of Technology was established in August 1958 and was originally known as Suzhou Normal Vocational College. It began recruiting graduates from teachers' colleges in 1978 after the restoration of the national college entrance examination. In 1989, with the approval of the Jiangsu provincial government and the State Education Commission, it joined Changshu Vocational University. In early 2002 CIT applied to offer undergraduate programs, and in 2004 the Ministry of Education approved the establishment of the Changshu Institute of Technology. May 12 is celebrated as the university's founding day. Specialties CIT offers seven main specialties and provides 44 undergraduate programs, including programs in arts, history, education, law, engineering, and science. Specialties: Arts, History, Education, Law, Engineering, Science, Management. CIT has established many campuses and facilities. Schools: School of Humanities; School of Mechanical Engineering; School of Computer Science and Engineering; School of Electrical Engineering and Automation; School of Management; School of Biology and Food Engineering; School of Mathematics and Statistics; School of Arts and Garments Engineering; School of Foreign Languages; School of Chemistry and Materials Engineering; School of Physics and Electronic Engineering; School of Marxism Studies; Continuing Education School; College of Teacher Education; Department of Physical Education; Library. The library consists of the Shaw Library on the East Lake campus and the Southeast campus library. The Shaw Library is located on Kuncheng Lake and has five floors. The building combines traditional Chinese architectural style with Western construction and incorporates Taoist water culture and Confucian mountain culture. In 2009, it ranked first in the Shaw Foundation's evaluation of university construction projects. The East Lake campus library has nine floors. The Southeast campus library is also located on Kuncheng Lake, across from the Shaw Library. The two libraries have 12 reading rooms. The library holds more than 1.2 million print books, 3 million e-books, 30,000 electronic journals, and 1,958 annual newspapers and periodicals. The staff totals 60: two hold doctoral degrees, three have master's degrees, and 38 have bachelor's degrees. The library has an organized, capable team with a sound structure. Adopting the tenet "reader first, service foremost," the library offers open-shelf access; each reading room is equipped with seating and digital information-retrieval stations. Opening hours are 8:15 to 21:30 daily, with adjusted hours during holidays. The library provides comprehensive information services, including book reservation, document delivery, and SDI (Selective Dissemination of Information). It also cooperates with related institutions to share resources. All operations are managed through integrated computer systems, achieving automation to support literature and information services. Each year the library organizes the "Scholarly Campus" reading month, which is highly regarded in Suzhou. Partner institutions include the University of San Francisco, the Belarusian State Academy of Arts, the University of Ulsan, Fachhochschule Nordhessen, and Tatung University. The school focuses on international communication and cooperation and seeks to increase its internationalization. Since its reform and opening up, CIT has strengthened academic exchanges with the outside world. It has sent experts and professors to countries including the United States, the United Kingdom, Japan, Germany, Singapore, Australia, Israel, and India to pursue advanced studies, and it regularly hosts foreign guests. CIT has established cooperative relationships with many overseas institutions and values its connection to the local economy and education. It cooperates with the Winchester School of Art at the University of Southampton on a "2+2" talent training program and with the German North Hessian Applied Technology University on mechanical and electronic engineering through a "3+1" program. After joining the "1+2+1" personnel training plan, CIT signed cooperative education agreements with six American universities. The Changshu Institute of Technology and the FOM University of Applied Sciences for Economics and Management signed an MSc Economics cooperation agreement. It also uses a "1+1" model for programs in three areas: accounting and finance, marketing, and marketing communication. All courses conform to UNESCO standards. Students who complete the program are awarded a Master of Arts degree recognized by the German government and other governments worldwide. The school promotes scientific research and faculty development. It currently has more than 1,100 staff members, including 600 with doctoral or master’s degrees, and has established research teams dedicated to scientific inquiry. The school has achieved numerous research outcomes: it has obtained 33 government-authorized patents and undertaken nearly 100 cross-disciplinary projects. Over the past five years, CIT faculty have published 57 monographs and 3,800 papers, about 600 of which were indexed in SCI, EI, or CPCI. CIT faculty currently lead more than 500 research projects at national, provincial, and municipal levels, including projects funded by the National Natural Science Foundation and agricultural technology programs. CIT also employs more than 50 foreign faculty from the United States, Canada, the United Kingdom, France, Japan, Denmark, and Finland, and hosts international visitors to promote academic exchange.	The Changshu Institute of Technology (CIT) is a provincial public university characterized by polytechnic education, applied technology, and teacher training. It is located in the center of the Yangtze River Delta in Changshu, a satellite city of Suzhou, Jiangsu, China, and is across a lake from Shajiabang. The university is divided into the East Lake Campus and the Southeast Campus. It grew out of Suzhou Normal College and officially opened on twelve May two thousand four. The school motto is "Breed integrity, seek truth—keep at it every day, and you'll go far along the way." History The Changshu Institute of Technology was established in August nineteen fifty eight and was originally known as Suzhou Normal Vocational College. It began recruiting graduates from teachers' colleges in nineteen seventy eight after the restoration of the national college entrance examination. In nineteen eighty nine, with the approval of the Jiangsu provincial government and the State Education Commission, it joined Changshu Vocational University. In early two thousand two CIT applied to offer undergraduate programs, and in two thousand four the Ministry of Education approved the establishment of the Changshu Institute of Technology. May twelve is celebrated as the university's founding day. Specialties CIT offers seven main specialties and provides forty four undergraduate programs, including programs in arts, history, education, law, engineering, and science. Specialties: Arts, History, Education, Law, Engineering, Science, Management. CIT has established many campuses and facilities. Schools: School of Humanities; School of Mechanical Engineering; School of Computer Science and Engineering; School of Electrical Engineering and Automation; School of Management; School of Biology and Food Engineering; School of Mathematics and Statistics; School of Arts and Garments Engineering; School of Foreign Languages; School of Chemistry and Materials Engineering; School of Physics and Electronic Engineering; School of Marxism Studies; Continuing Education School; College of Teacher Education; Department of Physical Education; Library. The library consists of the Shaw Library on the East Lake campus and the Southeast campus library. The Shaw Library is located on Kuncheng Lake and has five floors. The building combines traditional Chinese architectural style with Western construction and incorporates Taoist water culture and Confucian mountain culture. In two thousand nine, it ranked first in the Shaw Foundation's evaluation of university construction projects. The East Lake campus library has nine floors. The Southeast campus library is also located on Kuncheng Lake, across from the Shaw Library. The two libraries have twelve reading rooms. The library holds more than one point two million print books, three million e-books, thirty thousand electronic journals, and one thousand nine hundred fifty eight annual newspapers and periodicals. The staff totals sixty: two hold doctoral degrees, three have master's degrees, and thirty eight have bachelor's degrees. The library has an organized, capable team with a sound structure. Adopting the tenet "reader first, service foremost," the library offers open-shelf access; each reading room is equipped with seating and digital information-retrieval stations. Opening hours are eight colon fifteen to twenty one colon thirty daily, with adjusted hours during holidays. The library provides comprehensive information services, including book reservation, document delivery, and SDI (Selective Dissemination of Information). It also cooperates with related institutions to share resources. All operations are managed through integrated computer systems, achieving automation to support literature and information services. Each year the library organizes the "Scholarly Campus" reading month, which is highly regarded in Suzhou. Partner institutions include the University of San Francisco, the Belarusian State Academy of Arts, the University of Ulsan, Fachhochschule Nordhessen, and Tatung University. The school focuses on international communication and cooperation and seeks to increase its internationalization. Since its reform and opening up, CIT has strengthened academic exchanges with the outside world. It has sent experts and professors to countries including the United States, the United Kingdom, Japan, Germany, Singapore, Australia, Israel, and India to pursue advanced studies, and it regularly hosts foreign guests. CIT has established cooperative relationships with many overseas institutions and values its connection to the local economy and education. It cooperates with the Winchester School of Art at the University of Southampton on a "two plus two" talent training program and with the German North Hessian Applied Technology University on mechanical and electronic engineering through a "three plus one" program. After joining the "one plus two plus one" personnel training plan, CIT signed cooperative education agreements with six American universities. The Changshu Institute of Technology and the FOM University of Applied Sciences for Economics and Management signed an MSc Economics cooperation agreement. It also uses a "one plus one" model for programs in three areas: accounting and finance, marketing, and marketing communication. All courses conform to UNESCO standards. Students who complete the program are awarded a Master of Arts degree recognized by the German government and other governments worldwide. The school promotes scientific research and faculty development. It currently has more than one thousand one hundred staff members, including six hundred with doctoral or master’s degrees, and has established research teams dedicated to scientific inquiry. The school has achieved numerous research outcomes: it has obtained thirty three government-authorized patents and undertaken nearly one hundred cross-disciplinary projects. Over the past five years, CIT faculty have published fifty seven monographs and three thousand eight hundred papers, about six hundred of which were indexed in SCI, EI, or CPCI. CIT faculty currently lead more than five hundred research projects at national, provincial, and municipal levels, including projects funded by the National Natural Science Foundation and agricultural technology programs. CIT also employs more than fifty foreign faculty from the United States, Canada, the United Kingdom, France, Japan, Denmark, and Finland, and hosts international visitors to promote academic exchange.
long_en_287	wiki_en	963	en	In telecommunications, white spaces are radio frequencies allocated to a broadcasting service but not used locally. National and international bodies assign frequencies for specific uses and, in most cases, license the rights to broadcast over them. This frequency-allocation process creates bandplans that, for technical reasons, set aside white space between used radio bands or channels to avoid interference. In such cases, although the frequencies are unused, they have been specifically assigned for a purpose, such as a guard band. More commonly, these white spaces occur naturally between used channels, since assigning nearby transmissions to adjacent channels causes destructive interference. In addition to white space assigned for technical reasons, there is also unused radio spectrum that has either never been used or is becoming free due to technical changes. In particular, the switchover to digital television frees up large areas between about 50 MHz and 700 MHz, because digital transmissions can be packed into adjacent channels while analog ones cannot. This allows the band to be compressed into fewer channels while still supporting more transmissions. In the United States, the abandoned television frequencies are primarily in the upper UHF 700 MHz band, covering TV channels 52 to 69 (698 to 806 MHz). Television white spaces will continue to exist in UHF frequencies and in VHF frequencies, which require larger antennas for mobile users and white-space devices. In much of the world (except the U.S.), abandoned television channels have been in VHF, and the resulting large VHF white spaces are being reallocated for digital radio standards such as DAB, DAB+, and DMB. Various proposals, including IEEE 802.11af, IEEE 802.22, and proposals from the White Spaces Coalition, advocate using white spaces vacated by analog TV to provide wireless broadband Internet access. A device intended to use these channels is called a white-space device (WSD). WSDs are designed to detect unused portions of the spectrum and use them for Internet signals, potentially improving broadband and Wi-Fi availability in rural areas. Early proposals included using GNSS receivers and programming each WSD with a database of local TV stations, but that approach would not address non-stationary or unlicensed users or stations licensed or modified after the database was created. Reuse of these frequencies may also impact wireless microphones, medical telemetry, and other technologies that have historically relied on them. Professional wireless microphones have used TV white space for decades, long before the emergence of consumer "white space" devices. Like Wi‑Fi, TV white space is a wireless technology that uses radio frequency bands. TV white space operates in the 470 MHz to 698 MHz range, while Wi‑Fi operates in the 2.4 GHz and 5 GHz bands. Data rates depend on the radio model, vendor, antenna, and other factors; new TV white space radios can support more than 50 Mbit/s. Wi‑Fi data rates also vary with range and line of sight, and can reach up to 1,000 Mbit/s using the IEEE 802.11ac standard. Range is a crucial difference: TV white space typically covers about 6 miles on average, though this varies with noise, line of sight, and other conditions; some vendors, such as Carlson Wireless, advertise ranges up to 24.8 miles. Both technologies typically consume between 20 and 100 watts, depending on device, antenna, and vendor, and both can meet government security standards such as FIPS 197 (the Advanced Encryption Standard). Wi‑Fi performs well in urban environments, while TV white space is well suited for rural areas. Graph 1 illustrates the differences between TV white space and Wi‑Fi. By country Argentina Microsoft, in partnership with the communications authority of Argentina, Ente Nacional de Comunicaciones (ENACOM), planned to deliver wireless access to schools in the province of Mendoza in August 2017. Microsoft will loan the TV White Spaces hardware to ENACOM technicians, and national satellite operator ARSAT will act as the ISP. No further trial details have been released. Canada In August 2011, Industry Canada, the Canadian ministry for industry, launched a consultation titled "Consultation on a Policy and Technical Framework for the Use of Non‑Broadcasting Applications in the Television Broadcasting Bands Below 698 MHz." The consultation closed on November 4, 2011. Submissions were received from a wide range of organizations in the telecom and broadcast industries. Kenya A pilot project by Indigo Telecom/Microsoft and the Kenyan government is reportedly delivering bandwidth speeds of up to 16 Mbit/s to three rural communities that lack electricity — Male, Gakawa, and Laikipia — using a solar‑powered network. Namibia In Namibia, a pilot project called Citizen Connect, a collaboration between the Microsoft 4Afrika Initiative, the MyDigitalBridge Foundation, and the Millennium Challenge Account Namibia (MCA‑N), is slated to deliver broadband Internet to twenty-seven schools and seven circuit offices of the Ministry of Education in Omusati, Oshana, and Ohangwena using TV White Space technology. Philippines: In 2014, Microsoft worked with the Philippine government to pilot a program for digitizing the management of remote fishermen. Singapore: After the FCC, the Infocomm Media Development Authority became the second regulator in the world to regulate TV White Space, ahead of the UK and Canada. The Singapore efforts were driven mainly by the Singapore White Spaces Pilot Group (SWSPG), founded by the Institute for Infocomm Research, Microsoft, and StarHub. The Institute for Infocomm Research subsequently spun off Whizpace to commercialize TV White Space radio using intellectual property (IP) developed at the institute since 2006. South Africa: Google, in partnership with the Independent Communications Authority of South Africa (ICASA), the CSIR Meraka Institute, the Wireless Access Providers Association (WAPA), and Carlson Wireless, delivered wireless access to ten schools through three base stations on the campus of Stellenbosch University’s Faculty of Medicine and Health Sciences in Tygerberg, Cape Town.	In telecommunications, white spaces are radio frequencies allocated to a broadcasting service but not used locally. National and international bodies assign frequencies for specific uses and, in most cases, license the rights to broadcast over them. This frequency-allocation process creates bandplans that, for technical reasons, set aside white space between used radio bands or channels to avoid interference. In such cases, although the frequencies are unused, they have been specifically assigned for a purpose, such as a guard band. More commonly, these white spaces occur naturally between used channels, since assigning nearby transmissions to adjacent channels causes destructive interference. In addition to white space assigned for technical reasons, there is also unused radio spectrum that has either never been used or is becoming free due to technical changes. In particular, the switchover to digital television frees up large areas between about fifty megahertz and seven hundred megahertz, because digital transmissions can be packed into adjacent channels while analog ones cannot. This allows the band to be compressed into fewer channels while still supporting more transmissions. In the United States, the abandoned television frequencies are primarily in the upper UHF seven hundred megahertz band, covering TV channels fifty two to sixty nine (six hundred ninety eight to eight hundred six megahertz). Television white spaces will continue to exist in UHF frequencies and in VHF frequencies, which require larger antennas for mobile users and white-space devices. In much of the world (except the U.S.), abandoned television channels have been in VHF, and the resulting large VHF white spaces are being reallocated for digital radio standards such as DAB, DAB+, and DMB. Various proposals, including IEEE eight zero two point one one af, IEEE eight zero two point two two, and proposals from the White Spaces Coalition, advocate using white spaces vacated by analog TV to provide wireless broadband Internet access. A device intended to use these channels is called a white-space device (WSD). WSDs are designed to detect unused portions of the spectrum and use them for Internet signals, potentially improving broadband and Wi‑Fi availability in rural areas. Early proposals included using GNSS receivers and programming each WSD with a database of local TV stations, but that approach would not address non-stationary or unlicensed users or stations licensed or modified after the database was created. Reuse of these frequencies may also impact wireless microphones, medical telemetry, and other technologies that have historically relied on them. Professional wireless microphones have used TV white space for decades, long before the emergence of consumer "white space" devices. Like Wi‑Fi, TV white space is a wireless technology that uses radio frequency bands. TV white space operates in the four hundred seventy MHz to six hundred ninety-eight MHz range, while Wi‑Fi operates in the two point four GHz and five GHz bands. Data rates depend on the radio model, vendor, antenna, and other factors; new TV white space radios can support more than fifty Mbit per second. Wi‑Fi data rates also vary with range and line of sight, and can reach up to one thousand Mbit per second using the IEEE eight zero two point one one ac standard. Range is a crucial difference: TV white space typically covers about six miles on average, though this varies with noise, line of sight, and other conditions; some vendors, such as Carlson Wireless, advertise ranges up to twenty four point eight miles. Both technologies typically consume between twenty and one hundred watts, depending on device, antenna, and vendor, and both can meet government security standards such as FIPS one hundred ninety seven (the Advanced Encryption Standard). Wi‑Fi performs well in urban environments, while TV white space is well suited for rural areas. Graph one illustrates the differences between TV white space and Wi‑Fi. By country Argentina Microsoft, in partnership with the communications authority of Argentina, Ente Nacional de Comunicaciones (ENACOM), planned to deliver wireless access to schools in the province of Mendoza in August two thousand seventeen. Microsoft will loan the TV White Spaces hardware to ENACOM technicians, and national satellite operator ARSAT will act as the ISP. No further trial details have been released. Canada In August two thousand eleven, Industry Canada, the Canadian ministry for industry, launched a consultation titled "Consultation on a Policy and Technical Framework for the Use of Non‑Broadcasting Applications in the Television Broadcasting Bands Below six hundred ninety eight MHz." The consultation closed on November fourth, two thousand eleven. Submissions were received from a wide range of organizations in the telecom and broadcast industries. Kenya A pilot project by Indigo Telecom/Microsoft and the Kenyan government is reportedly delivering bandwidth speeds of up to sixteen Mbit/s to three rural communities that lack electricity — Male, Gakawa, and Laikipia — using a solar‑powered network. Namibia In Namibia, a pilot project called Citizen Connect, a collaboration between the Microsoft 4Afrika Initiative, the MyDigitalBridge Foundation, and the Millennium Challenge Account Namibia (MCA‑N), is slated to deliver broadband Internet to twenty-seven schools and seven circuit offices of the Ministry of Education in Omusati, Oshana, and Ohangwena using TV White Space technology. Philippines: In two thousand fourteen, Microsoft worked with the Philippine government to pilot a program for digitizing the management of remote fishermen. Singapore: After the FCC, the Infocomm Media Development Authority became the second regulator in the world to regulate TV White Space, ahead of the UK and Canada. The Singapore efforts were driven mainly by the Singapore White Spaces Pilot Group (SWSPG), founded by the Institute for Infocomm Research, Microsoft, and StarHub. The Institute for Infocomm Research subsequently spun off Whizpace to commercialize TV White Space radio using intellectual property (IP) developed at the institute since two thousand six. South Africa: Google, in partnership with the Independent Communications Authority of South Africa (ICASA), the CSIR Meraka Institute, the Wireless Access Providers Association (WAPA), and Carlson Wireless, delivered wireless access to ten schools through three base stations on the campus of Stellenbosch University’s Faculty of Medicine and Health Sciences in Tygerberg, Cape Town.
long_en_368	poet_en	886	en	I'm a simple girl who likes simple things. I'm just trying to finally get my life in order. I don't write to give advice; I kind of just talk my way through situations that happen to me. If I can help someone else along the way, that's good too. I don't profess to have all the answers; as a matter of fact, I mess up quite often. But I know that God loves me and wants nothing but the best for me and those around me. I hold on to that every day. I'm not looking for kudos; I'm just trying to make it one day at a time. My sons came home the other day in a panic. It seemed their cousins had gotten them all upset about strange things that had happened in their homes. They told my boys that chairs had moved by themselves, that they'd seen ghosts fly past windows, and that they heard things at night. They told the boys that demons were there and came out at night to get them. That led to a lot of questions and, in the end, the boys slept with me that night. I explained as best I could, because I do believe there are evil things in this world. I don't necessarily think about them the same way others do, but I do believe in demons. I talked to Greg, my minister, and asked if he would talk to the boys for me and explain; of course he agreed. He said he would tell them about Ephesians 6. No matter what is going on in the world, the way to fight Satan is by putting on the armor of God that will protect you against anything that could possibly happen. Yesterday was not bad, but it was a hard day. It seems that the armor he told me about Thursday morning, I needed to put on by Thursday night. I have said it before and I will say it again: Satan is very real and very much alive, and if he sees you struggling with one thing, he brings the rest of it on to pull you farther down than you started. But as much as Satan is real, so is God, and as the day went on I just prayed for some sense of calm. I am a person who really believes in signs. I ask God point-blank, "Okay, if this is it, then show me. If it's not, send me a sign." I don't know if He actually does it or if it's just my mind making it happen, but I would really like to believe God is answering me to help me. So yesterday, as usual, that is what I did. I said I needed a sense of calm, that I needed to know this was going to be okay, and I would appreciate it if He would let me know today. Hmmm, I got that last night. I was reminded that even when I feel overwhelmed. My daughter wants to be a lawyer/WNBA basketball star. My middle child wants to be an artist, and the baby, well, he has a lot of things he wants to do. I believe that with a whole lot of work and putting their minds to it, my children can do and be anything they want. They have the God-given right to make those choices that affect their lives. My children have dreams, but I also think God has plans for them that may or may not include what they have in mind. We are all called to be like Christ; no matter how many great things we may do that impress people here, God may not be as impressed if we are not following His plan. It took me nearly my whole life, but I finally realized that God has a specific purpose for me, as He does for everyone. It wasn't something that just hit me—I haven't always known, and honestly I'm not completely sure I understand it all, but I know it's there. God didn't knock me over the head; there were no neon signs. Actually, it wasn't until I started asking sincerely that I began hearing what He was trying to tell me. It wasn't on my timetable; it was all on His. I am an impatient person—I make no bones about that—and I still don't like waiting. But for God's plan to work as He intended, I had to be ready. What good would it have done if my mind and heart were not in the right place? I also had to realize (and we all do) that if God chose me to do something, it is going to be done right. He is not going to give you something He didn't think you could handle. A perfect example is Paul. Who would have thought that a man who persecuted Christians would become one of the greatest teachers of the gospel of all time? Romans 8:28: We know that in all things God works for the good of those who love Him, who have been called according to His purpose.	I'm a simple girl who likes simple things. I'm just trying to finally get my life in order. I don't write to give advice; I kind of just talk my way through situations that happen to me. If I can help someone else along the way, that's good too. I don't profess to have all the answers; as a matter of fact, I mess up quite often. But I know that God loves me and wants nothing but the best for me and those around me. I hold on to that every day. I'm not looking for kudos; I'm just trying to make it one day at a time. My sons came home the other day in a panic. It seemed their cousins had gotten them all upset about strange things that had happened in their homes. They told my boys that chairs had moved by themselves, that they'd seen ghosts fly past windows, and that they heard things at night. They told the boys that demons were there and came out at night to get them. That led to a lot of questions and, in the end, the boys slept with me that night. I explained as best I could, because I do believe there are evil things in this world. I don't necessarily think about them the same way others do, but I do believe in demons. I talked to Greg, my minister, and asked if he would talk to the boys for me and explain; of course he agreed. He said he would tell them about Ephesians six. No matter what is going on in the world, the way to fight Satan is by putting on the armor of God that will protect you against anything that could possibly happen. Yesterday was not bad, but it was a hard day. It seems that the armor he told me about Thursday morning, I needed to put on by Thursday night. I have said it before and I will say it again: Satan is very real and very much alive, and if he sees you struggling with one thing, he brings the rest of it on to pull you farther down than you started. But as much as Satan is real, so is God, and as the day went on I just prayed for some sense of calm. I am a person who really believes in signs. I ask God point-blank, "Okay, if this is it, then show me. If it's not, send me a sign." I don't know if He actually does it or if it's just my mind making it happen, but I would really like to believe God is answering me to help me. So yesterday, as usual, that is what I did. I said I needed a sense of calm, that I needed to know this was going to be okay, and I would appreciate it if He would let me know today. Hmmm, I got that last night. I was reminded that even when I feel overwhelmed. My daughter wants to be a lawyer/WNBA basketball star. My middle child wants to be an artist, and the baby, well, he has a lot of things he wants to do. I believe that with a whole lot of work and putting their minds to it, my children can do and be anything they want. They have the God-given right to make those choices that affect their lives. My children have dreams, but I also think God has plans for them that may or may not include what they have in mind. We are all called to be like Christ; no matter how many great things we may do that impress people here, God may not be as impressed if we are not following His plan. It took me nearly my whole life, but I finally realized that God has a specific purpose for me, as He does for everyone. It wasn't something that just hit me—I haven't always known, and honestly I'm not completely sure I understand it all, but I know it's there. God didn't knock me over the head; there were no neon signs. Actually, it wasn't until I started asking sincerely that I began hearing what He was trying to tell me. It wasn't on my timetable; it was all on His. I am an impatient person—I make no bones about that—and I still don't like waiting. But for God's plan to work as He intended, I had to be ready. What good would it have done if my mind and heart were not in the right place? I also had to realize (and we all do) that if God chose me to do something, it is going to be done right. He is not going to give you something He didn't think you could handle. A perfect example is Paul. Who would have thought that a man who persecuted Christians would become one of the greatest teachers of the gospel of all time? Romans eight:twenty eight: We know that in all things God works for the good of those who love Him, who have been called according to His purpose.
long_en_166	paper_en	1,916	en	DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrates remarkable reasoning capabilities. Through RL, DeepSeek-R1-Zero naturally emerges with numerous powerful and intriguing reasoning behaviors. However, it encounters challenges such as poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates multi-stage training and cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1-1217 on reasoning tasks. To support the research community, we open-source DeepSeek-R1-Zero, DeepSeek-R1, and six dense models (1.5B, 7B, 8B, 14B, 32B, 70B) distilled from DeepSeek-R1 based on Qwen and Llama. Section 1: Introduction In recent years, Large Language Models (LLMs) have been undergoing rapid iteration and evolution, progressively diminishing the gap towards Artificial General Intelligence (AGI). Recently, post-training has emerged as an important component of the full training pipeline. It has been shown to enhance accuracy on reasoning tasks, align with social values, and adapt to user preferences, all while requiring relatively minimal computational resources against pre-training. In the context of reasoning capabilities, OpenAI's o1 series models were the first to introduce inference-time scaling by increasing the length of the Chain-of-Thought reasoning process. This approach has achieved significant improvements in various reasoning tasks, such as mathematics, coding, and scientific reasoning. However, the challenge of effective test-time scaling remains an open question for the research community. Several prior works have explored various approaches, including process-based reward models, reinforcement learning, and search algorithms such as Monte Carlo Tree Search and Beam Search. However, none of these methods has achieved general reasoning performance comparable to OpenAI's o1 series models. In this paper, we take the first step toward improving language model reasoning capabilities using pure reinforcement learning (RL). Our goal is to explore the potential of LLMs to develop reasoning capabilities without any supervised data, focusing on their self-evolution through a pure RL process. Specifically, we use DeepSeek-V3-Base as the base model and employ GRPO as the RL framework to improve model performance in reasoning. During training, DeepSeek-R1-Zero naturally emerged with numerous powerful and interesting reasoning behaviors. After thousands of RL steps, DeepSeek-R1-Zero exhibits super performance on reasoning benchmarks. For instance, the pass at 1 score on AIME 2024 increases from 15.6% to 71.0%, and with majority voting, the score further improves to 86.7%, matching the performance of OpenAI-o1-0912. However, DeepSeek-R1-Zero encounters challenges such as poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates a small amount of cold-start data and a multi-stage training pipeline. Specifically, we begin by collecting thousands of cold-start data to fine-tune the DeepSeek-V3-Base model. Following this, we perform reasoning-oriented RL like DeepSeek-R1-Zero. Upon nearing convergence in the RL process, we create new SFT data through rejection sampling on the RL checkpoint, combined with supervised data from DeepSeek-V3 in domains such as writing, factual QA, and self-cognition, and then retrain the DeepSeek-V3-Base model. After fine-tuning with the new data, the checkpoint undergoes an additional RL process, taking into account prompts from all scenarios. After these steps, we obtained a checkpoint referred to as DeepSeek-R1, which achieves performance on par with OpenAI-o1-1217. We further explore distillation from DeepSeek-R1 to smaller dense models. Using Qwen2.5-32B as the base model, direct distillation from DeepSeek-R1 outperforms applying RL on it. This demonstrates that the reasoning patterns discovered by larger base models are crucial for improving reasoning capabilities. We open-source the distilled Qwen and Llama series. Notably, our distilled 14B model outperforms state-of-the-art open-source QwQ-32B-Preview by a large margin, and the distilled 32B and 70B models set a new record on the reasoning benchmarks among dense models. Contributions Post-Training: Large-Scale Reinforcement Learning on the Base Model We directly apply RL to the base model without relying on supervised fine-tuning (SFT) as a preliminary step. This approach allows the model to explore chain-of-thought (CoT) for solving complex problems, resulting in the development of DeepSeek-R1-Zero. DeepSeek-R1-Zero demonstrates capabilities such as self-verification, reflection, and generating long CoTs, marking a significant milestone for the research community. Notably, it is the first open research to validate that reasoning capabilities of LLMs can be incentivized purely through RL, without the need for SFT. This breakthrough paves the way for future advancements in this area. We introduce our pipeline to develop DeepSeek-R1. The pipeline incorporates two RL stages aimed at discovering improved reasoning patterns and aligning with human preferences, as well as two SFT stages that serve as the seed for the model's reasoning and non-reasoning capabilities. We believe the pipeline will benefit the industry by creating better models. Distillation: Smaller Models Can Be Powerful Too We demonstrate that the reasoning patterns of larger models can be distilled into smaller models, resulting in better performance compared to the reasoning patterns discovered through RL on small models. The open source DeepSeek-R1, as well as its API, will benefit the research community to distill better smaller models in the future. Using the reasoning data generated by DeepSeek-R1, we fine-tuned several dense models that are widely used in the research community. The evaluation results demonstrate that the distilled smaller dense models perform exceptionally well on benchmarks. DeepSeek-R1-Distill-Qwen-7B achieves 55.5% on AIME 2024, surpassing QwQ-32B-Preview. Additionally, DeepSeek-R1-Distill-Qwen-32B scores 72.6% on AIME 2024, 94.3% on MATH-500, and 57.2% on LiveCodeBench. These results significantly outperform previous open-source models and are comparable to o1-mini. We open-source distilled 1.5B, 7B, 8B, 14B, 32B, and 70B checkpoints based on Qwen2.5 and Llama3 series to the community. Summary of Evaluation Results Reasoning tasks: (1) DeepSeek-R1 achieves a score of 79.8% Pass at 1 on AIME 2024, slightly surpassing OpenAI-o1-1217. On MATH-500, it attains an impressive score of 97.3%, performing on par with OpenAI-o1-1217 and significantly outperforming other models. (2) On coding-related tasks, DeepSeek-R1 demonstrates expert level in code competition tasks, as it achieves 2,029 Elo rating on Codeforces outperforming 96.3% human participants in the competition. For engineering-related tasks, DeepSeek-R1 performs slightly better than DeepSeek-V3, which could help developers in real world tasks. Knowledge: On benchmarks such as MMLU, MMLU-Pro, and GPQA Diamond, DeepSeek-R1 achieves outstanding results, significantly outperforming DeepSeek-V3 with scores of 90.8% on MMLU, 84.0% on MMLU-Pro, and 71.5% on GPQA Diamond. While its performance is slightly below that of OpenAI-o1-1217 on these benchmarks, DeepSeek-R1 surpasses other closed-source models, demonstrating its competitive edge in educational tasks. On the factual benchmark SimpleQA, DeepSeek-R1 outperforms DeepSeek-V3, demonstrating its capability in handling fact-based queries. A similar trend is observed where OpenAI-o1 surpasses 4o on this benchmark. Others: DeepSeek-R1 also excels in a wide range of tasks, including creative writing, general question answering, editing, summarization, and more. It achieves an impressive length-controlled win-rate of 87.6% on AlpacaEval 2.0 and a win-rate of 92.3% on ArenaHard, showcasing its strong ability to intelligently handle non-exam-oriented queries. Additionally, DeepSeek-R1 demonstrates outstanding performance on tasks requiring long-context understanding, substantially outperforming DeepSeek-V3 on long-context benchmarks. Section 2: Approach Overview Previous work has heavily relied on large amounts of supervised data to enhance model performance. In this study, we demonstrate that reasoning capabilities can be significantly improved through large-scale reinforcement learning (RL), even without using supervised fine-tuning (SFT) as a cold start. Furthermore, performance can be further enhanced with the inclusion of a small amount of cold-start data. In the following sections, we present: (1) DeepSeek-R1-Zero, which applies RL directly to the base model without any SFT data, and (2) DeepSeek-R1, which applies RL starting from a checkpoint fine-tuned with thousands of long Chain-of-Thought (CoT) examples. 3) Distill the reasoning capability from DeepSeek-R1 to small dense models. DeepSeek-R1-Zero: Reinforcement Learning on the Base Model Reinforcement learning has demonstrated significant effectiveness in reasoning tasks, as evidenced by our previous works. However, these works heavily depended on supervised data, which are time-intensive to gather. In this section, we explore the potential of LLMs to develop reasoning capabilities without any supervised data, focusing on their self-evolution through a pure reinforcement learning process. We start with a brief overview of our RL algorithm, followed by the presentation of some exciting results, and hope this provides the community with valuable insights. Reinforcement Learning Algorithm Group Relative Policy Optimization In order to save the training costs of RL, we adopt Group Relative Policy Optimization (GRPO), which foregoes the critic model that is typically the same size as the policy model, and estimates the baseline from group scores instead. Specifically, for each question, GRPO samples a group of outputs from the old policy and then optimizes the policy model by maximizing a specific objective. The advantage is computed using a group of rewards corresponding to the outputs within each group. Reward Modeling The reward is the source of the training signal, which decides the optimization direction of RL. To train DeepSeek-R1-Zero, we adopt a rule-based reward system that mainly consists of two types of rewards: Accuracy rewards: The accuracy reward model evaluates whether the response is correct. For example, in the case of math problems with deterministic results, the model is required to provide the final answer in a specified format (e.g., within a box), enabling reliable rule-based verification of correctness. Similarly, for LeetCode problems, a compiler can be used to generate feedback based on predefined test cases. Format rewards: In addition to the accuracy reward model, we employ a format reward model that enforces the model to put its thinking process between ‘<think>’ and ‘</think>’ tags. We do not apply the outcome or process neural reward model in developing DeepSeek-R1-Zero, because we find that the neural reward model may suffer from reward hacking in the large-scale reinforcement learning process, and retraining the reward model needs additional training resources and it complicates the whole training pipeline. Training Template To train DeepSeek-R1-Zero, we begin by designing a straightforward template that guides the base model to adhere to our specified instructions. This template requires DeepSeek-R1-Zero to first produce a reasoning process, followed by the final answer. We intentionally limit our constraints to this structural format, avoiding any content-specific biases--such as mandating reflective reasoning or promoting particular problem-solving strategies--to ensure that we can accurately observe the model's natural progression during the RL process. Performance, Self-evolution Process and Aha Moment of DeepSeek-R1-Zero Performance of DeepSeek-R1-Zero The performance trajectory of DeepSeek-R1-Zero on the AIME 2024 benchmark throughout the RL training process shows a steady and consistent enhancement.	DeepSeek-R one: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning We introduce our first-generation reasoning models, DeepSeek-R one-Zero and DeepSeek-R one. DeepSeek-R one-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrates remarkable reasoning capabilities. Through RL, DeepSeek-R one-Zero naturally emerges with numerous powerful and intriguing reasoning behaviors. However, it encounters challenges such as poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R one, which incorporates multi-stage training and cold-start data before RL. DeepSeek-R one achieves performance comparable to OpenAI-o one-one thousand two hundred seventeen on reasoning tasks. To support the research community, we open-source DeepSeek-R one-Zero, DeepSeek-R one, and six dense models (one point five B, seven B, eight B, fourteen B, thirty two B, seventy B) distilled from DeepSeek-R one based on Qwen and Llama. Section one: Introduction In recent years, Large Language Models (LLMs) have been undergoing rapid iteration and evolution, progressively diminishing the gap towards Artificial General Intelligence (AGI). Recently, post-training has emerged as an important component of the full training pipeline. It has been shown to enhance accuracy on reasoning tasks, align with social values, and adapt to user preferences, all while requiring relatively minimal computational resources against pre-training. In the context of reasoning capabilities, OpenAI's o one series models were the first to introduce inference-time scaling by increasing the length of the Chain-of-Thought reasoning process. This approach has achieved significant improvements in various reasoning tasks, such as mathematics, coding, and scientific reasoning. However, the challenge of effective test-time scaling remains an open question for the research community. Several prior works have explored various approaches, including process-based reward models, reinforcement learning, and search algorithms such as Monte Carlo Tree Search and Beam Search. However, none of these methods has achieved general reasoning performance comparable to OpenAI's o one series models. In this paper, we take the first step toward improving language model reasoning capabilities using pure reinforcement learning (RL). Our goal is to explore the potential of LLMs to develop reasoning capabilities without any supervised data, focusing on their self-evolution through a pure RL process. Specifically, we use DeepSeek V three Base as the base model and employ GRPO as the RL framework to improve model performance in reasoning. During training, DeepSeek R one Zero naturally emerged with numerous powerful and interesting reasoning behaviors. After thousands of RL steps, DeepSeek R one Zero exhibits super performance on reasoning benchmarks. For instance, the pass at one score on AIME two thousand twenty four increases from fifteen point six percent to seventy one point zero percent, and with majority voting, the score further improves to eighty six point seven percent, matching the performance of OpenAI o one zero nine one two. However, DeepSeek R one Zero encounters challenges such as poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek R one, which incorporates a small amount of cold start data and a multi stage training pipeline. Specifically, we begin by collecting thousands of cold start data to fine tune the DeepSeek V three Base model. Following this, we perform reasoning oriented RL like DeepSeek R one Zero. Upon nearing convergence in the RL process, we create new SFT data through rejection sampling on the RL checkpoint, combined with supervised data from DeepSeek V three in domains such as writing, factual QA, and self cognition, and then retrain the DeepSeek V three Base model. After fine tuning with the new data, the checkpoint undergoes an additional RL process, taking into account prompts from all scenarios. After these steps, we obtained a checkpoint referred to as DeepSeek R one, which achieves performance on par with OpenAI o one one two one seven. We further explore distillation from DeepSeek R one to smaller dense models. Using Qwen two point five thirty two B as the base model, direct distillation from DeepSeek R one outperforms applying RL on it. This demonstrates that the reasoning patterns discovered by larger base models are crucial for improving reasoning capabilities. We open-source the distilled Qwen and Llama series. Notably, our distilled fourteen B model outperforms state-of-the-art open-source QwQ-thirty-two B-Preview by a large margin, and the distilled thirty-two B and seventy B models set a new record on the reasoning benchmarks among dense models. Contributions Post-Training: Large-Scale Reinforcement Learning on the Base Model We directly apply RL to the base model without relying on supervised fine-tuning (SFT) as a preliminary step. This approach allows the model to explore chain-of-thought (CoT) for solving complex problems, resulting in the development of DeepSeek-R one-Zero. DeepSeek-R one-Zero demonstrates capabilities such as self-verification, reflection, and generating long CoTs, marking a significant milestone for the research community. Notably, it is the first open research to validate that reasoning capabilities of LLMs can be incentivized purely through RL, without the need for SFT. This breakthrough paves the way for future advancements in this area. We introduce our pipeline to develop DeepSeek-R one. The pipeline incorporates two RL stages aimed at discovering improved reasoning patterns and aligning with human preferences, as well as two SFT stages that serve as the seed for the model's reasoning and non-reasoning capabilities. We believe the pipeline will benefit the industry by creating better models. Distillation: Smaller Models Can Be Powerful Too We demonstrate that the reasoning patterns of larger models can be distilled into smaller models, resulting in better performance compared to the reasoning patterns discovered through RL on small models. The open source DeepSeek R one, as well as its API, will benefit the research community to distill better smaller models in the future. Using the reasoning data generated by DeepSeek R one, we fine-tuned several dense models that are widely used in the research community. The evaluation results demonstrate that the distilled smaller dense models perform exceptionally well on benchmarks. DeepSeek R one Distill Qwen seven B achieves fifty five point five percent on AIME two thousand twenty four, surpassing QwQ thirty two B Preview. Additionally, DeepSeek R one Distill Qwen thirty two B scores seventy two point six percent on AIME two thousand twenty four, ninety four point three percent on MATH five hundred, and fifty seven point two percent on LiveCodeBench. These results significantly outperform previous open-source models and are comparable to o one mini. We open-source distilled one point five B, seven B, eight B, fourteen B, thirty two B, and seventy B checkpoints based on Qwen two point five and Llama three series to the community. Summary of Evaluation Results Reasoning tasks: (one) DeepSeek R one achieves a score of seventy nine point eight percent Pass at one on AIME two thousand twenty four, slightly surpassing OpenAI o one one two one seven. On MATH five hundred, it attains an impressive score of ninety seven point three percent, performing on par with OpenAI o one one two one seven and significantly outperforming other models. (two) On coding-related tasks, DeepSeek R one demonstrates expert level in code competition tasks, as it achieves two thousand twenty-nine Elo rating on Codeforces outperforming ninety-six point three percent human participants in the competition. For engineering-related tasks, DeepSeek R one performs slightly better than DeepSeek V three, which could help developers in real world tasks. Knowledge: On benchmarks such as MMLU, MMLU-Pro, and GPQA Diamond, DeepSeek R one achieves outstanding results, significantly outperforming DeepSeek V three with scores of ninety point eight percent on MMLU, eighty-four point zero percent on MMLU-Pro, and seventy-one point five percent on GPQA Diamond. While its performance is slightly below that of OpenAI o one one two one seven on these benchmarks, DeepSeek R one surpasses other closed-source models, demonstrating its competitive edge in educational tasks. On the factual benchmark SimpleQA, DeepSeek R one outperforms DeepSeek V three, demonstrating its capability in handling fact-based queries. A similar trend is observed where OpenAI o one surpasses four o on this benchmark. Others: DeepSeek R one also excels in a wide range of tasks, including creative writing, general question answering, editing, summarization, and more. It achieves an impressive length-controlled win-rate of eighty-seven point six percent on AlpacaEval two point zero and a win-rate of ninety-two point three percent on ArenaHard, showcasing its strong ability to intelligently handle non-exam-oriented queries. Additionally, DeepSeek R one demonstrates outstanding performance on tasks requiring long-context understanding, substantially outperforming DeepSeek V three on long-context benchmarks. Section two: Approach Overview Previous work has heavily relied on large amounts of supervised data to enhance model performance. In this study, we demonstrate that reasoning capabilities can be significantly improved through large-scale reinforcement learning (RL), even without using supervised fine-tuning (SFT) as a cold start. Furthermore, performance can be further enhanced with the inclusion of a small amount of cold-start data. In the following sections, we present: (one) DeepSeek-R one-Zero, which applies RL directly to the base model without any SFT data, and (two) DeepSeek-R one, which applies RL starting from a checkpoint fine-tuned with thousands of long Chain-of-Thought (CoT) examples. three) Distill the reasoning capability from DeepSeek-R one to small dense models. DeepSeek-R one-Zero: Reinforcement Learning on the Base Model Reinforcement learning has demonstrated significant effectiveness in reasoning tasks, as evidenced by our previous works. However, these works heavily depended on supervised data, which are time-intensive to gather. In this section, we explore the potential of LLMs to develop reasoning capabilities without any supervised data, focusing on their self-evolution through a pure reinforcement learning process. We start with a brief overview of our RL algorithm, followed by the presentation of some exciting results, and hope this provides the community with valuable insights. Reinforcement Learning Algorithm Group Relative Policy Optimization In order to save the training costs of RL, we adopt Group Relative Policy Optimization (GRPO), which foregoes the critic model that is typically the same size as the policy model, and estimates the baseline from group scores instead. Specifically, for each question, GRPO samples a group of outputs from the old policy and then optimizes the policy model by maximizing a specific objective. The advantage is computed using a group of rewards corresponding to the outputs within each group. Reward Modeling The reward is the source of the training signal, which decides the optimization direction of RL. To train DeepSeek-R one-Zero, we adopt a rule-based reward system that mainly consists of two types of rewards: Accuracy rewards: The accuracy reward model evaluates whether the response is correct. For example, in the case of math problems with deterministic results, the model is required to provide the final answer in a specified format (e.g., within a box), enabling reliable rule-based verification of correctness. Similarly, for LeetCode problems, a compiler can be used to generate feedback based on predefined test cases. Format rewards: In addition to the accuracy reward model, we employ a format reward model that enforces the model to put its thinking process between ‘<think>’ and ‘</think>’ tags. We do not apply the outcome or process neural reward model in developing DeepSeek R one Zero, because we find that the neural reward model may suffer from reward hacking in the large-scale reinforcement learning process, and retraining the reward model needs additional training resources and it complicates the whole training pipeline. Training Template To train DeepSeek R one Zero, we begin by designing a straightforward template that guides the base model to adhere to our specified instructions. This template requires DeepSeek R one Zero to first produce a reasoning process, followed by the final answer. We intentionally limit our constraints to this structural format, avoiding any content-specific biases--such as mandating reflective reasoning or promoting particular problem-solving strategies--to ensure that we can accurately observe the model's natural progression during the RL process. Performance, Self-evolution Process and Aha Moment of DeepSeek R one Zero Performance of DeepSeek R one Zero The performance trajectory of DeepSeek R one Zero on the AIME two thousand twenty four benchmark throughout the RL training process shows a steady and consistent enhancement.
long_en_337	poet_en	1,020	en	You can read part 1 of my story here. The night after we went to the lake, I spent the night at Lynne's house and we watched The Bachelorette — I'm totally hooked on this season, by the way! We had a nice, relaxing night, which was a great way to end the trip. The next morning I stopped by Lynne's work to say goodbye and then drove to Front Royal, about 2.5 hours north. When I got there, I got Izzie all situated, and Torrie (my aunt) and I decided to go to lunch and run some errands. As we were leaving her road, we saw a UPS truck. I had decided to have my new phone sent to Torrie's house so I could have it before I drove back to New York, so we turned around and followed the UPS truck. He drove straight to her house, where I met him and he handed me my new phone. It was quite exciting! We had some lovely Mexican food for lunch. I love Mexican food, and there isn't any good Mexican food here, but this place was delicious. When Torrie went to pay, she realized she'd left her wallet at the house, so we went back home. We hung out with the dogs a little and did some things around her house. Then we left again to go to Target and get pedicures. After that, we decided to spend some time at the house with the dogs. Torrie and I talked about our weight-loss journeys; she had just started going for runs and walks every night. So we decided to take the dogs and the wine we had bought at Target. My uncle and she live on five acres of land right on a river, so it's perfect for the dogs to run around. We ran and walked for 30 minutes, and it felt really good! We ran by the river, in the water, and it was a great bonding experience. We set up some little chairs right by the river and even lit some tiki torches. My uncle came out and we just hung out and drank quite a bit. I was in the middle of a super interesting story when I heard a super loud crack! I jumped up out of my chair and so did my uncle. I looked behind me and a huge tree log had fallen from the tree. It landed five feet from where we were sitting and two feet from the cars. Somehow it didn't hit us, the cars, or the dogs! When I turned around, Torrie was still sitting in her chair; apparently my story was so interesting she didn't really react to the huge cracking sound. I looked for my uncle and he was in a ball about 20 feet away from us. I just started laughing and asked how he got there so quickly. He said he thought the whole tree was falling and just ran while looking back at the tree and tripped over a rock! He was fine, and we were just lucky that all of us and the dogs were okay. I left pretty early the next day to head back to NY. I had so much fun in VA, but I wish I had stayed longer — the time went by so fast and I feel like I spent most of it in the car. I can't wait to go back. I'm going back in three weeks when my uncle turns 40. They're having a huge party with a band by the river, but they're making sure all the dead tree limbs are gone first. Y'all read about my trip there — it took forever, but I made it safe and sound. I got into Lynchburg around 5:30 p.m. and raced to my sister's house. Basically I dropped Izzie off, said a quick hello to friends from Canada, and then bolted because I had dinner plans at 5:30 p.m. I was a little late but met Laurel at Buffalo Wild Wings, where I finally satisfied my recent craving for fried pickles. I'm not sure what that was about or where it came from. After that I went home and went to bed; I was exhausted from my drive. Lynne came home from her rehearsal dinner and we hung out for a little bit before I called it a night. The next day Lynne and I went to see my grandma, who had just had shoulder surgery, and it was good to see her. After that we went shopping, which was amazing. We took a little too long shopping, and when we got home we had an hour to get ready for the wedding. You know an hour is not long enough to get ready! Lynne only has one shower, so I opted to wash my hair in the sink — a girl has to do what a girl has to do. We managed to get there in time, by the grace of God, because I'm not sure how we did. The wedding was beautiful; just look at the pictures. Megan and Daniel are a great couple and they have an awesome relationship. The next day Justin, my brother-in-law, was working at the lake, and Lynne, some friends, and I drove out there to hang out. Another friend—are you following?—had gotten a boat with his parents and invited us to go tubing. Uhm, heck yeah (not the tubing part). We hung out on the boat, and Lynne was a beast out there on the tube; the driver definitely didn't hold back because she was a girl. It was all fun and games until we hit a tidal wave that struck my bag, which was canvas and not quite waterproof. There was a lot of water, and for some stupid reason I had brought my phone on the boat; I'm not sure why, because I didn't use it or need it.	You can read part one of my story here. The night after we went to the lake, I spent the night at Lynne's house and we watched The Bachelorette — I'm totally hooked on this season, by the way! We had a nice, relaxing night, which was a great way to end the trip. The next morning I stopped by Lynne's work to say goodbye and then drove to Front Royal, about two point five hours north. When I got there, I got Izzie all situated, and Torrie (my aunt) and I decided to go to lunch and run some errands. As we were leaving her road, we saw a UPS truck. I had decided to have my new phone sent to Torrie's house so I could have it before I drove back to New York, so we turned around and followed the UPS truck. He drove straight to her house, where I met him and he handed me my new phone. It was quite exciting! We had some lovely Mexican food for lunch. I love Mexican food, and there isn't any good Mexican food here, but this place was delicious. When Torrie went to pay, she realized she'd left her wallet at the house, so we went back home. We hung out with the dogs a little and did some things around her house. Then we left again to go to Target and get pedicures. After that, we decided to spend some time at the house with the dogs. Torrie and I talked about our weight-loss journeys; she had just started going for runs and walks every night. So we decided to take the dogs and the wine we had bought at Target. My uncle and she live on five acres of land right on a river, so it's perfect for the dogs to run around. We ran and walked for thirty minutes, and it felt really good! We ran by the river, in the water, and it was a great bonding experience. We set up some little chairs right by the river and even lit some tiki torches. My uncle came out and we just hung out and drank quite a bit. I was in the middle of a super interesting story when I heard a super loud crack! I jumped up out of my chair and so did my uncle. I looked behind me and a huge tree log had fallen from the tree. It landed five feet from where we were sitting and two feet from the cars. Somehow it didn't hit us, the cars, or the dogs! When I turned around, Torrie was still sitting in her chair; apparently my story was so interesting she didn't really react to the huge cracking sound. I looked for my uncle and he was in a ball about twenty feet away from us. I just started laughing and asked how he got there so quickly. He said he thought the whole tree was falling and just ran while looking back at the tree and tripped over a rock! He was fine, and we were just lucky that all of us and the dogs were okay. I left pretty early the next day to head back to NY. I had so much fun in VA, but I wish I had stayed longer — the time went by so fast and I feel like I spent most of it in the car. I can't wait to go back. I'm going back in three weeks when my uncle turns forty. They're having a huge party with a band by the river, but they're making sure all the dead tree limbs are gone first. Y'all read about my trip there — it took forever, but I made it safe and sound. I got into Lynchburg around five thirty p.m. and raced to my sister's house. Basically I dropped Izzie off, said a quick hello to friends from Canada, and then bolted because I had dinner plans at five thirty p.m. I was a little late but met Laurel at Buffalo Wild Wings, where I finally satisfied my recent craving for fried pickles. I'm not sure what that was about or where it came from. After that I went home and went to bed; I was exhausted from my drive. Lynne came home from her rehearsal dinner and we hung out for a little bit before I called it a night. The next day Lynne and I went to see my grandma, who had just had shoulder surgery, and it was good to see her. After that we went shopping, which was amazing. We took a little too long shopping, and when we got home we had an hour to get ready for the wedding. You know an hour is not long enough to get ready! Lynne only has one shower, so I opted to wash my hair in the sink — a girl has to do what a girl has to do. We managed to get there in time, by the grace of God, because I'm not sure how we did. The wedding was beautiful; just look at the pictures. Megan and Daniel are a great couple and they have an awesome relationship. The next day Justin, my brother-in-law, was working at the lake, and Lynne, some friends, and I drove out there to hang out. Another friend—are you following?—had gotten a boat with his parents and invited us to go tubing. Uhm, heck yeah (not the tubing part). We hung out on the boat, and Lynne was a beast out there on the tube; the driver definitely didn't hold back because she was a girl. It was all fun and games until we hit a tidal wave that struck my bag, which was canvas and not quite waterproof. There was a lot of water, and for some stupid reason I had brought my phone on the boat; I'm not sure why, because I didn't use it or need it.
long_en_331	poet_en	1,042	en	Earlier this evening, I saw a Facebook post from a cat owner announcing that her cat was pregnant. She asked people not to post any negative comments. She knew she should have had her cat spayed, but the pregnancy was an accident; the kittens would go to a good home. A post like that bothers me. How can a pet owner claim the pregnancy was an accident? Anyone with an ounce of responsibility knows that pets should be spayed or neutered at an appropriate age. It's that simple. Don't claim that you don't have the time—you don't have to do the procedure yourself: make an appointment with the vet, hop in your car or call a taxi, and take the pet to the doctor. Others claim they don't have the money for such a procedure. In that case, don't have a pet. When you adopt a pet, you accept responsibility for its well-being. That means three sets of inoculations and having the pet spayed or neutered. One shouldn't just think of the animal but also of the litters that a single cat or dog can produce. If one cat or dog has five litters of six kittens or puppies, that's 30 offspring. If each of those 30 has five litters of six, that's 900. When someone stated that unspayed or unneutered pets should be kept in the house, another replied that it's cruel to keep a pet indoors. I remember my father saying much the same when I told him about my indoor cats. He said that surely they miss walking the neighborhood and feeling the grass under their paws. So I asked him, "Dad, do you miss climbing Mount Everest and feeling the snow crunching under your feet with every step to reach the summit?" "Of course not," he said. "I've never climbed Everest, so how can I..." And right there and then he understood that it's not cruel to keep a cat indoors. The animal cannot miss what it never experienced. Finally, I want to address the matter of finding good homes for kittens or puppies. Never assume that when a person shows up to adopt a cat, he or she has the pets' best interests at heart. Good Friday is a day I have a problem with. Christians the world over are supposed to remember this day as the day when Jesus Christ died, but when exactly did he die? This year we remember the tragic event on March 25th, but let's have a look at other years: 2015 — April 3rd; 2014 — April 18th; 2013 — March 29th; 2012 — April 6th. How are we supposed to properly pay our respects when the date keeps changing? Yes, I know the date of Easter is based on the lunar calendar. When I looked it up on Google, I got the following explanation: Unlike a lot of Christian feast days, Easter does not have a fixed date. The feast is based on the lunar calendar, so Easter is scheduled to fall on the Sunday that follows the full moon on or after March 21, also known as the spring equinox. While this might be a perfectly rational explanation, to me it makes no sense. How can one remember the most important death in history but keep moving it around? Let's look at some other historical figures: Cleopatra — August 12, 30 BC; Julius Caesar — March 15, 44 BC; Nero — June 9, AD 68; Augustus — August 19, AD 14; Leonardo da Vinci — May 2, 1519; Galileo Galilei — January 8, 1642. If someone was able to record these dates for future generations, why mess with Jesus Christ's death? If it was, for instance, April 5, it should always be April 5, regardless of which day that date happens to fall on. We are equally wrong when we observe a moment of silence at 3:00 p.m., the time Jesus is supposed to have died. To be correct, on the East Coast of the United States and Canada this moment of silence should be observed at 8:00 a.m., as Jerusalem is seven hours ahead of the East Coast. I find it mind-boggling that throughout history people have drawn and written what was important to them, but nobody made an accurate notation of the date of Jesus Christ's death. Or did they and was this date conveniently "lost"? It is generally known that Constantine, together with 300 religious leaders, decided what information was included in the Bible. The cats' noses are a little out of joint because they can't go outside. Quite frankly, my nose is out of joint too. In the first week of March, the temperature reached 12 and even 14 degrees Celsius (that's 53.6 and 57.2 degrees F), and everyone thought spring had sprung early. I washed heavy winter jackets, sweaters, scarves, and gloves, all to be put away clean until we needed them again. While I washed and cleaned, the cats enjoyed a watery sun on the balcony. Yes, it was still a bit chilly, but in their fur coats they seemed comfortable. As the balcony door remained open, they were able to come and go as they pleased. That is to say, Charlotte, Mickey, and Gabriel came back in when it got too cold for them, but Holly had her own set of rules. She sat by the door and meowed when she wanted to come in. When I said, "Come on then; the others managed—you'll fit, too," she gave no sign of movement. So I opened the door a little more, enough for a poodle to fit through. Holly remained seated. I opened the door wider still; a German Shepherd would have had no problem squeezing through, but still no movement on Holly's part. So I opened the door so wide a horse could have galloped in, and, yes, Miss Holly daintily stepped over the threshold. Honestly, that cat has illusions of grandeur. Of course, now that they were used to going outside, they couldn't understand, when the weather turned colder again, why they couldn't go on the balcony any longer.	Earlier this evening, I saw a Facebook post from a cat owner announcing that her cat was pregnant. She asked people not to post any negative comments. She knew she should have had her cat spayed, but the pregnancy was an accident; the kittens would go to a good home. A post like that bothers me. How can a pet owner claim the pregnancy was an accident? Anyone with an ounce of responsibility knows that pets should be spayed or neutered at an appropriate age. It's that simple. Don't claim that you don't have the time—you don't have to do the procedure yourself: make an appointment with the vet, hop in your car or call a taxi, and take the pet to the doctor. Others claim they don't have the money for such a procedure. In that case, don't have a pet. When you adopt a pet, you accept responsibility for its well-being. That means three sets of inoculations and having the pet spayed or neutered. One shouldn't just think of the animal but also of the litters that a single cat or dog can produce. If one cat or dog has five litters of six kittens or puppies, that's thirty offspring. If each of those thirty has five litters of six, that's nine hundred. When someone stated that unspayed or unneutered pets should be kept in the house, another replied that it's cruel to keep a pet indoors. I remember my father saying much the same when I told him about my indoor cats. He said that surely they miss walking the neighborhood and feeling the grass under their paws. So I asked him, "Dad, do you miss climbing Mount Everest and feeling the snow crunching under your feet with every step to reach the summit?" "Of course not," he said. "I've never climbed Everest, so how can I..." And right there and then he understood that it's not cruel to keep a cat indoors. The animal cannot miss what it never experienced. Finally, I want to address the matter of finding good homes for kittens or puppies. Never assume that when a person shows up to adopt a cat, he or she has the pets' best interests at heart. Good Friday is a day I have a problem with. Christians the world over are supposed to remember this day as the day when Jesus Christ died, but when exactly did he die? This year we remember the tragic event on March twenty fifth, but let's have a look at other years: two thousand fifteen — April third; two thousand fourteen — April eighteenth; two thousand thirteen — March twenty ninth; two thousand twelve — April sixth. How are we supposed to properly pay our respects when the date keeps changing? Yes, I know the date of Easter is based on the lunar calendar. When I looked it up on Google, I got the following explanation: Unlike a lot of Christian feast days, Easter does not have a fixed date. The feast is based on the lunar calendar, so Easter is scheduled to fall on the Sunday that follows the full moon on or after March twenty first, also known as the spring equinox. While this might be a perfectly rational explanation, to me it makes no sense. How can one remember the most important death in history but keep moving it around? Let's look at some other historical figures: Cleopatra — August twelve, thirty BC; Julius Caesar — March fifteen, forty four BC; Nero — June nine, AD sixty eight; Augustus — August nineteen, AD fourteen; Leonardo da Vinci — May two, one thousand five hundred nineteen; Galileo Galilei — January eight, one thousand six hundred forty two. If someone was able to record these dates for future generations, why mess with Jesus Christ's death? If it was, for instance, April five, it should always be April five, regardless of which day that date happens to fall on. We are equally wrong when we observe a moment of silence at three p.m., the time Jesus is supposed to have died. To be correct, on the East Coast of the United States and Canada this moment of silence should be observed at eight a.m., as Jerusalem is seven hours ahead of the East Coast. I find it mind-boggling that throughout history people have drawn and written what was important to them, but nobody made an accurate notation of the date of Jesus Christ's death. Or did they and was this date conveniently "lost"? It is generally known that Constantine, together with three hundred religious leaders, decided what information was included in the Bible. The cats' noses are a little out of joint because they can't go outside. Quite frankly, my nose is out of joint too. In the first week of March, the temperature reached twelve and even fourteen degrees Celsius (that's fifty three point six and fifty seven point two degrees F), and everyone thought spring had sprung early. I washed heavy winter jackets, sweaters, scarves, and gloves, all to be put away clean until we needed them again. While I washed and cleaned, the cats enjoyed a watery sun on the balcony. Yes, it was still a bit chilly, but in their fur coats they seemed comfortable. As the balcony door remained open, they were able to come and go as they pleased. That is to say, Charlotte, Mickey, and Gabriel came back in when it got too cold for them, but Holly had her own set of rules. She sat by the door and meowed when she wanted to come in. When I said, "Come on then; the others managed—you'll fit, too," she gave no sign of movement. So I opened the door a little more, enough for a poodle to fit through. Holly remained seated. I opened the door wider still; a German Shepherd would have had no problem squeezing through, but still no movement on Holly's part. So I opened the door so wide a horse could have galloped in, and, yes, Miss Holly daintily stepped over the threshold. Honestly, that cat has illusions of grandeur. Of course, now that they were used to going outside, they couldn't understand, when the weather turned colder again, why they couldn't go on the balcony any longer.
long_en_357	poet_en	987	en	So, there's this thing out in the cyberspace world called Twitter. I'm sure most of you have heard of it: www.twitter.com. I signed up about a year ago, but really didn't do much with it. I mean, how do you write in 160 characters or less, what you are doing RIGHT NOW! Well, mainly I was sitting at my computer. But then I realized that you could "tweet" from your cell phone. I wasn't doing much from there either and I didn't really have anyone to follow. Over the course of the year I found some of my friends on there and kept tabs on what they were doing. I have one friend who tweets like crazy. She's a real smart cookie with a long history... oh wait, that's another song. Backtrack... well, she is really smart and uses every new toy to her advantage. She's got Twitter wired for sound among her blogs, museums, and other techno stuff. I recently figured out that if I clicked on "Who I'm following" and turned the device status to on, then I would get text messages when they tweeted. I know, I'm slow, but I'm learning! I think I'm going to have to upgrade my text messaging plan to unlimited pretty soon. This is a great networking tool, especially in the business world. Not only can I tweet with my friends, but I'm also able to keep up with my colleagues. I can already see this will be great when I attend the HSMG conference in May. So, all of you in the guild and otherwise who are on Twitter, come follow me @shadowfax0704. Those who haven't yet joined Twitter can find me at http://twitter.com/shadowfax0704. Just put in the comments section that you found me on my blog so I'll know who you are! Oh, and if any of you Twitter experts know of more tips and tricks out there, please let me know. I am still learning my way around Twitter and would love to know more. Just add your comments below. And now, in 160 characters or less: putting out a great blog on the joy of tweeting! Who says living out in the middle of nowhere isn't interesting? Many of you hear me talk about the musings of country life. Well, it's an endless stream of bizarre goings-on. We have the dead lady that lives next door, the fake illegals across the street and, with them, their horse who apparently hates living there! She's always breaking out and coming to our house—pretty much every day (or night) she comes for a visit. I guess it wouldn't be bad if we wanted a third horse, but considering she comes into our barn and then breaks into the food bins, it gets a little tiresome. This last time she, yet again, broke into the feed bags and consumed half a bag of feed. Yes, I know it is bad for a horse to eat that much food. We called her owners only to find out he was out of town. All throughout the day we chased her and ran her down the road only to see she returned. We let the dogs out after her and she nonchalantly brushed them off. I guess she got tired of my chasing her out of the barn. She sauntered to the end of our driveway and lay down for a nap, much like a stray dog deciding to catch some rays in the sun. She didn't even move when my husband came home from work. She casually lifted her head as if he might disturb her beauty sleep, then stretched out even more. At that time I looked across the street and noticed the neighbor's son jumping on the trampoline. Did he not notice his horse had been missing all day and all the previous night? I yelled to the boy to come and retrieve his horse. He promptly came over, and when I asked him if he had been home, he said, "Why yes, I had." I told him that we had his horse, and all I received was a grumbled, "Oh." I then proceeded to tell him that his horse had eaten a bag of our feed and that I had been chasing the horse out of our barn all day. Nary a word of sorrow or concern for the horse or any trouble we had gone through was spoken by this creature of a boy. Needless to say, I was not happy about the situation. Every day we meet someone—whether it's someone helping to carry boxes from your car, a new acquaintance, or a friend for life, even if they have different political or religious persuasions. Just the other day I met a few more interesting people, all extremely different from one another. One was Native American. He was a healer and had finished making a documentary; I had the privilege of watching a bit of his unfinished film. It was very interesting, and what made it more so was that, sitting next to him, I could ask questions about the different elements and why they were included. He told me he burned sage at the beginning to purify the surroundings. I did not know that. Another person I met that evening was a scriptwriter who is filming a comedic-horror zombie movie. It's a well-thought-out project and sounds like it will be a great film. He's quite a character and very humorous, so I'm sure this translates well on screen. His producer had picked up the movie because it was different from most slasher, gore-filled horror films. I like that on any given day you never know who you may meet.	So, there's this thing out in the cyberspace world called Twitter. I'm sure most of you have heard of it: www dot twitter dot com. I signed up about a year ago, but really didn't do much with it. I mean, how do you write in one hundred sixty characters or less, what you are doing RIGHT NOW! Well, mainly I was sitting at my computer. But then I realized that you could "tweet" from your cell phone. I wasn't doing much from there either and I didn't really have anyone to follow. Over the course of the year I found some of my friends on there and kept tabs on what they were doing. I have one friend who tweets like crazy. She's a real smart cookie with a long history... oh wait, that's another song. Backtrack... well, she is really smart and uses every new toy to her advantage. She's got Twitter wired for sound among her blogs, museums, and other techno stuff. I recently figured out that if I clicked on "Who I'm following" and turned the device status to on, then I would get text messages when they tweeted. I know, I'm slow, but I'm learning! I think I'm going to have to upgrade my text messaging plan to unlimited pretty soon. This is a great networking tool, especially in the business world. Not only can I tweet with my friends, but I'm also able to keep up with my colleagues. I can already see this will be great when I attend the HSMG conference in May. So, all of you in the guild and otherwise who are on Twitter, come follow me at shadowfax zero seven zero four. Those who haven't yet joined Twitter can find me at http colon slash slash twitter dot com slash shadowfax zero seven zero four. Just put in the comments section that you found me on my blog so I'll know who you are! Oh, and if any of you Twitter experts know of more tips and tricks out there, please let me know. I am still learning my way around Twitter and would love to know more. Just add your comments below. And now, in one hundred sixty characters or less: putting out a great blog on the joy of tweeting! Who says living out in the middle of nowhere isn't interesting? Many of you hear me talk about the musings of country life. Well, it's an endless stream of bizarre goings-on. We have the dead lady that lives next door, the fake illegals across the street and, with them, their horse who apparently hates living there! She's always breaking out and coming to our house—pretty much every day (or night) she comes for a visit. I guess it wouldn't be bad if we wanted a third horse, but considering she comes into our barn and then breaks into the food bins, it gets a little tiresome. This last time she, yet again, broke into the feed bags and consumed half a bag of feed. Yes, I know it is bad for a horse to eat that much food. We called her owners only to find out he was out of town. All throughout the day we chased her and ran her down the road only to see she returned. We let the dogs out after her and she nonchalantly brushed them off. I guess she got tired of my chasing her out of the barn. She sauntered to the end of our driveway and lay down for a nap, much like a stray dog deciding to catch some rays in the sun. She didn't even move when my husband came home from work. She casually lifted her head as if he might disturb her beauty sleep, then stretched out even more. At that time I looked across the street and noticed the neighbor's son jumping on the trampoline. Did he not notice his horse had been missing all day and all the previous night? I yelled to the boy to come and retrieve his horse. He promptly came over, and when I asked him if he had been home, he said, "Why yes, I had." I told him that we had his horse, and all I received was a grumbled, "Oh." I then proceeded to tell him that his horse had eaten a bag of our feed and that I had been chasing the horse out of our barn all day. Nary a word of sorrow or concern for the horse or any trouble we had gone through was spoken by this creature of a boy. Needless to say, I was not happy about the situation. Every day we meet someone—whether it's someone helping to carry boxes from your car, a new acquaintance, or a friend for life, even if they have different political or religious persuasions. Just the other day I met a few more interesting people, all extremely different from one another. One was Native American. He was a healer and had finished making a documentary; I had the privilege of watching a bit of his unfinished film. It was very interesting, and what made it more so was that, sitting next to him, I could ask questions about the different elements and why they were included. He told me he burned sage at the beginning to purify the surroundings. I did not know that. Another person I met that evening was a scriptwriter who is filming a comedic-horror zombie movie. It's a well-thought-out project and sounds like it will be a great film. He's quite a character and very humorous, so I'm sure this translates well on screen. His producer had picked up the movie because it was different from most slasher, gore-filled horror films. I like that on any given day you never know who you may meet.
long_en_354	poet_en	851	en	Getting used to this stupid routine of singing, along with work and training, made everything so much harder. I wanted to think about our next step, but they made me shift twice a day and had someone in my head at all times. Even when I tried to keep them out of my head, it seemed they were doing the same to me. I couldn't learn anything more about the bomb or their plans. I was still an outsider, and I accepted it — I didn't want to be part of their pack, especially when I couldn't even think normally. I grew so sick of singing the same songs that I started singing ones I didn't know all the words to, then lost track of the melody and wasted an hour. I knew they would always talk about me and read my thoughts, even laugh, but I kept up the pretense and carried on as I had for weeks, with only minimal time to figure out what to do next. Six weeks — six weeks and I hadn't seen Rexxie. How much longer would it take to get back to our mate? I was really glad Isilesah was with me; the only other person I talked to was Pete, apart from the times they forced me to speak, despite knowing I could. I wanted to go back to Rexxie. The next week, things went from bad to worse. It wasn’t just that I knew they were up to something — something big. Whatever was happening outside the pack seemed to have picked up pace, and apparently I was now a useful pawn. It was "time," I’d heard, and I’d already noticed that some shifters had arrived and many had left; it was very different from the previous six weeks. I had not removed the bomb. I suppose I had been so focused on not thinking about it that I never did. It was worrying, because even if I had had the chance to take it out, I felt I would simply be given another one. So it was put directly on hold. Every time the "bomb" tried to call me to remind me it was there, there was only a long beep on the other end. But this week things got desperate. First, they discussed my leaving right in front of me. They said I should follow the path they had shown me; I shook my head and said I was part of their pack now. One of them got mad, stood up, and said I was needed for their plan to work perfectly, but was told to be quiet. Then the Alpha's aura was used and I was forced to leave — with the bomb still implanted in my shoulder. What they hadn't thought about was my peculiar way of thinking; when the Alpha commanded that I leave the boundary, that is what I did. I left the boundary, then sat right next to it and did not move. I did what I was told. Singing songs in my head while still in wolf form, it wasn't long before they realized I wasn't going to move. When they brought me back to the pack, that's when things really hit the fan. They ganged up on me, called me names, and threw things that even stank. There were only about ten of them, but ten against one felt dishonorable; I could only think they were all bastards. I had been told to shift and was forced to kneel in front of that witch of a woman. "Luna, she's here!" Hearing a chuckle, I saw her relishing the moment. "You might think you're clever, little one, but you're not." She pulled my hair and pressed my face into the dirt. "You like this place so much? Then have some—eat it!" Laughter rang out and I desperately tried not to eat the dirt next to my mouth. A minute later, when the pressure on my head lessened, I kept my face bowed, terrified. I knew they needed me alive, but torture didn't sit well with me, and I knew they had no time to waste—they were in a hurry to use me. "Mmm, it seems the little one thinks she's blind…" The woman's ugly voice sighed, cheers went up, and I shifted from panic to a deeper fear of what was coming next. And just as I feared, it went from bad to worse. That witch of a woman savagely lifted my head, bared her wolf-like claws, and began gouging my eyes. I screamed. It was unlike anything I had ever experienced — and I suspect few others had either: to go from being able to see to seeing nothing, and to feel that kind of pain. I felt blood run down my cheeks and chin, and I think one of my eyes popped like a balloon. My screams, swallowed by the woman's merciless actions, were cut off when a hand closed around my neck.	Getting used to this stupid routine of singing, along with work and training, made everything so much harder. I wanted to think about our next step, but they made me shift twice a day and had someone in my head at all times. Even when I tried to keep them out of my head, it seemed they were doing the same to me. I couldn't learn anything more about the bomb or their plans. I was still an outsider, and I accepted it — I didn't want to be part of their pack, especially when I couldn't even think normally. I grew so sick of singing the same songs that I started singing ones I didn't know all the words to, then lost track of the melody and wasted an hour. I knew they would always talk about me and read my thoughts, even laugh, but I kept up the pretense and carried on as I had for weeks, with only minimal time to figure out what to do next. six weeks — six weeks and I hadn't seen Rexxie. How much longer would it take to get back to our mate? I was really glad Isilesah was with me; the only other person I talked to was Pete, apart from the times they forced me to speak, despite knowing I could. I wanted to go back to Rexxie. The next week, things went from bad to worse. It wasn’t just that I knew they were up to something — something big. Whatever was happening outside the pack seemed to have picked up pace, and apparently I was now a useful pawn. It was "time," I’d heard, and I’d already noticed that some shifters had arrived and many had left; it was very different from the previous six weeks. I had not removed the bomb. I suppose I had been so focused on not thinking about it that I never did. It was worrying, because even if I had had the chance to take it out, I felt I would simply be given another one. So it was put directly on hold. Every time the "bomb" tried to call me to remind me it was there, there was only a long beep on the other end. But this week things got desperate. First, they discussed my leaving right in front of me. They said I should follow the path they had shown me; I shook my head and said I was part of their pack now. One of them got mad, stood up, and said I was needed for their plan to work perfectly, but was told to be quiet. Then the Alpha's aura was used and I was forced to leave — with the bomb still implanted in my shoulder. What they hadn't thought about was my peculiar way of thinking; when the Alpha commanded that I leave the boundary, that is what I did. I left the boundary, then sat right next to it and did not move. I did what I was told. Singing songs in my head while still in wolf form, it wasn't long before they realized I wasn't going to move. When they brought me back to the pack, that's when things really hit the fan. They ganged up on me, called me names, and threw things that even stank. There were only about ten of them, but ten against one felt dishonorable; I could only think they were all bastards. I had been told to shift and was forced to kneel in front of that witch of a woman. "Luna, she's here!" Hearing a chuckle, I saw her relishing the moment. "You might think you're clever, little one, but you're not." She pulled my hair and pressed my face into the dirt. "You like this place so much? Then have some—eat it!" Laughter rang out and I desperately tried not to eat the dirt next to my mouth. A minute later, when the pressure on my head lessened, I kept my face bowed, terrified. I knew they needed me alive, but torture didn't sit well with me, and I knew they had no time to waste—they were in a hurry to use me. "Mmm, it seems the little one thinks she's blind…" The woman's ugly voice sighed, cheers went up, and I shifted from panic to a deeper fear of what was coming next. And just as I feared, it went from bad to worse. That witch of a woman savagely lifted my head, bared her wolf-like claws, and began gouging my eyes. I screamed. It was unlike anything I had ever experienced — and I suspect few others had either: to go from being able to see to seeing nothing, and to feel that kind of pain. I felt blood run down my cheeks and chin, and I think one of my eyes popped like a balloon. My screams, swallowed by the woman's merciless actions, were cut off when a hand closed around my neck.
long_en_279	wiki_en	991	en	Secret Project Revolution is a 2013 American short film directed by Madonna and Steven Klein that deals with artistic freedom and human rights. The film launched a global initiative, Art for Freedom, to further freedom of expression; the initiative was created by Madonna, curated by Vice, and distributed by BitTorrent. The goal of Art for Freedom is to promote and facilitate free expression and thereby combat the repression of artistic expression. Synopsis Secret Project Revolution is a 17-minute black-and-white film that features Madonna's off-camera commentary throughout, consisting of a narrative recorded exclusively for the project as well as excerpts from her onstage speeches during The MDNA Tour. It opens with a scene depicting Madonna locked in a prison cell, holding the bars; this is projected against a voice-over composed of excerpts from her onstage speech at the Olympia in Paris in July 2012. The speech reflected on the global economic situation at the time, arguing that fear is the source of intolerance. The film then shows a pantomime in which the tour dancers sit and stand motionless in a room while Madonna enters holding a gun — representing branding — and begins shooting them. This section ends with the words "All you need for a movie is a gun and a girl," a quote from Jean-Luc Godard, displayed across the screen. The following scene shows Madonna being dragged by police officers to a cell, where she is thrown to the floor. Her narrative begins, explaining the circumstances that prompted the project. For several minutes, footage of Madonna in her cell alternates with scenes of a man being tortured and a dancer performing routines. Madonna is then shown restrained on a bed while a dancer teases her and simulates torture. The next scene depicts an imprisoned man who has clearly been physically tortured, blood on his face and chest. In the following segment, dancers perform routines and adopt expressive poses as downtempo piano music plays and Madonna continues her narrative. An almost nude man dances in front of guards, and Madonna's a cappella performance of "My Country, 'Tis of Thee" begins, followed by excerpts from her Saint Petersburg speech. Footage of Madonna killing the dancers is played backwards, restoring them to life. Madonna is again shown being dragged by police officers, who drop her on the floor and look down on her. This is followed by a Jean-Paul Sartre quote: "Freedom is what you do with what's been done to you." The film ends with a dedication to those who have been, are, or may be persecuted for the color of their skin, their religious beliefs, their artistic expression, their gender, or their sexual preferences. For anyone whose human rights have been violated. Production Background In late 2012, Madonna and Steven Klein were due to work together on a photo shoot for the promotional campaign of shoes and lingerie she had designed as part of her Truth or Dare collection. However, her products were met with an unfavorable reception from the distributing company as being too provocative, which led to cancellation of the whole project. Instead, Madonna decided to use the set to film the project with Klein and her dancers, using lingerie she had designed. The work began in Buenos Aires, Argentina, in mid-December 2012, during Madonna's break between performances in the country. In the following weeks, the singer began writing the narration and more footage was produced. The film was her artistic response to a series of social and political events that took place around the world while she was performing The MDNA Tour in the second half of 2012. These events included, among others, the threat of Israel striking Iran, the imprisonment of Yulia Tymoshenko in Ukraine and Pussy Riot in Russia, gay rights violations, the United States presidential election, and the attempted assassination of Malala Yousafzai. One scene was inspired by the film The Night Porter, while another was inspired by the scene in the film noir Caged in which Eleanor Parker is jailed. Madonna has described secretprojectrevolution as one of the most important things she has ever done. Development: On December 14, 2012, a picture taken on the set surfaced online, sparking rumors that Madonna was working on a new music video, which many fans thought would be "Gang Bang." On December 19, Steven Klein first referred to the film as "#secretproject" on Twitter, which would serve as its official name. Later that month he contacted Rihanna, Lady Gaga, Cher, Dita Von Teese, and Naomi Campbell on Twitter, inviting them to participate. A picture captioned "From Madonna to Rihanna 2 of 3" emerged in early 2013, followed by a photograph of Naomi Campbell and Kate Moss associated with the project. More set pictures surfaced, and Madonna began posting stills and related images on Instagram in March 2013, using the hashtag #secretproject, later changed to #secretprojectrevolution. On March 20, 2013, Klein published the first trailer for the film. In July, rumors circulated that the project was actually a launch of a new clothing line designed by Madonna's then-partner Brahim Zaibat. Release: Initially announced for release on May 12, 2013, the film ultimately premiered on September 23, 2013, with eighteen outdoor screenings across five cities in four countries: New York, Los Angeles, Toronto, Berlin, and London. The following day, Secretprojectrevolution was released in the same form in Chicago, San Francisco, Rome, Paris, and Tel Aviv, the latter added at the last minute as the final location. All locations were revealed shortly before the events via Madonna's social media profiles. On September 24, the full film was uploaded to YouTube and released via BitTorrent's new Bundle feature. The Bundle included unlocked trailers and stills, plus three videos and a written message from Madonna; these items were locked but could be unlocked by submitting a valid email address. BitTorrent said the Bundle feature was still in alpha.	Secret Project Revolution is a two thousand thirteen American short film directed by Madonna and Steven Klein that deals with artistic freedom and human rights. The film launched a global initiative, Art for Freedom, to further freedom of expression; the initiative was created by Madonna, curated by Vice, and distributed by BitTorrent. The goal of Art for Freedom is to promote and facilitate free expression and thereby combat the repression of artistic expression. Synopsis Secret Project Revolution is a seventeen-minute black-and-white film that features Madonna's off-camera commentary throughout, consisting of a narrative recorded exclusively for the project as well as excerpts from her onstage speeches during The MDNA Tour. It opens with a scene depicting Madonna locked in a prison cell, holding the bars; this is projected against a voice-over composed of excerpts from her onstage speech at the Olympia in Paris in July two thousand twelve. The speech reflected on the global economic situation at the time, arguing that fear is the source of intolerance. The film then shows a pantomime in which the tour dancers sit and stand motionless in a room while Madonna enters holding a gun — representing branding — and begins shooting them. This section ends with the words "All you need for a movie is a gun and a girl," a quote from Jean-Luc Godard, displayed across the screen. The following scene shows Madonna being dragged by police officers to a cell, where she is thrown to the floor. Her narrative begins, explaining the circumstances that prompted the project. For several minutes, footage of Madonna in her cell alternates with scenes of a man being tortured and a dancer performing routines. Madonna is then shown restrained on a bed while a dancer teases her and simulates torture. The next scene depicts an imprisoned man who has clearly been physically tortured, blood on his face and chest. In the following segment, dancers perform routines and adopt expressive poses as downtempo piano music plays and Madonna continues her narrative. An almost nude man dances in front of guards, and Madonna's a cappella performance of "My Country, 'Tis of Thee" begins, followed by excerpts from her Saint Petersburg speech. Footage of Madonna killing the dancers is played backwards, restoring them to life. Madonna is again shown being dragged by police officers, who drop her on the floor and look down on her. This is followed by a Jean-Paul Sartre quote: "Freedom is what you do with what's been done to you." The film ends with a dedication to those who have been, are, or may be persecuted for the color of their skin, their religious beliefs, their artistic expression, their gender, or their sexual preferences. For anyone whose human rights have been violated. Production Background In late two thousand twelve, Madonna and Steven Klein were due to work together on a photo shoot for the promotional campaign of shoes and lingerie she had designed as part of her Truth or Dare collection. However, her products were met with an unfavorable reception from the distributing company as being too provocative, which led to cancellation of the whole project. Instead, Madonna decided to use the set to film the project with Klein and her dancers, using lingerie she had designed. The work began in Buenos Aires, Argentina, in mid-December two thousand twelve, during Madonna's break between performances in the country. In the following weeks, the singer began writing the narration and more footage was produced. The film was her artistic response to a series of social and political events that took place around the world while she was performing The MDNA Tour in the second half of two thousand twelve. These events included, among others, the threat of Israel striking Iran, the imprisonment of Yulia Tymoshenko in Ukraine and Pussy Riot in Russia, gay rights violations, the United States presidential election, and the attempted assassination of Malala Yousafzai. One scene was inspired by the film The Night Porter, while another was inspired by the scene in the film noir Caged in which Eleanor Parker is jailed. Madonna has described secretprojectrevolution as one of the most important things she has ever done. Development: On December fourteenth, two thousand twelve, a picture taken on the set surfaced online, sparking rumors that Madonna was working on a new music video, which many fans thought would be "Gang Bang." On December nineteenth, two thousand twelve, Steven Klein first referred to the film as "hashtag secretproject" on Twitter, which would serve as its official name. Later that month he contacted Rihanna, Lady Gaga, Cher, Dita Von Teese, and Naomi Campbell on Twitter, inviting them to participate. A picture captioned "From Madonna to Rihanna two of three" emerged in early two thousand thirteen, followed by a photograph of Naomi Campbell and Kate Moss associated with the project. More set pictures surfaced, and Madonna began posting stills and related images on Instagram in March two thousand thirteen, using the hashtag hashtag secretproject, later changed to hashtag secretprojectrevolution. On March twentieth, two thousand thirteen, Klein published the first trailer for the film. In July, rumors circulated that the project was actually a launch of a new clothing line designed by Madonna's then-partner Brahim Zaibat. Release: Initially announced for release on May twelfth, two thousand thirteen, the film ultimately premiered on September twenty third, two thousand thirteen, with eighteen outdoor screenings across five cities in four countries: New York, Los Angeles, Toronto, Berlin, and London. The following day, Secretprojectrevolution was released in the same form in Chicago, San Francisco, Rome, Paris, and Tel Aviv, the latter added at the last minute as the final location. All locations were revealed shortly before the events via Madonna's social media profiles. On September twenty four, the full film was uploaded to YouTube and released via BitTorrent's new Bundle feature. The Bundle included unlocked trailers and stills, plus three videos and a written message from Madonna; these items were locked but could be unlocked by submitting a valid email address. BitTorrent said the Bundle feature was still in alpha.
long_en_292	wiki_en	1,017	en	The March for Science Portland (also known as March for Science PDX and Portland March for Science) was a protest held in Portland, Oregon, on April 22, 2017 (Earth Day). It was part of the global March for Science, a series of rallies in Washington, D.C., and more than 600 cities worldwide. Organized by Portland Science Advocates, the march supported science and protested President Donald Trump's proposed cuts to funding for the Environmental Protection Agency and the National Institutes of Health. Funding for the event, which cost approximately $30,000, was crowdsourced. Amid rainy weather, thousands attended, gathering at Tom McCall Waterfront Park before marching 44 blocks through downtown Portland. Speakers included Earl Blumenauer, Suzanne Bonamici, and Elizabeth Steiner Hayward. Supporting organizations included the Audubon Society of Portland, the Oregon Environmental Council, Oregon Health & Science University, and the Xerces Society. Reporters noted organizers' efforts to create a political yet nonpartisan atmosphere, though participants publicly criticized Trump. The event featured activities for children and was described as family-friendly. The group included people from diverse disciplines, such as Jackie Wirz, an assistant dean at Oregon Health & Science University, and a store clerk with a passion for science and writing. The event's steering committee was co-chaired by Curt Waltman and Wirz; Rich Hatfield served as communications committee co-chair, and Sumi Malik, a transportation planner for CH2M Hill, served as community outreach co-chair. Alex Conley was also an organizer. The organizers met through the event's Facebook page, launched in February 2017, and sought advice from the organizers of the Women's March on Portland, held in January 2017. They aimed to support science and protest President Donald Trump's proposed cuts to funding for the Environmental Protection Agency and the National Institutes of Health, as well as his threat to withdraw from the Paris Agreement and reduce funding for scientific research and K–12 science programs, among other policies. Prior to the march, Waltman and Wirz wrote a guest column in The Oregonian praising Oregon's achievements in science and encouraging people to "stand up for science." They wrote, "We stand in support of scientific inquiry, researchers, availability of data and evidence-based policy," and invited people who "love" science to participate. Preparation and planning: funding for the event, which cost approximately $30,000, was crowdsourced. Obtaining a permit to host the event at the park cost $3,000. Monetary donations were accepted, and merchandise sales also funded the demonstration. Funds were transferred through the Xerces Society. Portland Science Advocates was established as a nonprofit organization for members to continue the demonstration's mission after the march. Prior to the march, organizers surveyed Tom McCall Waterfront Park and posted flyers about the event in the surrounding area. They emphasized that all people were invited to participate, not just scientists, and anticipated participation by approximately 10,000 demonstrators. Nearly 7,000 had committed to attending the event on its Facebook page. Organizers planned to use drones to help estimate the crowd size. More than a thousand people had gathered by 10:00 a.m. at Tom McCall Waterfront Park, near the Morrison Bridge. Six business leaders, innovators, politicians, and scientists spoke at the event. Politicians included U.S. Representatives Earl Blumenauer and Suzanne Bonamici, and Elizabeth Steiner Hayward, a Democratic member of the Oregon State Senate for the 17th District. Blumenauer referred to the Trump administration in his speech, saying, "We have failed in the political process because they've made science partisan." "They've undermined the credibility and confidence that people have, for example in science and in dealing with climate," he said. He also spoke about the importance of evidence-based decision-making. In her speech, Bonamici said science should not be a partisan issue, and told the crowd: "We are going to go back to Washington, D.C., with all of you in our minds — this beautiful sea of science supporters — and fight for research, fight for clean air, fight for clean water, and fight for our planet. Stay engaged; this is just the beginning." The other speakers included Xerces Society senior conservation biologist Rich Hatfield, Intel educator and innovation manager Shashi Jain, and Gabe Sheoships, a citizen of the Cayuse and Walla Walla tribes of the Confederated Tribes of the Umatilla Indian Reservation, who serves as education director for Friends of Tryon Creek and is an adjunct faculty member at Portland State University. At about 11:00 a.m., protesters began marching south along Naito Parkway. They navigated a 44-block portion of downtown before returning to the park around noon for a "Science Expo" and "Kid Zone," as well as live music. The route was identical to the one used during the Women's March on Portland; participants followed Naito to Jefferson Street, to Fourth Avenue, to Pine Street, then revisited Tom McCall Waterfront Park. Family-friendly educational booths and vendors were stationed near the park's Battleship Oregon Memorial. The demonstration ended at about 4:00 p.m. Amid rainy weather, thousands of people attended; some estimates far exceeded the anticipated 10,000 participants. Many protesters wore science-themed costumes, such as bio suits, space suits, and white lab coats. The city experienced increased traffic because of street and bridge closures for the march and other events. The march was endorsed by the Audubon Society of Portland, Oregon Health & Science University, and the Xerces Society, and was promoted by the Democratic Party of Washington County, the Oregon Environmental Council, Portland State University's biology department, and Rose City Astronomers. The Center for Biological Diversity issued a press release confirming some of its scientists planned to march in Portland and other cities across the United States. The educational nonprofit Science Project had a booth at the park. The American Chemical Society, which publishes Chemical & Engineering News, supported the parent March for Science demonstration as long as organizers remained nonpartisan; the magazine's former editor-in-chief, Rudy Baum, covered the Portland march. Portland's Badass Women's Protest Choir, formed after the 2016 presidential election, also planned to attend.	The March for Science Portland (also known as March for Science PDX and Portland March for Science) was a protest held in Portland, Oregon, on April twenty-two, two thousand seventeen (Earth Day). It was part of the global March for Science, a series of rallies in Washington, D.C., and more than six hundred cities worldwide. Organized by Portland Science Advocates, the march supported science and protested President Donald Trump's proposed cuts to funding for the Environmental Protection Agency and the National Institutes of Health. Funding for the event, which cost approximately thirty thousand dollars, was crowdsourced. Amid rainy weather, thousands attended, gathering at Tom McCall Waterfront Park before marching forty-four blocks through downtown Portland. Speakers included Earl Blumenauer, Suzanne Bonamici, and Elizabeth Steiner Hayward. Supporting organizations included the Audubon Society of Portland, the Oregon Environmental Council, Oregon Health and Science University, and the Xerces Society. Reporters noted organizers' efforts to create a political yet nonpartisan atmosphere, though participants publicly criticized Trump. The event featured activities for children and was described as family-friendly. The group included people from diverse disciplines, such as Jackie Wirz, an assistant dean at Oregon Health and Science University, and a store clerk with a passion for science and writing. The event's steering committee was co-chaired by Curt Waltman and Wirz; Rich Hatfield served as communications committee co-chair, and Sumi Malik, a transportation planner for CH2M Hill, served as community outreach co-chair. Alex Conley was also an organizer. The organizers met through the event's Facebook page, launched in February two thousand seventeen, and sought advice from the organizers of the Women's March on Portland, held in January two thousand seventeen. They aimed to support science and protest President Donald Trump's proposed cuts to funding for the Environmental Protection Agency and the National Institutes of Health, as well as his threat to withdraw from the Paris Agreement and reduce funding for scientific research and K to twelve science programs, among other policies. Prior to the march, Waltman and Wirz wrote a guest column in The Oregonian praising Oregon's achievements in science and encouraging people to "stand up for science." They wrote, "We stand in support of scientific inquiry, researchers, availability of data and evidence-based policy," and invited people who "love" science to participate. Preparation and planning: funding for the event, which cost approximately thirty thousand dollars, was crowdsourced. Obtaining a permit to host the event at the park cost three thousand dollars. Monetary donations were accepted, and merchandise sales also funded the demonstration. Funds were transferred through the Xerces Society. Portland Science Advocates was established as a nonprofit organization for members to continue the demonstration's mission after the march. Prior to the march, organizers surveyed Tom McCall Waterfront Park and posted flyers about the event in the surrounding area. They emphasized that all people were invited to participate, not just scientists, and anticipated participation by approximately ten thousand demonstrators. Nearly seven thousand had committed to attending the event on its Facebook page. Organizers planned to use drones to help estimate the crowd size. More than a thousand people had gathered by ten a.m. at Tom McCall Waterfront Park, near the Morrison Bridge. Six business leaders, innovators, politicians, and scientists spoke at the event. Politicians included U.S. Representatives Earl Blumenauer and Suzanne Bonamici, and Elizabeth Steiner Hayward, a Democratic member of the Oregon State Senate for the seventeenth District. Blumenauer referred to the Trump administration in his speech, saying, "We have failed in the political process because they've made science partisan." "They've undermined the credibility and confidence that people have, for example in science and in dealing with climate," he said. He also spoke about the importance of evidence-based decision-making. In her speech, Bonamici said science should not be a partisan issue, and told the crowd: "We are going to go back to Washington, D.C., with all of you in our minds — this beautiful sea of science supporters — and fight for research, fight for clean air, fight for clean water, and fight for our planet. Stay engaged; this is just the beginning." The other speakers included Xerces Society senior conservation biologist Rich Hatfield, Intel educator and innovation manager Shashi Jain, and Gabe Sheoships, a citizen of the Cayuse and Walla Walla tribes of the Confederated Tribes of the Umatilla Indian Reservation, who serves as education director for Friends of Tryon Creek and is an adjunct faculty member at Portland State University. At about eleven a.m., protesters began marching south along Naito Parkway. They navigated a forty-four-block portion of downtown before returning to the park around noon for a "Science Expo" and "Kid Zone," as well as live music. The route was identical to the one used during the Women's March on Portland; participants followed Naito to Jefferson Street, to Fourth Avenue, to Pine Street, then revisited Tom McCall Waterfront Park. Family-friendly educational booths and vendors were stationed near the park's Battleship Oregon Memorial. The demonstration ended at about four p.m. Amid rainy weather, thousands of people attended; some estimates far exceeded the anticipated ten thousand participants. Many protesters wore science-themed costumes, such as bio suits, space suits, and white lab coats. The city experienced increased traffic because of street and bridge closures for the march and other events. The march was endorsed by the Audubon Society of Portland, Oregon Health and Science University, and the Xerces Society, and was promoted by the Democratic Party of Washington County, the Oregon Environmental Council, Portland State University's biology department, and Rose City Astronomers. The Center for Biological Diversity issued a press release confirming some of its scientists planned to march in Portland and other cities across the United States. The educational nonprofit Science Project had a booth at the park. The American Chemical Society, which publishes Chemical and Engineering News, supported the parent March for Science demonstration as long as organizers remained nonpartisan; the magazine's former editor-in-chief, Rudy Baum, covered the Portland march. Portland's Badass Women's Protest Choir, formed after the two thousand sixteen presidential election, also planned to attend.
long_en_348	poet_en	1,000	en	I set up my father's computer a couple of weeks ago. He needed written directions to use Word. I had to fix the inevitable paper jam and explain the directions again. Update: He wrote a letter in Word to a friend from the Navy, printed it, put it in an envelope, and sent it via USPS. I saw the first attempt. I'm sure the recipient is wondering what's up. His spelling is atrocious; I did not remind him about spell check because it would have been too confusing. His words had numbers and stuff. The first line of the letter I saw, without the numbers, read: "Well, I'm back in business with email. My daughter got the computer running again for me." Early tomorrow morning the boys, soon to be 3 and 6, will be coming to our house for a sleepover. Their mother, my daughter, is having her tonsils removed. After spending the last year with one throat infection after another, and ending a month ago with an abscess that wouldn't go away, the ear, nose, and throat specialist strongly urged her to do so. Part of her problem is that she is allergic to most antibiotics. Her regular doctor rarely prescribes antibiotics, preferring to let her fight off most infections naturally and save the drugs for emergencies. This last infection took two rounds of different drugs to finally cure her. The boys are more excited about the sleepover than having any sympathy for their mother. I'm sure they have big plans for the week. They were mad that it didn't start tonight. Auntie, my other daughter, gave them Spider-Man beds for the special sleepover when they were in Chicago visiting her a few weeks ago. Let's hope their parents have packed enough for this sleepover. Last time, one child had no underwear or socks. We spent four nights and five days at the family camp by the lake. It was such a wonderful, relaxing place. It has all the modern conveniences that would make an early pioneer woman jump for joy. Lucky me! This is the stove I get to cook on. Oops — this picture shows the stove needs polishing. Funny thing: the burners don't turn on by simply turning a dial. This is what we had to do so we could cook. We were completely out of wood. Of course, we also had to gather more wood for the fireplace in case we needed heat. After you prepare a lovely meal, you have to pump water and heat it on the stove to wash the dishes. Here's our "ice box" — just kidding. We do have electricity, so we have a modern refrigerator; the ice box is used as a pantry. We had the boys spend one night with us. They were able to spend two full days playing in the woods, swimming, going for boat rides, and fishing. We didn't catch anything big enough to keep, but this little guy was so cute I had to post the picture. On a side note, this is my husband's fishing boat — the boat he just had to have. In the five years we've had this boat, he has never caught a fish that was a keeper — not for lack of trying. Woo-hoo! I just got off the phone with my father. Today he received a letter in the mail notifying him that he won a sweepstakes. Lucky me; I know he will share the wealth. I don't know what I'm going to buy first — should it be jewelry, a new car, a trip? It really burns me that they market to senior citizens, raising their hopes with all the double talk. He wasn't listening; he won't be getting a check for hundreds of thousands of dollars. I tried to tell him not to send any money. At times he gets so angry at my mother if she explains that it is just a con to get his money; other times he understands. I have tried to explain different ways to my mother to make him think she sends the check, and she is getting much better at it. Our anniversary didn't turn out to be the wonderful romantic evening I had planned. My husband's cousin and fiancée came into town unexpectedly. They canoed on Lake Superior in the morning and met us at camp later in the day. We went for a paddle-boat ride on the lake and then went swimming. The rain only lasted for a minute. A quick change and we were on our way to dinner. We have a great lodge with good food only three minutes from camp. We'll have to make a date for the romantic dinner another time. Party chair Isn't this the cutest? This is one of the bar stools that they had painted for the patio. I wanted to get a picture of the bar, but didn't have a clear shot when I thought about it. They found four more stools at a garage sale that morning and are getting those done too. Tee shirt We had a grown-up night this past weekend. My husband and I were invited to my cousin's house for his and his girlfriend's second annual margarita party. They have a great backyard, lots of room; I'm not sure how many acres. Of course there was too much food and drink. I wish I could invite my own friends to this party. We don't really know very many people who get invited, except the family members, of course. Some of them I don't really talk to much. She bought him this new shirt for the occasion. I went to my parents' tonight to set up his computer again. He took it apart a few months back and moved it into another room in his basement.	I set up my father's computer a couple of weeks ago. He needed written directions to use Word. I had to fix the inevitable paper jam and explain the directions again. Update: He wrote a letter in Word to a friend from the Navy, printed it, put it in an envelope, and sent it via USPS. I saw the first attempt. I'm sure the recipient is wondering what's up. His spelling is atrocious; I did not remind him about spell check because it would have been too confusing. His words had numbers and stuff. The first line of the letter I saw, without the numbers, read: "Well, I'm back in business with email. My daughter got the computer running again for me." Early tomorrow morning the boys, soon to be three and six, will be coming to our house for a sleepover. Their mother, my daughter, is having her tonsils removed. After spending the last year with one throat infection after another, and ending a month ago with an abscess that wouldn't go away, the ear, nose, and throat specialist strongly urged her to do so. Part of her problem is that she is allergic to most antibiotics. Her regular doctor rarely prescribes antibiotics, preferring to let her fight off most infections naturally and save the drugs for emergencies. This last infection took two rounds of different drugs to finally cure her. The boys are more excited about the sleepover than having any sympathy for their mother. I'm sure they have big plans for the week. They were mad that it didn't start tonight. Auntie, my other daughter, gave them Spider-Man beds for the special sleepover when they were in Chicago visiting her a few weeks ago. Let's hope their parents have packed enough for this sleepover. Last time, one child had no underwear or socks. We spent four nights and five days at the family camp by the lake. It was such a wonderful, relaxing place. It has all the modern conveniences that would make an early pioneer woman jump for joy. Lucky me! This is the stove I get to cook on. Oops — this picture shows the stove needs polishing. Funny thing: the burners don't turn on by simply turning a dial. This is what we had to do so we could cook. We were completely out of wood. Of course, we also had to gather more wood for the fireplace in case we needed heat. After you prepare a lovely meal, you have to pump water and heat it on the stove to wash the dishes. Here's our "ice box" — just kidding. We do have electricity, so we have a modern refrigerator; the ice box is used as a pantry. We had the boys spend one night with us. They were able to spend two full days playing in the woods, swimming, going for boat rides, and fishing. We didn't catch anything big enough to keep, but this little guy was so cute I had to post the picture. On a side note, this is my husband's fishing boat — the boat he just had to have. In the five years we've had this boat, he has never caught a fish that was a keeper — not for lack of trying. Woo-hoo! I just got off the phone with my father. Today he received a letter in the mail notifying him that he won a sweepstakes. Lucky me; I know he will share the wealth. I don't know what I'm going to buy first — should it be jewelry, a new car, a trip? It really burns me that they market to senior citizens, raising their hopes with all the double talk. He wasn't listening; he won't be getting a check for hundreds of thousands of dollars. I tried to tell him not to send any money. At times he gets so angry at my mother if she explains that it is just a con to get his money; other times he understands. I have tried to explain different ways to my mother to make him think she sends the check, and she is getting much better at it. Our anniversary didn't turn out to be the wonderful romantic evening I had planned. My husband's cousin and fiancée came into town unexpectedly. They canoed on Lake Superior in the morning and met us at camp later in the day. We went for a paddle-boat ride on the lake and then went swimming. The rain only lasted for a minute. A quick change and we were on our way to dinner. We have a great lodge with good food only three minutes from camp. We'll have to make a date for the romantic dinner another time. Party chair Isn't this the cutest? This is one of the bar stools that they had painted for the patio. I wanted to get a picture of the bar, but didn't have a clear shot when I thought about it. They found four more stools at a garage sale that morning and are getting those done too. Tee shirt We had a grown-up night this past weekend. My husband and I were invited to my cousin's house for his and his girlfriend's second annual margarita party. They have a great backyard, lots of room; I'm not sure how many acres. Of course there was too much food and drink. I wish I could invite my own friends to this party. We don't really know very many people who get invited, except the family members, of course. Some of them I don't really talk to much. She bought him this new shirt for the occasion. I went to my parents' tonight to set up his computer again. He took it apart a few months back and moved it into another room in his basement.
long_en_365	poet_en	814	en	Bargain Bin Appraisal is a unique skill granted by an unknown source that allows enhanced perception. While the skill's normal state is passive and lets me perceive better, it can also be used actively, which consumes MP. Active use gives better descriptions, and the more unique the thing or concept I'm appraising, the higher the cost. I appraised the skill earlier, but now I understand the context for passive appraisal. Apparently the passive side of this skill does not consume MP; it also grants different perceptions than active appraisal. Great — I don't have to spend twenty years learning languages. Could that mean I can talk to animals and dragons if they exist? That's a scary thought. I'd be horribly awkward if I started meowing at a cat in front of people; I hope the ability doesn't force me to do that. I've determined that my Redescription ability probably grants me new skills. I have no idea what the catalyst is other than hair, because I can't yet appraise it. I might as well use it to spawn powers for now. I asked to remain outside the gate for the time being so I can explore my skills a bit more. Clyde reluctantly agreed to my proposal; it would have seemed suspicious for a no-name man to appear at the city gate expecting entry. When I returned to the city, I walked outside the gate for a few hours and discovered a forest. A forest! What splendor. I encountered some goblins there, but I beat them without using any skills. I had a distinct feeling that "Flame Finger" wouldn't help in battle and would just waste MP I was trying to regenerate. After about six hours, when the moon was at its highest (probably midnight), my MP reached 4,500. I felt like I could use Redescription again. My gamer's intuition told me the depressingly high creation cost of Flame Finger was probably related to its ability to evolve. I could see great potential in it, but I wouldn't use it for now. As I set out to create another ability, I thought of possibilities. Should I focus on another element? Could Redescription do more than create skills? I'd probably try that next. I moved my hand toward a tree and considered using Redescription to change it into an inanimate object — a sword? a magic staff? — the possibilities were endless. Holding my hand out, I settled on what to try. How exciting! I figured Redescription ran on Dungeon Master rules and could do almost anything; its power and cost were probably determined entirely by what I asked for, like wishes. "Become an infinite storage bag!" I demanded. As soon as I did, black lightning shot from my hand into the tree, gradually transforming it into a bag of pristine leather with a string strap—exactly as I had imagined. Even a strand of my hair was consumed in the process. I looked at the bag sitting on the floor. Is it infinite? Can such a little bag be infinite? I giggled and picked it up. A few hours later, after my MP recovered, I appraised it and strange text appeared: Bargain Bin Bag of Below a Basic Build's Ability [??MP] This bag is a pocket-dimension container capable of storing the weight of one tree: exactly 1,222 kg. Deposits and withdrawals are determined by the mass of the items and whether they are alive. What the heck is with that long name describing how the bag is terrible? I used the little magic I had left to rename the poorly named item "Magic Bag." What a pity—why did I pick such a tiny tree? It appears the 1,222 kg limit is based on the weight of the tree I converted. The phrase "Below a Basic Build's Ability," aside from being poetic, seems to mean the bag is pretty awful. I went to sleep intending to test it on another tree in the morning. When I woke the next morning, I was stunned. Shocked. Unraveled. Peeled. Shaven. Why don't I need to shave? Anyway, I tried using my roughly 4,500 MP on a tree again and a window popped up: "You can only use Redescription one time for similar effects." That window not only confirmed my belief that it could do almost anything, but also showed that its power was limited only by the uniqueness of its uses. Holy crap! If I'd spilled the beans about my ability to the people in the car, they probably would have discovered this earlier, which could have put me in serious danger. That's another reason I didn't tell them, and I'm extremely grateful I restrained myself. I wasn't lying; I simply didn't know.	Bargain Bin Appraisal is a unique skill granted by an unknown source that allows enhanced perception. While the skill's normal state is passive and lets me perceive better, it can also be used actively, which consumes MP. Active use gives better descriptions, and the more unique the thing or concept I'm appraising, the higher the cost. I appraised the skill earlier, but now I understand the context for passive appraisal. Apparently the passive side of this skill does not consume MP; it also grants different perceptions than active appraisal. Great — I don't have to spend twenty years learning languages. Could that mean I can talk to animals and dragons if they exist? That's a scary thought. I'd be horribly awkward if I started meowing at a cat in front of people; I hope the ability doesn't force me to do that. I've determined that my Redescription ability probably grants me new skills. I have no idea what the catalyst is other than hair, because I can't yet appraise it. I might as well use it to spawn powers for now. I asked to remain outside the gate for the time being so I can explore my skills a bit more. Clyde reluctantly agreed to my proposal; it would have seemed suspicious for a no-name man to appear at the city gate expecting entry. When I returned to the city, I walked outside the gate for a few hours and discovered a forest. A forest! What splendor. I encountered some goblins there, but I beat them without using any skills. I had a distinct feeling that "Flame Finger" wouldn't help in battle and would just waste MP I was trying to regenerate. After about six hours, when the moon was at its highest (probably midnight), my MP reached four thousand five hundred. I felt like I could use Redescription again. My gamer's intuition told me the depressingly high creation cost of Flame Finger was probably related to its ability to evolve. I could see great potential in it, but I wouldn't use it for now. As I set out to create another ability, I thought of possibilities. Should I focus on another element? Could Redescription do more than create skills? I'd probably try that next. I moved my hand toward a tree and considered using Redescription to change it into an inanimate object — a sword? a magic staff? — the possibilities were endless. Holding my hand out, I settled on what to try. How exciting! I figured Redescription ran on Dungeon Master rules and could do almost anything; its power and cost were probably determined entirely by what I asked for, like wishes. "Become an infinite storage bag!" I demanded. As soon as I did, black lightning shot from my hand into the tree, gradually transforming it into a bag of pristine leather with a string strap—exactly as I had imagined. Even a strand of my hair was consumed in the process. I looked at the bag sitting on the floor. Is it infinite? Can such a little bag be infinite? I giggled and picked it up. A few hours later, after my MP recovered, I appraised it and strange text appeared: Bargain Bin Bag of Below a Basic Build's Ability [??MP] This bag is a pocket-dimension container capable of storing the weight of one tree: exactly one thousand two hundred twenty two kg. Deposits and withdrawals are determined by the mass of the items and whether they are alive. What the heck is with that long name describing how the bag is terrible? I used the little magic I had left to rename the poorly named item "Magic Bag." What a pity—why did I pick such a tiny tree? It appears the one thousand two hundred twenty two kg limit is based on the weight of the tree I converted. The phrase "Below a Basic Build's Ability," aside from being poetic, seems to mean the bag is pretty awful. I went to sleep intending to test it on another tree in the morning. When I woke the next morning, I was stunned. Shocked. Unraveled. Peeled. Shaven. Why don't I need to shave? Anyway, I tried using my roughly four thousand five hundred MP on a tree again and a window popped up: "You can only use Redescription one time for similar effects." That window not only confirmed my belief that it could do almost anything, but also showed that its power was limited only by the uniqueness of its uses. Holy crap! If I'd spilled the beans about my ability to the people in the car, they probably would have discovered this earlier, which could have put me in serious danger. That's another reason I didn't tell them, and I'm extremely grateful I restrained myself. I wasn't lying; I simply didn't know.
long_en_349	poet_en	896	en	It was May 2001 when several visitors arrived from abroad to meet Adna. They were relatives of Adna’s biological mother, Liz, and it was the first time Adna would meet them. When they arrived, Adna learned that one of them was getting married; the wedding would be held in their paternal grandfather’s hometown on one of the country’s beautiful islands. They asked Adna to attend, but she declined: she couldn’t leave Aunt Felip, who was bedridden, and she couldn’t travel alone. They went to the island without her and returned to Adna’s home after the wedding. Although it was their first meeting, it felt as if they had known each other for a long time because they had always talked about Adna and wanted to see her in person. When Mom Rona was alive and Adna was still a baby, whenever a relative of Liz visited, Mom Rona would keep Adna away from them. They had no chance to get close to her and sometimes could only watch her from afar. When the relatives returned, they learned that Lex was the representative of their town in the church’s May festivity, competing against many others. It was a money contest—the highest total during the count would win. In the first count Lex was ahead, but in the final count another candidate had the higher total by only a small margin over Lex. One of Adna’s relatives brought Lex to the mall and bought her the most expensive gown, matched with a beautiful pair of high-heeled sandals. The sandals were set with glittering stones that would attract anyone who saw them. Lex’s gown was also beautiful and came with a shawl. Though Lex placed second, she was more stunning than the winner. It was also the first time Lex wore heavy makeup and had her hair done. The priest said to Azia, “It will be your turn next year.” Azia replied, “No—I don’t like to wear high-heeled sandals.” At the end of June, the older sister of Adna’s biological mother arrived with her husband, who was of a different nationality; he was a merchant seaman. Adna welcomed her aunt and uncle; they gave presents to everyone in the family, especially Lex, Azia, and Al the Second. Adna’s aunt told Al that her uncles on the island longed to see Adna. They were teachers and kept asking when they could see her. Adna was afraid to travel by ship, afraid of drowning like the time she and Uncle Mike were swimming at the beach—she would have drowned if she had not been able to hold on to Uncle Mike’s shorts. Adna’s aunt asked her to go with her to purchase a plane ticket. Adna insisted it be a round-trip ticket. Round-trip tickets were bought for the three of them: Adna, her aunt, and her aunt’s husband. It was Adna's first time traveling without her family; she had been compelled to go to fulfill the last wish of relatives on a distant island. She sat near the window at the back of a 19-seater plane and was amazed to see clouds like cotton. Looking down, the houses looked like matchboxes. In her mind she prayed, "God, please, if this plane must crash, let it be over land, not the sea — I don't know how to swim; let there be trees that might catch the plane." When she looked at the sea she imagined sharks, but later realized they were only small boats. At the airport someone was waiting for them with a vehicle. The place was beautiful, with pretty beaches along the way. When they arrived at the house where they would stay, Adna learned that the woman was a first-degree cousin of her mother and that her husband, a lawyer, held a high government position. Their garden was well landscaped, and the glass walls offered a view of it from inside the house. Adna couldn't believe how elaborate every meal was; she was encouraged to eat a lot. At home, Adna would buy two large fried chickens—enough for Al, Adna, Aunt Felip, Lex, Azia, Al the second, and the nanny. But in the house where they stayed, one large fried chicken was meant for each person. Every twenty minutes the chef served many dishes, and Adna complained that she couldn't eat it all. “I have helped so many hungry people and I know hunger is a big problem, but here I am with a big problem: how to finish my share of food,” she thought. A few days later Adna noticed she could no longer fit her pants; she had gained weight. It had been a nice tour of the island—a small place that could be seen in half a day. Adna bathed in the cold and hot springs and climbed part of an inactive volcano that had erupted long ago. There was a family reunion, and she met many old relatives, all retired teachers. One of them told her, “I'm ready to die now that I've seen you,” which deeply touched her. When it was time to say goodbye, Adna prepared to return to her family. Her aunt's husband noticed how much she missed them.	It was May two thousand one when several visitors arrived from abroad to meet Adna. They were relatives of Adna’s biological mother, Liz, and it was the first time Adna would meet them. When they arrived, Adna learned that one of them was getting married; the wedding would be held in their paternal grandfather’s hometown on one of the country’s beautiful islands. They asked Adna to attend, but she declined: she couldn’t leave Aunt Felip, who was bedridden, and she couldn’t travel alone. They went to the island without her and returned to Adna’s home after the wedding. Although it was their first meeting, it felt as if they had known each other for a long time because they had always talked about Adna and wanted to see her in person. When Mom Rona was alive and Adna was still a baby, whenever a relative of Liz visited, Mom Rona would keep Adna away from them. They had no chance to get close to her and sometimes could only watch her from afar. When the relatives returned, they learned that Lex was the representative of their town in the church’s May festivity, competing against many others. It was a money contest—the highest total during the count would win. In the first count Lex was ahead, but in the final count another candidate had the higher total by only a small margin over Lex. One of Adna’s relatives brought Lex to the mall and bought her the most expensive gown, matched with a beautiful pair of high-heeled sandals. The sandals were set with glittering stones that would attract anyone who saw them. Lex’s gown was also beautiful and came with a shawl. Though Lex placed second, she was more stunning than the winner. It was also the first time Lex wore heavy makeup and had her hair done. The priest said to Azia, “It will be your turn next year.” Azia replied, “No—I don’t like to wear high-heeled sandals.” At the end of June, the older sister of Adna’s biological mother arrived with her husband, who was of a different nationality; he was a merchant seaman. Adna welcomed her aunt and uncle; they gave presents to everyone in the family, especially Lex, Azia, and Al the Second. Adna’s aunt told Al that her uncles on the island longed to see Adna. They were teachers and kept asking when they could see her. Adna was afraid to travel by ship, afraid of drowning like the time she and Uncle Mike were swimming at the beach—she would have drowned if she had not been able to hold on to Uncle Mike’s shorts. Adna’s aunt asked her to go with her to purchase a plane ticket. Adna insisted it be a round-trip ticket. Round-trip tickets were bought for the three of them: Adna, her aunt, and her aunt’s husband. It was Adna's first time traveling without her family; she had been compelled to go to fulfill the last wish of relatives on a distant island. She sat near the window at the back of a nineteen-seater plane and was amazed to see clouds like cotton. Looking down, the houses looked like matchboxes. In her mind she prayed, "God, please, if this plane must crash, let it be over land, not the sea — I don't know how to swim; let there be trees that might catch the plane." When she looked at the sea she imagined sharks, but later realized they were only small boats. At the airport someone was waiting for them with a vehicle. The place was beautiful, with pretty beaches along the way. When they arrived at the house where they would stay, Adna learned that the woman was a first degree cousin of her mother and that her husband, a lawyer, held a high government position. Their garden was well landscaped, and the glass walls offered a view of it from inside the house. Adna couldn't believe how elaborate every meal was; she was encouraged to eat a lot. At home, Adna would buy two large fried chickens—enough for Al, Adna, Aunt Felip, Lex, Azia, Al the second, and the nanny. But in the house where they stayed, one large fried chicken was meant for each person. Every twenty minutes the chef served many dishes, and Adna complained that she couldn't eat it all. “I have helped so many hungry people and I know hunger is a big problem, but here I am with a big problem: how to finish my share of food,” she thought. A few days later Adna noticed she could no longer fit her pants; she had gained weight. It had been a nice tour of the island—a small place that could be seen in half a day. Adna bathed in the cold and hot springs and climbed part of an inactive volcano that had erupted long ago. There was a family reunion, and she met many old relatives, all retired teachers. One of them told her, “I'm ready to die now that I've seen you,” which deeply touched her. When it was time to say goodbye, Adna prepared to return to her family. Her aunt's husband noticed how much she missed them.
long_en_276	wiki_en	967	en	Indira Gandhi Institute of Technology (IGIT), Sarang, was established in 1982 and was initially managed directly by the Government of Odisha as the Odisha College of Engineering (OCE). Since 1981 the institute had operated as the Modern Polytechnic (MPT), offering diploma courses in Civil, Electrical, Mechanical, Mining, and Survey Engineering. In 1987 OCE and MPT were merged, renamed IGIT Sarang, and management was transferred to an autonomous society. The institute currently offers nine undergraduate engineering programs: Civil, Chemical, Electrical, Mechanical, Metallurgical, Electronics and Telecommunication, Computer Science Engineering, Production Engineering, and Architecture. It offers two part-time postgraduate engineering programs: Industrial Power Control and Drives, and Environmental Science and Engineering; nine full-time postgraduate programs: Computer Science Engineering, Electronics and Telecommunication Engineering, Geotechnical Engineering, Mechanical System Design, Metallurgy and Materials Engineering, Power Electronics and Drives, Power System Engineering, Production Engineering, and Structural Engineering; and a Master in Computer Applications (MCA). Additionally, it offers five diploma courses in Civil, Electrical, Electronics and Telecommunication, Mechanical, and Metallurgical Engineering. Currently, Dr. Satyabrata Mohanta is the director of the institute. Five undergraduate branches were accredited by the NBA in 2016. It is affiliated with Biju Patnaik University of Technology and the State Council for Technical Education and Vocational Training. In 2014, the Government of Odisha decided to confer unitary university status on IGIT. In 2017, the University Grants Commission granted autonomous status to the institution. IGIT was established in 1982 and was originally managed directly by the Government of Odisha as Odisha College of Engineering (OCE). Prior to that, since 1981, the institute—under the name Modern Polytechnic (MPT)—offered diploma courses in civil, electrical, mechanical, and mining survey engineering. In 1987, OCE and MPT merged and were renamed IGIT, Sarang, with management transferred to an autonomous society. It is the first government engineering college in Odisha to receive accreditation from the National Board of Accreditation (NBA) of the All India Council for Technical Education (AICTE). The institute is residential, with an integrated campus covering 179 acres that includes hostels, staff quarters, and the Dr. M. P. Mishra Memorial Stadium, featuring basketball, volleyball, and badminton courts. Facilities include a central library with 27,000 volumes, a central computer centre, a central workshop, a knowledge centre, and eight student hostels providing accommodation for over 1,300 students. The current annual intake is 1,200 students for B.Tech programs only (excluding M.Tech, M.Sc., and diploma programs). Other amenities include SBI (core banking facilities), a guesthouse, an on‑campus dispensary, a post office, a canteen, a gymnasium, and a students' and employees' co‑operative. The serene atmosphere at Sarang offers students an optimal opportunity to concentrate on their studies. Courses The institution offers undergraduate and postgraduate programmes in engineering and natural sciences. The following academic programmes are available at IGIT: Bachelor of Architecture (B.Arch), Bachelor of Technology (B.Tech), Master of Technology (M.Tech), Diploma in Engineering, Master of Computer Applications (MCA), Master of Science (M.Sc), and Doctor of Philosophy (Ph.D). Undergraduate programmes The institute offers 4‑year undergraduate degree programmes in the following disciplines: Chemical Engineering (accredited by NBA); Civil Engineering (accredited by NBA); Computer Science and Engineering; Electrical Engineering (accredited by NBA); Electronics and Telecommunication Engineering; Metallurgy and Material Engineering (accredited by NBA); Mechanical Engineering (accredited by NBA); Production Engineering; and Architecture. IGIT also offers a 3‑year B.Tech degree for diploma holders under the lateral entry scheme, with the approval of AICTE, New Delhi, recognised by the Government of Odisha and affiliated to Biju Patnaik University of Technology, Rourkela. Postgraduate courses: The institute offers the following postgraduate programs. Part-time: - Environmental Science (2.5 years) - Production Engineering (2 years) Full-time: - Electronics and Telecommunication Engineering - Geotechnical Engineering - Power Electronics and Drives Engineering - Power Systems Engineering - Production Engineering - Metallurgy and Materials Science Engineering - Industrial Metallurgy - Mechanical System Design Engineering - Structural Engineering - Master in Computer Applications (MCA) In addition, IGIT Sarang has been recognised by Utkal University, Vani Vihar, Bhubaneswar, Odisha as a centre of excellence in engineering research. Diploma courses: The institute offers a three-year diploma in the following disciplines: - Civil Engineering - Electrical Engineering - Electronics and Telecommunication Engineering - Metallurgy and Materials Engineering Central Library: The library has a collection of 27,000 text and reference books, handbooks, and Indian Standards (IS) codes. About 50 journals are subscribed to. The library's transaction service is fully automated. Student halls of residence: The institute is fully residential and provides hostel accommodation for all students. The institute has the following residential halls: - Akash Bhawan — all first-year boys (B.Tech) - Aryabhatta Bhawan — all second-year boys (B.Tech) - Bhaskar Bhawan — third-year boys (B.Tech/B.Arch) - Brahmos Bhawan — third-year, final-year and fifth-year boys (B.Tech/B.Arch) - Surya Bhawan — all final-year boys (B.Tech) - Agni Bhawan — all diploma boys - Rohini Bhawan — all girl students of B.Tech/B.Arch first and second year, M.Tech, MCA, and diploma - Prithvi Bhawan — all third-year and final-year girls (B.Tech/B.Arch) College festivals: - Horizon — the institute's annual techno-cultural fest - Technovation — techno-cultural fest of the Department of CSE&A, IGIT Sarang, held under the aegis of the Mycomp Society - Admantium — a national-level technical symposium held by the Department of Metallurgical and Materials Engineering every year since 2014 Student organisations: - Society for Physical Education & Recreation (SPER) - Social Service Guild (SSG) - IGIT Cultural Association - Mycomp Society - Society Of Literary Enthusiasts (SOLE) - Audio Visual Club - IGIT Robotics Society - National Cadet Corps (NCC) External links: - IGIT Sarang - IGIT Sarang Alumni Association - State Council for Technical Education & Vocational Training - All India Council for Technical Education Categories: - Engineering colleges in Odisha - Colleges affiliated with Biju Patnaik University of Technology - Dhenkanal district - Monuments and memorials to Indira Gandhi - Educational institutions established in 1982 - 1982 establishments in Orissa	Indira Gandhi Institute of Technology (IGIT), Sarang, was established in nineteen eighty two and was initially managed directly by the Government of Odisha as the Odisha College of Engineering (OCE). Since nineteen eighty one the institute had operated as the Modern Polytechnic (MPT), offering diploma courses in Civil, Electrical, Mechanical, Mining, and Survey Engineering. In nineteen eighty seven OCE and MPT were merged, renamed IGIT Sarang, and management was transferred to an autonomous society. The institute currently offers nine undergraduate engineering programs: Civil, Chemical, Electrical, Mechanical, Metallurgical, Electronics and Telecommunication, Computer Science Engineering, Production Engineering, and Architecture. It offers two part-time postgraduate engineering programs: Industrial Power Control and Drives, and Environmental Science and Engineering; nine full-time postgraduate programs: Computer Science Engineering, Electronics and Telecommunication Engineering, Geotechnical Engineering, Mechanical System Design, Metallurgy and Materials Engineering, Power Electronics and Drives, Power System Engineering, Production Engineering, and Structural Engineering; and a Master in Computer Applications (MCA). Additionally, it offers five diploma courses in Civil, Electrical, Electronics and Telecommunication, Mechanical, and Metallurgical Engineering. Currently, Dr. Satyabrata Mohanta is the director of the institute. Five undergraduate branches were accredited by the NBA in two thousand sixteen. It is affiliated with Biju Patnaik University of Technology and the State Council for Technical Education and Vocational Training. In two thousand fourteen, the Government of Odisha decided to confer unitary university status on IGIT. In two thousand seventeen, the University Grants Commission granted autonomous status to the institution. IGIT was established in nineteen eighty-two and was originally managed directly by the Government of Odisha as Odisha College of Engineering (OCE). Prior to that, since nineteen eighty-one, the institute—under the name Modern Polytechnic (MPT)—offered diploma courses in civil, electrical, mechanical, and mining survey engineering. In nineteen eighty-seven, OCE and MPT merged and were renamed IGIT, Sarang, with management transferred to an autonomous society. It is the first government engineering college in Odisha to receive accreditation from the National Board of Accreditation (NBA) of the All India Council for Technical Education (AICTE). The institute is residential, with an integrated campus covering one hundred seventy-nine acres that includes hostels, staff quarters, and the Dr. M. P. Mishra Memorial Stadium, featuring basketball, volleyball, and badminton courts. Facilities include a central library with twenty-seven thousand volumes, a central computer centre, a central workshop, a knowledge centre, and eight student hostels providing accommodation for over one thousand three hundred students. The current annual intake is one thousand two hundred students for B.Tech programs only (excluding M.Tech, M.Sc., and diploma programs). Other amenities include SBI (core banking facilities), a guesthouse, an on‑campus dispensary, a post office, a canteen, a gymnasium, and a students' and employees' co‑operative. The serene atmosphere at Sarang offers students an optimal opportunity to concentrate on their studies. Courses The institution offers undergraduate and postgraduate programmes in engineering and natural sciences. The following academic programmes are available at IGIT: Bachelor of Architecture (B.Arch), Bachelor of Technology (B.Tech), Master of Technology (M.Tech), Diploma in Engineering, Master of Computer Applications (MCA), Master of Science (M.Sc), and Doctor of Philosophy (Ph.D). Undergraduate programmes The institute offers four‑year undergraduate degree programmes in the following disciplines: Chemical Engineering (accredited by NBA); Civil Engineering (accredited by NBA); Computer Science and Engineering; Electrical Engineering (accredited by NBA); Electronics and Telecommunication Engineering; Metallurgy and Material Engineering (accredited by NBA); Mechanical Engineering (accredited by NBA); Production Engineering; and Architecture. IGIT also offers a three‑year B.Tech degree for diploma holders under the lateral entry scheme, with the approval of AICTE, New Delhi, recognised by the Government of Odisha and affiliated to Biju Patnaik University of Technology, Rourkela. Postgraduate courses: The institute offers the following postgraduate programs. Part-time: - Environmental Science (two point five years) - Production Engineering (two years) Full-time: - Electronics and Telecommunication Engineering - Geotechnical Engineering - Power Electronics and Drives Engineering - Power Systems Engineering - Production Engineering - Metallurgy and Materials Science Engineering - Industrial Metallurgy - Mechanical System Design Engineering - Structural Engineering - Master in Computer Applications (MCA) In addition, IGIT Sarang has been recognised by Utkal University, Vani Vihar, Bhubaneswar, Odisha as a centre of excellence in engineering research. Diploma courses: The institute offers a three-year diploma in the following disciplines: - Civil Engineering - Electrical Engineering - Electronics and Telecommunication Engineering - Metallurgy and Materials Engineering Central Library: The library has a collection of twenty seven thousand text and reference books, handbooks, and Indian Standards (IS) codes. About fifty journals are subscribed to. The library's transaction service is fully automated. Student halls of residence: The institute is fully residential and provides hostel accommodation for all students. The institute has the following residential halls: - Akash Bhawan — all first-year boys (B.Tech) - Aryabhatta Bhawan — all second-year boys (B.Tech) - Bhaskar Bhawan — third-year boys (B.Tech/B.Arch) - Brahmos Bhawan — third-year, final-year and fifth-year boys (B.Tech/B.Arch) - Surya Bhawan — all final-year boys (B.Tech) - Agni Bhawan — all diploma boys - Rohini Bhawan — all girl students of B.Tech/B.Arch first and second year, M.Tech, MCA, and diploma - Prithvi Bhawan — all third-year and final-year girls (B.Tech/B.Arch) College festivals: - Horizon — the institute's annual techno-cultural fest - Technovation — techno-cultural fest of the Department of CSE&A, IGIT Sarang, held under the aegis of the Mycomp Society - Admantium — a national-level technical symposium held by the Department of Metallurgical and Materials Engineering every year since two thousand fourteen Student organisations: - Society for Physical Education & Recreation (SPER) - Social Service Guild (SSG) - IGIT Cultural Association - Mycomp Society - Society Of Literary Enthusiasts (SOLE) - Audio Visual Club - IGIT Robotics Society - National Cadet Corps (NCC) External links: - IGIT Sarang - IGIT Sarang Alumni Association - State Council for Technical Education & Vocational Training - All India Council for Technical Education Categories: - Engineering colleges in Odisha - Colleges affiliated with Biju Patnaik University of Technology - Dhenkanal district - Monuments and memorials to Indira Gandhi - Educational institutions established in nineteen eighty two - nineteen eighty two establishments in Orissa.
long_en_178	paper_en	2,073	en	Section 2. Related Work Multimodal Instruction-following Agents. In computer vision, existing works that build instruction-following agents can be broadly categorized into two classes: First, end-to-end trained models, which are separately explored for each specific research topic. For example, the vision-language navigation task and Habitat require the embodied AI agent to follow natural language instructions and take a sequence of actions to complete goals in visual environments. In the image editing domain, given an input image and a written instruction that tells the agent what to do, InstructPix2Pix edits images by following the human instructions. Second, a system that coordinates various models via LangChain or LLMs, such as Visual ChatGPT, X-GPT, MM-REACT, VisProg, and ViperGPT. While sharing the same goal in building instruction-following agents, we focus on developing an end-to-end trained language-vision multimodal model for multiple tasks. Instruction Tuning. In the natural language processing (NLP) community, to enable LLMs such as GPT-3, T5, PaLM, and OPT to follow natural language instructions and complete real-world tasks, researchers have explored methods for LLM instruction-tuning, leading to instruction-tuned counterparts such as InstructGPT/ChatGPT, FLAN-T5, FLAN-PaLM, and OPT-IML, respectively. It turns out that this simple approach can effectively improve the zero- and few-shot generalization abilities of LLMs. It is thus natural to borrow the idea from NLP to computer vision. More broadly, the teacher-student distillation ideas with foundation models have been studied in other topics such as image classification. Flamingo can be viewed as the GPT-3 moment in the multimodal domain, due to its strong performance on zero-shot task transfer and in-context-learning. Other LMMs trained on image-text pairs include BLIP-2, FROMAGe, and KOSMOS-1. PaLM-E is an LMM for embodied AI. Based on the recent "best" open-source LLM LLaMA, OpenFlamingo and LLaMA-Adapter are open-source efforts that enable LLaMA to use image inputs, paving the way to build open-source multimodal LLMs. While these models present promising task transfer generalization performance, they are not explicitly tuned with vision-language instruction data, and their performance in multimodal tasks usually falls short compared to language-only tasks. In this paper, we aim to fill this gap and study its effectiveness. Finally, note that visual instruction tuning is different from visual prompt tuning: the former aims to improve the model's instruction-following abilities, while the latter aims to improve the parameter-efficiency in model adaptation. Section 3. GPT-assisted Visual Instruction Data Generation The community has witnessed a surge in the amount of public multimodal data such as image-text pairs, ranging from CC to LAION. However, when it comes to multimodal instruction-following data, the available amount is limited, partially because the process for creating such data is time-consuming and less well-defined when human crowd-scouring is considered. Inspired by the success of recent GPT models in text-annotation tasks, we propose to leverage ChatGPT/GPT-4 for multimodal instruction-following data collection, based on the widely existing image-pair data. For an image X_v and its associated caption X_c, it is natural to create a set of questions X_q with the intent to instruct the assistant to describe the image content. We prompt GPT-4 to curate such a list of questions. Therefore, a simple way to expand an image-text pair to its instruction-following version is: Human: X_q X_v <STOP> Assistant: X_c <STOP>. Though cheap to construct, this simple expanded version lacks diversity and in-depth reasoning in both the instructions and responses. To mitigate this issue, we leverage language-only GPT-4 or ChatGPT as the strong teacher (both accept only text as input), to create instruction-following data involving visual content. Specifically, in order to encode an image into its visual features to prompt a text-only GPT, we use two types of symbolic representations: First, Captions typically describe the visual scene from various perspectives. Second, Bounding boxes usually localize the objects in the scene, and each box encodes the object concept and its spatial location. This symbolic representation allows us to encode the image as an LLM-recognizable sequence. We use COCO images and generate three types of instruction-following data. For each type, we first manually design a few examples. They are the only human annotations we have during data collection, and are used as seed examples in in-context-learning to query GPT-4. The first type is Conversation. We design a conversation between the assistant and a person asking questions about this photo. The answers are in a tone as if the assistant is seeing the image and answering the question. A diverse set of questions are asked about the visual content of the image, including the object types, counting the objects, object actions, object locations, and relative positions between objects. Only questions that have definite answers are considered. The second type is Detailed description. To include a rich and comprehensive description for an image, we create a list of questions with such an intent. We prompt GPT-4 then curate the list. For each image, we randomly sample one question from the list to ask GPT-4 to generate the detailed description. The third type is Complex reasoning. The above two types focus on the visual content itself, based on which we further create in-depth reasoning questions. The answers typically require a step-by-step reasoning process by following rigorous logic. We collect 158K unique language-image instruction-following samples in total, including 58K in conversations, 23K in detailed description, and 77k in complex reasoning, respectively. We ablated the use of ChatGPT and GPT-4 in our early experiments, and found that GPT-4 consistently provides higher quality instruction-following data, such as spatial reasoning. Section 4. Visual Instruction Tuning Subsection 4.1. Architecture The primary goal is to effectively leverage the capabilities of both the pre-trained LLM and visual model. We choose Vicuna as our LLM, as it has the best instruction following capabilities in language tasks among publicly available checkpoints. For an input image X_v, we consider the pre-trained CLIP visual encoder ViT-L/14, which provides the visual feature Z_v equals g of X_v. The grid features before and after the last Transformer layer are considered in our experiments. We consider a simple linear layer to connect image features into the word embedding space. Specifically, we apply a trainable projection matrix W to convert Z_v into language embedding tokens H_v, which have the same dimensionality as the word embedding space in the language model. The equation is H_v equals W dot Z_v. Thus, we have a sequence of visual tokens H_v. Note that our simple projection scheme is lightweight, which allows us to iterate data centric experiments quickly. More sophisticated schemes to connect the image and language representations can also be considered, such as gated cross-attention in Flamingo and Q-former in BLIP-2. We leave exploring possibly more effective and sophisticated architecture designs for LLaVA as future work. Subsection 4.2. Training For each image X_v, we generate multi-turn conversation data. We organize them as a sequence, by treating all answers as the assistant's response. The instruction at the t-th turn, X_instruct^t, is defined as follows: for the first turn (t=1), we randomly choose between [X_q^1, X_v] or [X_v, X_q^1]. For all remaining turns (t>1), the instruction is simply X_q^t. This leads to a unified format for the multimodal instruction-following sequence. We perform instruction-tuning of the LLM on the prediction tokens, using its original auto-regressive training objective. Specifically, for a sequence of length L, we compute the probability of the target answers X_a. This probability, p(X_a \| X_v, X_instruct), is the product from i equals 1 to L of the probability of each token x_i, conditioned on the image X_v, and all preceding instruction and answer tokens. Here, theta represents the trainable parameters. For readability, we note that the image is grounded for all answers, and system messages and stop tokens are omitted from the notation. For LLaVA model training, we consider a two-stage instruction-tuning procedure. Stage 1: Pre-training for Feature Alignment. To strike a balance between concept coverage and training efficiency, we filter CC3M to 595K image-text pairs. These pairs are converted to the instruction-following data using the naive expansion method described earlier. Each sample can be treated as a single-turn conversation. To construct the input instruction, for an image X_v, a question X_q is randomly sampled, which is a language instruction to request the assistant to describe the image briefly. The ground-truth prediction answer X_a is the original caption. In training, we keep both the visual encoder and LLM weights frozen, and maximize the likelihood with trainable parameters being only the projection matrix W. In this way, the image features H_v can be aligned with the pre-trained LLM word embedding. This stage can be understood as training a compatible visual tokenizer for the frozen LLM. Stage 2: Fine-tuning End-to-End. We always keep the visual encoder weights frozen, and continue to update both the pre-trained weights of the projection layer and LLM in LLaVA; i.e., the trainable parameters are W and the LLM parameters phi. We consider two specific use case scenarios: First, Multimodal Chatbot. We develop a Chatbot by fine-tuning on the 158K language-image instruction-following data. Among the three types of responses, conversation is multi-turn while the other two are single-turn. They are uniformly sampled in training. Second, Science QA. We study our method on the ScienceQA benchmark, the first large-scale multimodal science question dataset that annotates the answers with detailed lectures and explanations. Each question is provided a context in the form of natural language or an image. The assistant provides the reasoning process in natural language and selects the answer among multiple choices. For training, we organize the data as a single turn conversation, the question and context as the instruction, and reasoning and answer as the response. Section 5. Experiments We assess the performance of LLaVA in instruction-following and visual reasoning capabilities with two primary experimental settings: multimodal chatbot and the ScienceQA dataset, respectively. We train all models with 8 A100s, following Vicuna's hyperparameters. We pre-train our model on the filtered CC-595K subset for 1 epoch with a learning rate of 2e-3 and a batch size of 128, and fine-tune on the proposed LLaVA-Instruct-158K dataset for 3 epochs, with a learning rate of 2e-5 and a batch size of 32. Multimodal Chatbot We developed a chatbot demo to show the image understanding and conversation abilities of LLaVA, and to study how well LLaVA is able to digest visual inputs and exhibit instruction-following capabilities. We first use the examples in the original GPT-4 paper. For comparisons, we quote the prompt and response of the multimodal GPT-4 from their paper, and query BLIP-2 and OpenFlamingo model checkpoints to get their response. Surprisingly, although LLaVA is trained with a small multimodal instruction-following dataset (around 80K unique images), it demonstrates quite similar reasoning results with multimodal GPT-4 on these examples. Note that while these images are out-of-domain for LLaVA, LLaVA is still able to understand the scenes and follow the question instruction to provide a reasonable response. In contrast, BLIP-2 and OpenFlamingo focus on describing the image, instead of following the user instruction to answer in an appropriate manner. Quantitative Evaluation. To gain a systematic understanding of the performance of LLaVA, we propose a quantitative metric to measure the model's instruction-following capability on multimodal data. Inspired by previous work, we leverage GPT-4 to measure the quality of generated responses. Specifically, we create triplets consisting of image, ground-truth textual descriptions, and question. The candidate models (e.g., LLaVA) predict the answers based on the question and the image. To provide an approximate theoretical upper bound, we create a reference prediction based on the question and the ground-truth textual descriptions, using the text-only GPT-4.	Section two. Related Work Multimodal Instruction-following Agents. In computer vision, existing works that build instruction-following agents can be broadly categorized into two classes: First, end-to-end trained models, which are separately explored for each specific research topic. For example, the vision-language navigation task and Habitat require the embodied AI agent to follow natural language instructions and take a sequence of actions to complete goals in visual environments. In the image editing domain, given an input image and a written instruction that tells the agent what to do, InstructPix two Pix two edits images by following the human instructions. Second, a system that coordinates various models via LangChain or LLMs, such as Visual ChatGPT, X-GPT, MM-REACT, VisProg, and ViperGPT. While sharing the same goal in building instruction-following agents, we focus on developing an end-to-end trained language-vision multimodal model for multiple tasks. Instruction Tuning. In the natural language processing (NLP) community, to enable LLMs such as GPT three, T five, PaLM, and OPT to follow natural language instructions and complete real-world tasks, researchers have explored methods for LLM instruction-tuning, leading to instruction-tuned counterparts such as InstructGPT/ChatGPT, FLAN-T five, FLAN-PaLM, and OPT-IML, respectively. It turns out that this simple approach can effectively improve the zero- and few-shot generalization abilities of LLMs. It is thus natural to borrow the idea from NLP to computer vision. More broadly, the teacher student distillation ideas with foundation models have been studied in other topics such as image classification. Flamingo can be viewed as the GPT three moment in the multimodal domain, due to its strong performance on zero shot task transfer and in context learning. Other LMMs trained on image text pairs include BLIP two, FROMAGe, and KOSMOS one. PaLM-E is an LMM for embodied AI. Based on the recent "best" open source LLM LLaMA, OpenFlamingo and LLaMA-Adapter are open source efforts that enable LLaMA to use image inputs, paving the way to build open source multimodal LLMs. While these models present promising task transfer generalization performance, they are not explicitly tuned with vision language instruction data, and their performance in multimodal tasks usually falls short compared to language only tasks. In this paper, we aim to fill this gap and study its effectiveness. Finally, note that visual instruction tuning is different from visual prompt tuning: the former aims to improve the model's instruction following abilities, while the latter aims to improve the parameter efficiency in model adaptation. Section three. GPT assisted Visual Instruction Data Generation The community has witnessed a surge in the amount of public multimodal data such as image text pairs, ranging from CC to LAION. However, when it comes to multimodal instruction following data, the available amount is limited, partially because the process for creating such data is time consuming and less well-defined when human crowd scouring is considered. Inspired by the success of recent GPT models in text annotation tasks, we propose to leverage ChatGPT slash GPT four for multimodal instruction following data collection, based on the widely existing image pair data. For an image X v and its associated caption X c, it is natural to create a set of questions X q with the intent to instruct the assistant to describe the image content. We prompt GPT four to curate such a list of questions. Therefore, a simple way to expand an image text pair to its instruction following version is: Human: X q X v <STOP> Assistant: X c <STOP>. Though cheap to construct, this simple expanded version lacks diversity and in depth reasoning in both the instructions and responses. To mitigate this issue, we leverage language only GPT four or ChatGPT as the strong teacher (both accept only text as input), to create instruction following data involving visual content. Specifically, in order to encode an image into its visual features to prompt a text only GPT, we use two types of symbolic representations: First, Captions typically describe the visual scene from various perspectives. Second, Bounding boxes usually localize the objects in the scene, and each box encodes the object concept and its spatial location. This symbolic representation allows us to encode the image as an LLM-recognizable sequence. We use COCO images and generate three types of instruction-following data. For each type, we first manually design a few examples. They are the only human annotations we have during data collection, and are used as seed examples in in-context-learning to query GPT four. The first type is Conversation. We design a conversation between the assistant and a person asking questions about this photo. The answers are in a tone as if the assistant is seeing the image and answering the question. A diverse set of questions are asked about the visual content of the image, including the object types, counting the objects, object actions, object locations, and relative positions between objects. Only questions that have definite answers are considered. The second type is Detailed description. To include a rich and comprehensive description for an image, we create a list of questions with such an intent. We prompt GPT four then curate the list. For each image, we randomly sample one question from the list to ask GPT-4 to generate the detailed description. The third type is Complex reasoning. The above two types focus on the visual content itself, based on which we further create in-depth reasoning questions. The answers typically require a step-by-step reasoning process by following rigorous logic. We collect one hundred fifty eight K unique language-image instruction-following samples in total, including fifty eight K in conversations, twenty three K in detailed description, and seventy seven k in complex reasoning, respectively. We ablated the use of ChatGPT and GPT four in our early experiments, and found that GPT four consistently provides higher quality instruction-following data, such as spatial reasoning. Section four. Visual Instruction Tuning Subsection four point one. Architecture The primary goal is to effectively leverage the capabilities of both the pre-trained LLM and visual model. We choose Vicuna as our LLM, as it has the best instruction following capabilities in language tasks among publicly available checkpoints. For an input image X v, we consider the pre-trained CLIP visual encoder ViT-L/14, which provides the visual feature Z v equals g of X v. The grid features before and after the last Transformer layer are considered in our experiments. We consider a simple linear layer to connect image features into the word embedding space. Specifically, we apply a trainable projection matrix W to convert Z v into language embedding tokens H v, which have the same dimensionality as the word embedding space in the language model. The equation is H v equals W dot Z v. Thus, we have a sequence of visual tokens H v. Note that our simple projection scheme is lightweight, which allows us to iterate data centric experiments quickly. More sophisticated schemes to connect the image and language representations can also be considered, such as gated cross-attention in Flamingo and Q-former in BLIP-2. We leave exploring possibly more effective and sophisticated architecture designs for LLaVA as future work. Subsection four point two. Training For each image X v, we generate multi-turn conversation data. We organize them as a sequence, by treating all answers as the assistant's response. The instruction at the t-th turn, X instruct t, is defined as follows: for the first turn (t equals one), we randomly choose between [X q one, X v] or [X v, X q one]. For all remaining turns (t is greater than one), the instruction is simply X q t. This leads to a unified format for the multimodal instruction-following sequence. We perform instruction-tuning of the LLM on the prediction tokens, using its original auto-regressive training objective. Specifically, for a sequence of length L, we compute the probability of the target answers X a. This probability, p(X a given X v, X instruct), is the product from i equals one to L of the probability of each token x i, conditioned on the image X v, and all preceding instruction and answer tokens. Here, theta represents the trainable parameters. For readability, we note that the image is grounded for all answers, and system messages and stop tokens are omitted from the notation. For LLaVA model training, we consider a two-stage instruction-tuning procedure. Stage one: Pre-training for Feature Alignment. To strike a balance between concept coverage and training efficiency, we filter CC3M to five hundred ninety five K image-text pairs. These pairs are converted to the instruction-following data using the naive expansion method described earlier. Each sample can be treated as a single-turn conversation. To construct the input instruction, for an image X v, a question X q is randomly sampled, which is a language instruction to request the assistant to describe the image briefly. The ground-truth prediction answer X a is the original caption. In training, we keep both the visual encoder and LLM weights frozen, and maximize the likelihood with trainable parameters being only the projection matrix W. In this way, the image features H v can be aligned with the pre-trained LLM word embedding. This stage can be understood as training a compatible visual tokenizer for the frozen LLM. Stage two: Fine-tuning End-to-End. We always keep the visual encoder weights frozen, and continue to update both the pre-trained weights of the projection layer and LLM in LLaVA; i.e., the trainable parameters are W and the LLM parameters phi. We consider two specific use case scenarios: First, Multimodal Chatbot. We develop a Chatbot by fine-tuning on the one hundred fifty eight K language-image instruction-following data. Among the three types of responses, conversation is multi-turn while the other two are single-turn. They are uniformly sampled in training. Second, Science QA. We study our method on the ScienceQA benchmark, the first large-scale multimodal science question dataset that annotates the answers with detailed lectures and explanations. Each question is provided a context in the form of natural language or an image. The assistant provides the reasoning process in natural language and selects the answer among multiple choices. For training, we organize the data as a single turn conversation, the question and context as the instruction, and reasoning and answer as the response. Section five. Experiments We assess the performance of LLaVA in instruction-following and visual reasoning capabilities with two primary experimental settings: multimodal chatbot and the ScienceQA dataset, respectively. We train all models with eight A one hundreds, following Vicuna's hyperparameters. We pre-train our model on the filtered CC five hundred ninety five K subset for one epoch with a learning rate of two e minus three and a batch size of one hundred twenty eight, and fine-tune on the proposed LLaVA Instruct one hundred fifty eight K dataset for three epochs, with a learning rate of two e minus five and a batch size of thirty two. Multimodal Chatbot We developed a chatbot demo to show the image understanding and conversation abilities of LLaVA, and to study how well LLaVA is able to digest visual inputs and exhibit instruction-following capabilities. We first use the examples in the original GPT four paper. For comparisons, we quote the prompt and response of the multimodal GPT four from their paper, and query BLIP two and OpenFlamingo model checkpoints to get their response. Surprisingly, although LLaVA is trained with a small multimodal instruction-following dataset (around eighty thousand unique images), it demonstrates quite similar reasoning results with multimodal GPT four on these examples. Note that while these images are out-of-domain for LLaVA, LLaVA is still able to understand the scenes and follow the question instruction to provide a reasonable response. In contrast, BLIP two and OpenFlamingo focus on describing the image, instead of following the user instruction to answer in an appropriate manner. Quantitative Evaluation. To gain a systematic understanding of the performance of LLaVA, we propose a quantitative metric to measure the model's instruction-following capability on multimodal data. Inspired by previous work, we leverage GPT four to measure the quality of generated responses. Specifically, we create triplets consisting of image, ground-truth textual descriptions, and question. The candidate models (e.g., LLaVA) predict the answers based on the question and the image. To provide an approximate theoretical upper bound, we create a reference prediction based on the question and the ground-truth textual descriptions, using the text-only GPT four.
long_en_272	wiki_en	957	en	Revolution By Night were an electronic music band based in London, England. Often referred to simply as RBN, the only continuous members were founding members Steve Weeks (vocals/keyboards/programming) and Bryon Adamson (keyboards/programming); Phil Eaton (keyboards) joined in 2006. Revolution By Night Mk I - 1994 to 1999 Revolution By Night were originally a four-piece band featuring vocalist Steve Weeks and bassist Bryon Adamson. They were heavily influenced by bands such as Fields of the Nephilim and The Sisters of Mercy. They recorded the album Breathe in 1994, which was released the following year. No further releases followed other than a handful of compilation tracks, and the group ceased activity in their then-current format in 1999. The band was originally known as Restoration II from 1992 until 1994, a continuation of Weeks’ previous band Restoration, with all new members (including Bryon Adamson, Eric Gandoin and Richard Pyne) sourced via an advert placed in the UK music newspaper Melody Maker. Richard Pyne from Restoration II went on to form the bands Killing Miranda and Uberbyte. The ex-members of Restoration went on to form goth rock act Vendemmian. Revolution By Night Mk II - 2000 to 2005 In late 1999, with previously unrealised ambitions to pursue electronic music, Weeks set up a small home studio and began working on new synth-based material with Bryon Adamson; together they developed a new, full-blooded electronic sound. The first fruits of their labour surfaced the following year at the Wave-Gotik-Treffen festival in Leipzig, where the band played a synthesiser-heavy set, debuting the future RBN anthem "Faithless," written by Weeks on a Novation Supernova and a Korg Trinity rack. Although Weeks and Adamson had been increasing the use of synthesisers in unreleased material for some time, the new sound was unexpected for some fans but was well received. In June 2003 the Faithless EP was released. Featuring remixes of the title track by Ronan Harris of VNV Nation and Tom Shear of Assemblage 23, the EP contained two other new songs, "Schadenfreude" and "Higher Ground (voxless)," and two completely reworked older songs. With the EP reaching No. 14 in the DAC (German Alternative Chart) and No. 3 in the Belgian Side-Line chart, the band again appeared at Wave-Gotik-Treffen in Leipzig. The VNV Nation remix became a dance-floor hit at the Slimelight club in London. RBN then spent most of 2004 and 2005 largely on hiatus; although new material was written during this time, it was subsequently scrapped. RBN (also known as Revolution By Night Mk III) was active from 2006 to 2012; Stok:holm/DKAG has been in effect from 2013 to the present. In 2006 the band relaunched, supporting Covenant on their UK tour and playing a set of almost entirely new songs. A steady following began to develop, mainly in the London scene, with RBN supporting many of the scene's well-known names. The following year, the band's remix of the Reaper track "X-Junkie" was released on Reaper's CD The Devil Is Female, which reached number one on the German DAC chart in January 2008. Writing continued that year for what would become the City Lights album. Pre-production studio work on the album was completed by late 2010 and handed over to producer Krischan Wesenberg of Rotersand; the finished album was slated for release in 2011. In November 2010, Weeks remixed Komor Kommando's track "Shrapnel". Apart from a couple of live dates in 2011, the band did little else. In November 2012, RBN announced that they had changed their name to Stok:holm and quietly released the much-delayed City Lights album on 1 January 2013 under the new name. Stok:holm have neither played live dates nor produced new music and appear to be on permanent hiatus. Since then, Weeks has worked with fellow London DJ Emmerick Gortz in the EBM/tech band DKAG; they released an EP titled "Short Wave" in 2015 and have played multiple live shows supporting bands such as Reaper, Suicide Commando, Phosgore, Combichrist, Apoptygma Berzerk, Velvet Acid Christ, Icon of Coil, and Aesthetic Perfection. Discography Revolution By Night (Mk I) - Breathe (1995, M&A Musicart) — album - "Touch" — track appeared on a compilation album - "Selling Heaven" — track appeared on a compilation album - "Kingdom Come" — track appeared on a Mission covers compilation album Revolution By Night (Mk II and Mk III) / Stok:holm / DKAG - Faithless EP (2003, Sonic-X) — includes remixes by Ronan Harris (VNV Nation) and Assemblage 23 - Zillo Club Hits 9 (2004, Zillo) — contains "Faithless (Ronan Harris remix)" - Into the Darkness Vol. 1 DVD (2004, CrazyClips) — includes the video for "Schadenfreude": https://www.youtube.com/watch?v=HOsIPcs7t1Q - Machine Code (2006, 5-track limited edition tour CD) — originally available only at RBN concerts - City Lights (2013, download-only, Sixty4bitMediaRelease) — all material written by Weeks and Adamson in 2006–2007, recorded by RBN 2007–2010, released under the band name Stok:holm and available on iTunes DKAG — Short Wave (2015, limited edition physical CD and digital release, Sixty4BitMediaRelease) Slimelight Muzik (14 October 2017, limited edition physical CD and digital release, Sixty4BitMediaRelease) Remixes for other artists: - Reaper — "The Devil Is Female" CD contains "X-Junkie (Revolution By Night remix)" (2007, release date Dec 2007, Infacted) - System:FX — "High:Definition:Violence" (2008) CD contains "Turn to Rust (RBN 'No Matter' remix)" - Straftanz — Straftanz UK (featuring RBN) (2009, unreleased apart from online streaming) - Nachtmahr — "Maedchen in Uniform" CD contains "Tanz Diktator (RBN remix)" (2009, release date Jan 2010, Trisol) - Komor Kommando — "Oil, Steel and Rhythm" 2 CD version contains "Shrapnel (RBN Mood Music mix)" (2010, release date Feb 2011, Alfa Matrix) References External links English electronic music groups English synth-pop groups	Revolution By Night were an electronic music band based in London, England. Often referred to simply as RBN, the only continuous members were founding members Steve Weeks (vocals/keyboards/programming) and Bryon Adamson (keyboards/programming); Phil Eaton (keyboards) joined in two thousand six. Revolution By Night Mk one - one thousand nine hundred ninety-four to one thousand nine hundred ninety-nine Revolution By Night were originally a four-piece band featuring vocalist Steve Weeks and bassist Bryon Adamson. They were heavily influenced by bands such as Fields of the Nephilim and The Sisters of Mercy. They recorded the album Breathe in one thousand nine hundred ninety-four, which was released the following year. No further releases followed other than a handful of compilation tracks, and the group ceased activity in their then-current format in one thousand nine hundred ninety-nine. The band was originally known as Restoration two from one thousand nine hundred ninety-two until one thousand nine hundred ninety-four, a continuation of Weeks’ previous band Restoration, with all new members (including Bryon Adamson, Eric Gandoin and Richard Pyne) sourced via an advert placed in the UK music newspaper Melody Maker. Richard Pyne from Restoration two went on to form the bands Killing Miranda and Uberbyte. The ex-members of Restoration went on to form goth rock act Vendemmian. Revolution By Night Mk two - two thousand to two thousand five In late one thousand nine hundred ninety-nine, with previously unrealised ambitions to pursue electronic music, Weeks set up a small home studio and began working on new synth-based material with Bryon Adamson; together they developed a new, full-blooded electronic sound. The first fruits of their labour surfaced the following year at the Wave-Gotik-Treffen festival in Leipzig, where the band played a synthesiser-heavy set, debuting the future RBN anthem "Faithless," written by Weeks on a Novation Supernova and a Korg Trinity rack. Although Weeks and Adamson had been increasing the use of synthesisers in unreleased material for some time, the new sound was unexpected for some fans but was well received. In June two thousand three the Faithless EP was released. Featuring remixes of the title track by Ronan Harris of VNV Nation and Tom Shear of Assemblage twenty three, the EP contained two other new songs, "Schadenfreude" and "Higher Ground (voxless)," and two completely reworked older songs. With the EP reaching No. fourteen in the DAC (German Alternative Chart) and No. three in the Belgian Side-Line chart, the band again appeared at Wave-Gotik-Treffen in Leipzig. The VNV Nation remix became a dance-floor hit at the Slimelight club in London. RBN then spent most of two thousand four and two thousand five largely on hiatus; although new material was written during this time, it was subsequently scrapped. RBN (also known as Revolution By Night Mk three) was active from two thousand six to two thousand twelve; Stok:holm/DKAG has been in effect from two thousand thirteen to the present. In two thousand six the band relaunched, supporting Covenant on their UK tour and playing a set of almost entirely new songs. A steady following began to develop, mainly in the London scene, with RBN supporting many of the scene's well-known names. The following year, the band's remix of the Reaper track "X-Junkie" was released on Reaper's CD The Devil Is Female, which reached number one on the German DAC chart in January two thousand eight. Writing continued that year for what would become the City Lights album. Pre-production studio work on the album was completed by late two thousand ten and handed over to producer Krischan Wesenberg of Rotersand; the finished album was slated for release in two thousand eleven. In November two thousand ten, Weeks remixed Komor Kommando's track "Shrapnel". Apart from a couple of live dates in two thousand eleven, the band did little else. In November two thousand twelve, RBN announced that they had changed their name to Stok:holm and quietly released the much-delayed City Lights album on one January two thousand thirteen under the new name. Stok:holm have neither played live dates nor produced new music and appear to be on permanent hiatus. Since then, Weeks has worked with fellow London DJ Emmerick Gortz in the EBM/tech band DKAG; they released an EP titled "Short Wave" in two thousand fifteen and have played multiple live shows supporting bands such as Reaper, Suicide Commando, Phosgore, Combichrist, Apoptygma Berzerk, Velvet Acid Christ, Icon of Coil, and Aesthetic Perfection. Discography Revolution By Night (Mk one) - Breathe (nineteen ninety five, M and A Musicart) — album - "Touch" — track appeared on a compilation album - "Selling Heaven" — track appeared on a compilation album - "Kingdom Come" — track appeared on a Mission covers compilation album Revolution By Night (Mk two and Mk three) / Stok:holm / DKAG - Faithless EP (two thousand three, Sonic-X) — includes remixes by Ronan Harris (VNV Nation) and Assemblage twenty three - Zillo Club Hits nine (two thousand four, Zillo) — contains "Faithless (Ronan Harris remix)" - Into the Darkness Vol. one DVD (two thousand four, CrazyClips) — includes the video for "Schadenfreude": https colon slash slash www dot youtube dot com slash watch question mark v equals HOsIPcs7t1Q - Machine Code (two thousand six, five-track limited edition tour CD) — originally available only at RBN concerts - City Lights (two thousand thirteen, download-only, Sixty4bitMediaRelease) — all material written by Weeks and Adamson in two thousand six–two thousand seven, recorded by RBN two thousand seven–two thousand ten, released under the band name Stok:holm and available on iTunes DKAG — Short Wave (two thousand fifteen, limited edition physical CD and digital release, Sixty4BitMediaRelease) Slimelight Muzik (fourteen October two thousand seventeen, limited edition physical CD and digital release, Sixty4BitMediaRelease) Remixes for other artists: - Reaper — "The Devil Is Female" CD contains "X-Junkie (Revolution By Night remix)" (two thousand seven, release date Dec two thousand seven, Infacted) - System:FX — "High:Definition:Violence" (two thousand eight) CD contains "Turn to Rust (RBN 'No Matter' remix)" - Straftanz — Straftanz UK (featuring RBN) (two thousand nine, unreleased apart from online streaming) - Nachtmahr — "Maedchen in Uniform" CD contains "Tanz Diktator (RBN remix)" (two thousand nine, release date Jan two thousand ten, Trisol) - Komor Kommando — "Oil, Steel and Rhythm" two CD version contains "Shrapnel (RBN Mood Music mix)" (two thousand ten, release date Feb two thousand eleven, Alfa Matrix) References External links English electronic music groups English synth-pop groups.
long_en_142	paper_en	2,446	en	To achieve this, we employ a conditional flow matching (CFM) model to sample the Mel spectrogram, given speech tokens, reference speech and speaker embedding as conditions. In the CFM model, the distribution of target Mel spectrogram is described by a probability density path from a prior distribution and the data distribution. For sampling efficiency, we employ the optimal-transport (OT) flow to match the vector field, which is given by an ordinary differential equation (ODE). A causal convolutional Transformer UNet is employed to learn the ODE with the up-sampled token, masked Mel spectrogram, speaker embedding and timestep as the conditions. At the training stage, the masked Mel spectrogram is obtained by randomly masking out 70% to 100% of the final frames in the target Mel spectrogram. As for the inference, it is provided by the Mel spectrogram extracted from the reference speech. By minimizing the L1 loss between the predicted and ground-truth ODE, we can optimize the UNet parameters. At the training stage, the timestep follows a uniform distribution. However, during the inference, we employ the cosine scheduler to offer more steps for the initial generation stage. Besides, we also train the model on both conditional and non-conditional situations to enable the classifier-free guidance (CFG) at the inference stage. The CFG strength and the number of flow estimation (NFE) are set to 0.7 and 10, respectively, according to the experimental results. The current flow matching models always work on a offline mode, i.e., only all the speech tokens are generated, the Mel spectragram can be sampled, which is not friendly for the streaming synthesis. To overcome this issue, we treat the multi-step flow estimation as a stacked deeper neural network, which repeats the UNet ten times. Thus, by making the unfolded neural network causal, we can apply it on the streaming synthesis. We construct four masks to satisfy different application situations: Non-causal Mask, Full-causal Mask, Chunk-M Mask, and Chunk-2M Mask. For each training case in a mini-batch, we randomly sample a mask from the above four masks under the uniform distribution. In this manner, one flow matching model can be compatible to different scenarios, lowering the deployment complexity. Another advantage of this chunk-aware training is that the masks with more context sever as a teacher for the ones with less context, benefiting from the implicit self-distillation scheme. Latency Analysis for Streaming Mode The first-package latency is an important metric for streaming synthesis models, which significantly affects the user experience especially in LLM-based voice chat applications. In the context of TTS, the to-synthesize text is known in advance, and the latency comes from the aspects of speech token generation, Mel spectrogram reconstruction and waveform synthesis. Thus, the first-package latency of CosyVoice 2 can be obtained by summing three components: M times the computation time of the LM to generate one speech token, M times the computation time of the Flow Matching model to generate the Mel spectrogram frames for one speech token, and M times the computation time of the vocoder to synthesize waveforms for one speech token. In the context of LLM-based voice chat, the length of first-package-required text should also be considered. The first-package latency becomes less than or equal to the sum of N times the computation time of a LLM to generate one text token, plus the TTS latency. Note that, since the multi-character tokens are masked out in CosyVoice 2's text tokenizer, the text tokens used by text LLMs always encode longer raw text than those of CosyVoice 2. Thus, the the first-package latency must be lower than the calculated sum. Instructed Generation To enhance the controllability of CosyVoice 2, we integrated the instructed dataset into the base training set. We have collected 1500 hours of instructed training data, which includes both natural language instructions and fine-grained instructions. For natural language instructions, we prepend a natural language description and a special end token, ``<\|endofprompt\|>'' before the to-synthesize input text. These descriptions cover aspects such as emotion, speaking rate, role-playing, and dialects. Emotions include Happy, Sad, Surprised, Angry, Fearful, Disgusted, Calm, and Serious. Speaking rates include Fast, Very Fast, Slow, and Very Slow. Dialects include Cantonese, Sichuan, Shanghai, Zhengzhou, Changsha, and Tianjin. Role-playing examples include Mysterious, Fierce, Curious, Elegant, Lonely, Robot, and Peppa Pig. For fine-grained instructions, we insert vocal bursts between text tokens, using markers like ``[laughter]'' and ``[breath]''. Additionally, we apply vocal feature tags to phrases; for instance, ``<strong>XXX</strong>'' indicates emphasis on certain words, while ``<laughter>XXX</laughter>'' signifies speaking with laughter. Multi-Speaker Fine-tuning Fine-tuning the pre-trained model on specific speakers (SFT) can further improve the generation quality and speaker similarity. In this report, we introduce the multi-speaker fine-tuning (mSFT), in which the pretrained model is fine-tuned on multiply speakers simultaneously rather than a single speaker. This approach ensures comprehensive prosody and pronunciation coverage across multiple speakers and mitigates potential catastrophic forgetting from the pretrained models. To avoid timbre confusion between various speakers, we prepend speaker-prompt tags, like ``Speaker A<\|endofprompt\|>'' to the input text for a specific speaker. If a training sample is not labeled to a speaker, a special tag, ``unknown<\|endofprompt\|>'', is utlized. The learning rate is set to 1e-5 during the whole multi-speaker fine-tuning process. Reinforcement Learning for SFT Reinforcement learning is a commonly used method in the training of large language models, which can make the LM output align with human preference. In CosyVoice 2, we employ speaker similarity (SS) and recognition word error rate (WER) from the ASR system as the reward function to improve speaker similarity and pronunciation accuracy in the fine-tuning stage. We use WER and SS to distinguish preferred samples and rejected samples and optimize the TTS system with direct preference optimization (DPO). However, this method is time-consuming and computation-consuming as it should synthesis the audios through the TTS system repeatedly to obtain distinguishable preference and rejected samples. During training, four forward operations are needed for one training step. To simplify the process, we recover the LM predicted token into quantized low-rank representations, and directly use the ASR backend of the speech tokenizer to re-predict the input text. Then the predicted log posterior can be regarded as the ASR reward function to optimize the text-speech language model. During training, the ASR backend parameters are frozen. As the sample operation of the tokens still prevent us to optimize the model directly, we use the gumbel softmax sampling to make it differentiated and then optimize the language model by the ASR loss. Experimental Settings Training Data for Speech Tokenizer A 200,000-hour dataset is used to train the speech tokenizer with normalized transcriptions as labels. The training data comes from three different resources: open source ASR datasets, internal industrial datasets and TTS generation datasets. The dataset includes 110,884 hours of Chinese and 99,918 hours of English data. Although we only used Chinese and English data when training the speech tokenizer, subsequent experiments revealed that the speech tokenizer had zero-shot capability for other languages. It can be also used for speech synthesis in languages such as Japanese and Korean. Training Data for CosyVoice 2 CosyVoice 2 shares the same training data as its previous version. We first collect the speech-only data with internal speech processing tools. Subsequently, the Paraformer and SenseVoice are employed to generate pseudo text labels for Chinese and other languages, respectively. We also employ an internal force-alignment model to filter out low-quality data and enhances the accuracy of punctuation. Data details are 130,000 hours for Chinese, 30,000 for English, 4,600 for Japanese, and 2,200 for Korean. Evaluation Settings We evaluate our CosyVoice 2 on two test sets. The first one is constructed from the test-clean set of Librispeech corpus, denoted as test-clean. This test set is used to evaluate CosyVoice 2 on a limited English domain. The Whisper-large V3 is used as the ASR model to evaluate the content consistency. As for the speaker similarity (SS), we employ the ERes2Net model to extract speaker embeddings of prompt and generated utterances, and their raw cosine similarity is treated as the speaker similarity. NMOS score is used to evaluate the objective quality. The second evaluation is conducted on the SEED test sets, which is widely used to evaluate recent TTS models, covering various text domains and reference speeches. In this evaluation, about 2,000 Chinese and 1,000 English samples are selected from CommonVoice datasets, denoted as test-zh and test-en, respectively. In addition, about 400 hard test cases are also included to evaluate the robustness of TTS models on text repetition, tongue twister and other challenging synthesis cases, denoted as test-hard in this report. The Paraformer is employed to recognize the synthesis results of test-zh and test-hard, while the Whisper-large V3 is adopted for test-en to evaluate the content consistency. We adopt two speaker verification (SV) models to evaluate speaker similarity: WavLM-finetuned SV model and ERes2Net. Benchmark for Japanese and Korean We prepare two test sets, denoted as test-ja and test-ko, for the evaluation on Japanese and Korean speech synthesis. The test-ja consists 1,000 samples extracted from the CommonVoice dataset, which are used to measure the model’s performance on various metrics, such as WER, SS, MOS. Specifically, we randomly shuffle and pair the entire CommonVoice JA-test set as reference utterance and target utterance spoken. Considering the wide range of utterances' text lengths of JA-test set, we randomly selected 1,000 pairs of reference-target utterances from the length range from 8 to 32 characters as our final test set. For the test-ko, we selected 1,000 speech samples with a WER of less than 5% and no deletion or insertion errors, utilizing the Whisper-Large V3 as the ASR model. These samples were used as reference utterances for the Korean speech synthesis. For the input text, we randomly selected 1,000 text samples from the remaining data. We have released the lists of prompt speeches, prompt transcriptions and input text from these two test sets are released to facilitate result reproduction. By providing this open-source data, we aim to establish a benchmark for evaluating Japanese and Korean TTS models. The Whisper-large V3 is used as the ASR model for Japanese and Korean evaluations. Experimental Results Evaluations on Speech Tokenizer An ideal speech tokenizer is supposed to effectively utilizes the codebook, preserves information at a high fidelity, and demonstrates speaker independence. In this part we evaluate our supervised speech tokenizer from four aspects: 1) Codebook utilization rate; 2) ASR error rate within the entire encoder; 3) Token visualization of different speakers; 4) Speaker identification training. It turns out that the FSQ-based tokenizer fully utilizes the codebook and maintains more effective information from the aspect of ASR, indicating more semantic information maintained by FSQ. We further analyze the characteristics of FSQ through the t-SNE visualization. As an upstream model for TTS tasks, the tokenizer should strive to minimize the entanglement of speaker identity information with the speech signal. We selected 100 speech samples from each of the three speakers in the VoxCeleb1 dataset and visualized the corresponding tokens. It is evident that before the quantization, Encoder 1's outputs exhibit different distributions among different speakers. In contrast, the distributions of quantized representations are nearly indistinguishable. In addition, the tokenizer fully utilizes the codebook. Subsequently, the S3prl toolkit is employed to further evaluate the speaker entanglement by performing speaker identification (SID) task. We use Sensevoice-large encoder with FSQ as an upstream feature extractor and train SID task with representations before or after the quantization. The convergence curves of SID training show that the SID layer with quantized tokens does not converge, which proves the decoupling function of the tokenizer on speaker information. Comparison Results with Baselines We first evaluated our CosyVoice 2 models on a limited English text domain and compared it with several open-source models, such as ChatTTS, GPT-SoVITs, OpenVoice, ParlerTTS, EmotiVoice, and its predecessor CosyVoice. The objective results include content consistency (WER), speech quality (NMOS) and speaker similarity (SS). The results show that CosyVoice 2 achieves state-of-the-art performance on the Librispeech test-clean set, surpassing all baseline models across all evaluation metrics. Notably, CosyVoice 2 even demonstrates higher content consistency, speech quality, and speaker similarity than human utterances, indicating its human-parity synthesis quality. We also evaluated CosyVoice 2 on the commonly-used test sets: SEED test-zh, test-en and test-hard, which include diverse input texts and reference speeches from various domains. On the test-zh set, CosyVoice 2 surpasses all open-sourced models in terms of CER and SS, falling short of the commercial model SEED-TTS by only a small margin. On the test-en set, CosyVoice 2 ranks fourth and third in terms of WER and SS, respectively. This may result from the imbalance in the volume of training data between Chinese and English. We plan to explore data scaling in future work to enhance content consistency in English. On the test-hard set, the offline CosyVoice 2 model achieves state-of-the-art performance across all compared baseline, demonstrating its robustness in challenging synthesis scenarios. Compared with human-generated speeches, CosyVoice 2 shows comparable content consistency and superior speaker similarity. Considering the recognition errors can also stem from the ASR model, it is reasonable to conclude that CosyVoice 2 achieves a human-parity synthesis capability. We also evaluated the streaming mode, denoted as ``CosyVoice 2-S''. For both evaluation settings, the streaming mode's performance is nearly lossless in typical test cases. Only in challenging cases is there a slight degradation in content consistency, highlighting the strength of our unified streaming/non-streaming framework. We found that the results of speaker similarity are not consistent on different SV models. This may indicate a new research topic on how to evaluate speaker similarity for TTS models automatically. Since different TTS models may use different SV models to extract speaker information, evaluating speaker similarity with the same SV model allows a more accurate evaluation on the utilization of speaker information. Therefore, we employ ERes2Net for evaluating speaker similarity in subsequent experiments. Modular Ablation Study We conducted a modular ablation study on the text-speech language model to assess the impacts of our modifications, including LLM initialization, removing speaker embedding, and utilizing FSQ.	To achieve this, we employ a conditional flow matching (CFM) model to sample the Mel spectrogram, given speech tokens, reference speech and speaker embedding as conditions. In the CFM model, the distribution of target Mel spectrogram is described by a probability density path from a prior distribution and the data distribution. For sampling efficiency, we employ the optimal-transport (OT) flow to match the vector field, which is given by an ordinary differential equation (ODE). A causal convolutional Transformer UNet is employed to learn the ODE with the up-sampled token, masked Mel spectrogram, speaker embedding and timestep as the conditions. At the training stage, the masked Mel spectrogram is obtained by randomly masking out seventy percent to one hundred percent of the final frames in the target Mel spectrogram. As for the inference, it is provided by the Mel spectrogram extracted from the reference speech. By minimizing the L one loss between the predicted and ground-truth ODE, we can optimize the UNet parameters. At the training stage, the timestep follows a uniform distribution. However, during the inference, we employ the cosine scheduler to offer more steps for the initial generation stage. Besides, we also train the model on both conditional and non-conditional situations to enable the classifier-free guidance (CFG) at the inference stage. The CFG strength and the number of flow estimation (NFE) are set to zero point seven and ten, respectively, according to the experimental results. The current flow matching models always work on a offline mode, i.e., only all the speech tokens are generated, the Mel spectragram can be sampled, which is not friendly for the streaming synthesis. To overcome this issue, we treat the multi-step flow estimation as a stacked deeper neural network, which repeats the UNet ten times. Thus, by making the unfolded neural network causal, we can apply it on the streaming synthesis. We construct four masks to satisfy different application situations: Non-causal Mask, Full-causal Mask, Chunk M Mask, and Chunk two M Mask. For each training case in a mini-batch, we randomly sample a mask from the above four masks under the uniform distribution. In this manner, one flow matching model can be compatible to different scenarios, lowering the deployment complexity. Another advantage of this chunk-aware training is that the masks with more context sever as a teacher for the ones with less context, benefiting from the implicit self-distillation scheme. Latency Analysis for Streaming Mode The first-package latency is an important metric for streaming synthesis models, which significantly affects the user experience especially in LLM-based voice chat applications. In the context of TTS, the to-synthesize text is known in advance, and the latency comes from the aspects of speech token generation, Mel spectrogram reconstruction and waveform synthesis. Thus, the first-package latency of CosyVoice two can be obtained by summing three components: M times the computation time of the LM to generate one speech token, M times the computation time of the Flow Matching model to generate the Mel spectrogram frames for one speech token, and M times the computation time of the vocoder to synthesize waveforms for one speech token. In the context of LLM-based voice chat, the length of first-package-required text should also be considered. The first-package latency becomes less than or equal to the sum of N times the computation time of a LLM to generate one text token, plus the TTS latency. Note that, since the multi-character tokens are masked out in CosyVoice two's text tokenizer, the text tokens used by text LLMs always encode longer raw text than those of CosyVoice two. Thus, the the first-package latency must be lower than the calculated sum. Instructed Generation To enhance the controllability of CosyVoice two, we integrated the instructed dataset into the base training set. We have collected one thousand five hundred hours of instructed training data, which includes both natural language instructions and fine-grained instructions. For natural language instructions, we prepend a natural language description and a special end token, ``<\|endofprompt\|>'' before the to-synthesize input text. These descriptions cover aspects such as emotion, speaking rate, role-playing, and dialects. Emotions include Happy, Sad, Surprised, Angry, Fearful, Disgusted, Calm, and Serious. Speaking rates include Fast, Very Fast, Slow, and Very Slow. Dialects include Cantonese, Sichuan, Shanghai, Zhengzhou, Changsha, and Tianjin. Role-playing examples include Mysterious, Fierce, Curious, Elegant, Lonely, Robot, and Peppa Pig. For fine-grained instructions, we insert vocal bursts between text tokens, using markers like [laughter] and [breath]. Additionally, we apply vocal feature tags to phrases; for instance, strong XXX strong indicates emphasis on certain words, while laughter XXX laughter signifies speaking with laughter. Multi-Speaker Fine-tuning Fine-tuning the pre-trained model on specific speakers (SFT) can further improve the generation quality and speaker similarity. In this report, we introduce the multi-speaker fine-tuning (mSFT), in which the pretrained model is fine-tuned on multiply speakers simultaneously rather than a single speaker. This approach ensures comprehensive prosody and pronunciation coverage across multiple speakers and mitigates potential catastrophic forgetting from the pretrained models. To avoid timbre confusion between various speakers, we prepend speaker-prompt tags, like Speaker A endofprompt to the input text for a specific speaker. If a training sample is not labeled to a speaker, a special tag, unknown endofprompt, is utlized. The learning rate is set to one e minus five during the whole multi-speaker fine-tuning process. Reinforcement Learning for SFT Reinforcement learning is a commonly used method in the training of large language models, which can make the LM output align with human preference. In CosyVoice two, we employ speaker similarity (SS) and recognition word error rate (WER) from the ASR system as the reward function to improve speaker similarity and pronunciation accuracy in the fine-tuning stage. We use WER and SS to distinguish preferred samples and rejected samples and optimize the TTS system with direct preference optimization (DPO). However, this method is time-consuming and computation-consuming as it should synthesis the audios through the TTS system repeatedly to obtain distinguishable preference and rejected samples. During training, four forward operations are needed for one training step. To simplify the process, we recover the LM predicted token into quantized low-rank representations, and directly use the ASR backend of the speech tokenizer to re-predict the input text. Then the predicted log posterior can be regarded as the ASR reward function to optimize the text-speech language model. During training, the ASR backend parameters are frozen. As the sample operation of the tokens still prevent us to optimize the model directly, we use the gumbel softmax sampling to make it differentiated and then optimize the language model by the ASR loss. Experimental Settings Training Data for Speech Tokenizer A two hundred thousand-hour dataset is used to train the speech tokenizer with normalized transcriptions as labels. The training data comes from three different resources: open source ASR datasets, internal industrial datasets and TTS generation datasets. The dataset includes one hundred ten thousand eight hundred eighty-four hours of Chinese and ninety-nine thousand nine hundred eighteen hours of English data. Although we only used Chinese and English data when training the speech tokenizer, subsequent experiments revealed that the speech tokenizer had zero-shot capability for other languages. It can be also used for speech synthesis in languages such as Japanese and Korean. Training Data for CosyVoice two CosyVoice two shares the same training data as its previous version. We first collect the speech-only data with internal speech processing tools. Subsequently, the Paraformer and SenseVoice are employed to generate pseudo text labels for Chinese and other languages, respectively. We also employ an internal force-alignment model to filter out low-quality data and enhances the accuracy of punctuation. Data details are one hundred thirty thousand hours for Chinese, thirty thousand for English, four thousand six hundred for Japanese, and two thousand two hundred for Korean. Evaluation Settings We evaluate our CosyVoice two on two test sets. The first one is constructed from the test-clean set of Librispeech corpus, denoted as test-clean. This test set is used to evaluate CosyVoice two on a limited English domain. The Whisper-large V three is used as the ASR model to evaluate the content consistency. As for the speaker similarity (SS), we employ the ERes two Net model to extract speaker embeddings of prompt and generated utterances, and their raw cosine similarity is treated as the speaker similarity. NMOS score is used to evaluate the objective quality. The second evaluation is conducted on the SEED test sets, which is widely used to evaluate recent TTS models, covering various text domains and reference speeches. In this evaluation, about two thousand Chinese and one thousand English samples are selected from CommonVoice datasets, denoted as test-zh and test-en, respectively. In addition, about four hundred hard test cases are also included to evaluate the robustness of TTS models on text repetition, tongue twister and other challenging synthesis cases, denoted as test-hard in this report. The Paraformer is employed to recognize the synthesis results of test-zh and test-hard, while the Whisper-large V three is adopted for test-en to evaluate the content consistency. We adopt two speaker verification (SV) models to evaluate speaker similarity: WavLM-finetuned SV model and ERes two Net. Benchmark for Japanese and Korean We prepare two test sets, denoted as test-ja and test-ko, for the evaluation on Japanese and Korean speech synthesis. The test-ja consists one thousand samples extracted from the CommonVoice dataset, which are used to measure the model’s performance on various metrics, such as WER, SS, MOS. Specifically, we randomly shuffle and pair the entire CommonVoice JA-test set as reference utterance and target utterance spoken. Considering the wide range of utterances' text lengths of JA-test set, we randomly selected one thousand pairs of reference-target utterances from the length range from eight to thirty-two characters as our final test set. For the test-ko, we selected one thousand speech samples with a WER of less than five percent and no deletion or insertion errors, utilizing the Whisper-Large V three as the ASR model. These samples were used as reference utterances for the Korean speech synthesis. For the input text, we randomly selected one thousand text samples from the remaining data. We have released the lists of prompt speeches, prompt transcriptions and input text from these two test sets are released to facilitate result reproduction. By providing this open-source data, we aim to establish a benchmark for evaluating Japanese and Korean TTS models. The Whisper-large V three is used as the ASR model for Japanese and Korean evaluations. Experimental Results Evaluations on Speech Tokenizer An ideal speech tokenizer is supposed to effectively utilizes the codebook, preserves information at a high fidelity, and demonstrates speaker independence. In this part we evaluate our supervised speech tokenizer from four aspects: one) Codebook utilization rate; two) ASR error rate within the entire encoder; three) Token visualization of different speakers; four) Speaker identification training. It turns out that the FSQ-based tokenizer fully utilizes the codebook and maintains more effective information from the aspect of ASR, indicating more semantic information maintained by FSQ. We further analyze the characteristics of FSQ through the t S N E visualization. As an upstream model for TTS tasks, the tokenizer should strive to minimize the entanglement of speaker identity information with the speech signal. We selected one hundred speech samples from each of the three speakers in the VoxCeleb1 dataset and visualized the corresponding tokens. It is evident that before the quantization, Encoder one's outputs exhibit different distributions among different speakers. In contrast, the distributions of quantized representations are nearly indistinguishable. In addition, the tokenizer fully utilizes the codebook. Subsequently, the S3prl toolkit is employed to further evaluate the speaker entanglement by performing speaker identification (SID) task. We use Sensevoice-large encoder with FSQ as an upstream feature extractor and train SID task with representations before or after the quantization. The convergence curves of SID training show that the SID layer with quantized tokens does not converge, which proves the decoupling function of the tokenizer on speaker information. Comparison Results with Baselines We first evaluated our CosyVoice two models on a limited English text domain and compared it with several open-source models, such as ChatTTS, GPT-SoVITs, OpenVoice, ParlerTTS, EmotiVoice, and its predecessor CosyVoice. The objective results include content consistency (WER), speech quality (NMOS) and speaker similarity (SS). The results show that CosyVoice two achieves state-of-the-art performance on the Librispeech test-clean set, surpassing all baseline models across all evaluation metrics. Notably, CosyVoice two even demonstrates higher content consistency, speech quality, and speaker similarity than human utterances, indicating its human-parity synthesis quality. We also evaluated CosyVoice two on the commonly-used test sets: SEED test-zh, test-en and test-hard, which include diverse input texts and reference speeches from various domains. On the test-zh set, CosyVoice two surpasses all open-sourced models in terms of CER and SS, falling short of the commercial model SEED-TTS by only a small margin. On the test-en set, CosyVoice two ranks fourth and third in terms of WER and SS, respectively. This may result from the imbalance in the volume of training data between Chinese and English. We plan to explore data scaling in future work to enhance content consistency in English. On the test-hard set, the offline CosyVoice two model achieves state-of-the-art performance across all compared baseline, demonstrating its robustness in challenging synthesis scenarios. Compared with human-generated speeches, CosyVoice two shows comparable content consistency and superior speaker similarity. Considering the recognition errors can also stem from the ASR model, it is reasonable to conclude that CosyVoice two achieves a human-parity synthesis capability. We also evaluated the streaming mode, denoted as ``CosyVoice two S''. For both evaluation settings, the streaming mode's performance is nearly lossless in typical test cases. Only in challenging cases is there a slight degradation in content consistency, highlighting the strength of our unified streaming/non-streaming framework. We found that the results of speaker similarity are not consistent on different SV models. This may indicate a new research topic on how to evaluate speaker similarity for TTS models automatically. Since different TTS models may use different SV models to extract speaker information, evaluating speaker similarity with the same SV model allows a more accurate evaluation on the utilization of speaker information. Therefore, we employ ERes two Net for evaluating speaker similarity in subsequent experiments. Modular Ablation Study We conducted a modular ablation study on the text-speech language model to assess the impacts of our modifications, including LLM initialization, removing speaker embedding, and utilizing FSQ.
long_en_367	poet_en	967	en	For this package it all started with wanting to make Jerome some banana nut bread, one of his favorites. The kids helped me decorate his box with paper and stickers I found at Hobby Lobby. It turned out to be a really cute package. We hope he likes it! Punch buggy is the kids' favorite game to play in the car. Many times I have to initiate a "don't punch so hard" rule because they can get a little carried away. But Jerome is the best at playing this game; he always manages to spot the bugs before anyone else. Before he left he showed me a house on our way home that has an old bug and told me to always remember that one so I can at least get one on the kids. Thanks, honey, but they have already figured that one out! The kids are all well and back in school. Spencer will have a lot of homework to catch up on; he has been absent since January 13. I just hope I will be able to help him with all of it—I was never very good at algebra. Jerome always helps him with his math homework. Katie was up early this morning; she was excited to put a stamp on the letter she wrote to her daddy. We were finally given an address to send things to him while he is in training. I wish you could have seen the smile on her face as she put her letter in the mailbox, raised the red flag, then ran to catch the school bus. The kids and I have been talking about sending Jerome his first care package. We have a surprise for that one, too. We are all excited about preparing what we'll put in his package. Jerome, when you read this, know that Molly misses you too. I laughed when you wanted me to put you on speaker so you could talk to her. I laughed when you told me what the other guys said, and it made me smile to see Molly perk her ears up and stare at the phone. I think she misses you too, but she keeps the kids busy and their minds occupied. And yes, they still feed her when they're supposed to. I love you, baby. I've hit a breaking point — I don't feel like I can be strong anymore, at least not tonight. I appreciate all the support and kind words we've received from everyone; it has meant so much. I'm just at a point where I can't pretend anymore. Tonight I'm not strong, but I'm trying really, really hard to be. The tears just won't stop. The next morning I woke up to a phone call from the school: Natalie was sick and needed to be picked up. We came home and she barely made it to the bathroom before throwing up; she had a stomach bug. I can't help replaying in my head what the "pastor lady" (as Spencer called her) said to me. She came to visit Spencer while we were waiting for the tests to be run and the results to come in. I know that when she entered our room she wasn't expecting to see a fourteen-year-old boy distraught and wishing his dad were there with him too. I think she was taken aback by it all, by the tears welling up in his eyes when she asked if she could pray with him. She prayed a very emotional prayer. She prayed for his health and that he would get better soon. She prayed for all the pain, loneliness, and longing to be together that filled the room. She prayed for Jerome's safety and for God's hand to protect him while he was away, and to protect us as well. When she finished talking to Spencer and was getting ready to leave, she hugged me and said that it wasn't His intention for us to go through all of this alone and that everything would be OK. I had kind of laughed off the emotion I was shoving down when she entered the room and told her it was a deployment curse. She knew I was holding it all in. She kept asking to call our church or our pastor and asked if he knew we were there. I didn't... because I had my brave face on. I was strong, or at least pretending to be. The other night Jerome texted me asking how Spencer and Natalie were feeling and if they were getting better. Then he asked how I was. I told him, "I think the kids being sick has just broken the strength I thought I had to get through this. They are feeling a little better now, so I'm feeling like I can take a big breath and just breathe out all the stress..." Tonight I can't. Maybe I've just hit a point where I can't hold these feelings inside anymore. Maybe it's because I'm exhausted, physically and emotionally. Maybe it's because I'm trying so very hard to be strong for everyone else. Maybe I just need to cry so I can move on and be strong again. Tonight Natalie climbed into bed with me and watched a little TV. When it was time for bed, I told her I needed her to sleep in her room. I was feeling all the emotions that had built up inside, and I just needed a good cry, but I didn't want her to know. I wanted to be strong and not let her know why I needed to sleep alone.	For this package it all started with wanting to make Jerome some banana nut bread, one of his favorites. The kids helped me decorate his box with paper and stickers I found at Hobby Lobby. It turned out to be a really cute package. We hope he likes it! Punch buggy is the kids' favorite game to play in the car. Many times I have to initiate a "don't punch so hard" rule because they can get a little carried away. But Jerome is the best at playing this game; he always manages to spot the bugs before anyone else. Before he left he showed me a house on our way home that has an old bug and told me to always remember that one so I can at least get one on the kids. Thanks, honey, but they have already figured that one out! The kids are all well and back in school. Spencer will have a lot of homework to catch up on; he has been absent since January thirteen. I just hope I will be able to help him with all of it—I was never very good at algebra. Jerome always helps him with his math homework. Katie was up early this morning; she was excited to put a stamp on the letter she wrote to her daddy. We were finally given an address to send things to him while he is in training. I wish you could have seen the smile on her face as she put her letter in the mailbox, raised the red flag, then ran to catch the school bus. The kids and I have been talking about sending Jerome his first care package. We have a surprise for that one, too. We are all excited about preparing what we'll put in his package. Jerome, when you read this, know that Molly misses you too. I laughed when you wanted me to put you on speaker so you could talk to her. I laughed when you told me what the other guys said, and it made me smile to see Molly perk her ears up and stare at the phone. I think she misses you too, but she keeps the kids busy and their minds occupied. And yes, they still feed her when they're supposed to. I love you, baby. I've hit a breaking point — I don't feel like I can be strong anymore, at least not tonight. I appreciate all the support and kind words we've received from everyone; it has meant so much. I'm just at a point where I can't pretend anymore. Tonight I'm not strong, but I'm trying really, really hard to be. The tears just won't stop. The next morning I woke up to a phone call from the school: Natalie was sick and needed to be picked up. We came home and she barely made it to the bathroom before throwing up; she had a stomach bug. I can't help replaying in my head what the "pastor lady" (as Spencer called her) said to me. She came to visit Spencer while we were waiting for the tests to be run and the results to come in. I know that when she entered our room she wasn't expecting to see a fourteen-year-old boy distraught and wishing his dad were there with him too. I think she was taken aback by it all, by the tears welling up in his eyes when she asked if she could pray with him. She prayed a very emotional prayer. She prayed for his health and that he would get better soon. She prayed for all the pain, loneliness, and longing to be together that filled the room. She prayed for Jerome's safety and for God's hand to protect him while he was away, and to protect us as well. When she finished talking to Spencer and was getting ready to leave, she hugged me and said that it wasn't His intention for us to go through all of this alone and that everything would be OK. I had kind of laughed off the emotion I was shoving down when she entered the room and told her it was a deployment curse. She knew I was holding it all in. She kept asking to call our church or our pastor and asked if he knew we were there. I didn't... because I had my brave face on. I was strong, or at least pretending to be. The other night Jerome texted me asking how Spencer and Natalie were feeling and if they were getting better. Then he asked how I was. I told him, "I think the kids being sick has just broken the strength I thought I had to get through this. They are feeling a little better now, so I'm feeling like I can take a big breath and just breathe out all the stress..." Tonight I can't. Maybe I've just hit a point where I can't hold these feelings inside anymore. Maybe it's because I'm exhausted, physically and emotionally. Maybe it's because I'm trying so very hard to be strong for everyone else. Maybe I just need to cry so I can move on and be strong again. Tonight Natalie climbed into bed with me and watched a little TV. When it was time for bed, I told her I needed her to sleep in her room. I was feeling all the emotions that had built up inside, and I just needed a good cry, but I didn't want her to know. I wanted to be strong and not let her know why I needed to sleep alone.
long_en_314	wiki_en	891	en	Kind Space (formerly Pink Triangle Services, or PTS) is an LGBT community centre in Ottawa, Ontario, Canada. It is the oldest registered LGBT-specific charity in Canada, having been registered in 1984. The organization serves people of all ages in the National Capital Region, including gay, lesbian, bisexual, transgender, Two-Spirit, non-binary, queer, questioning, intersex, and asexual individuals, as well as QTBIPoC (queer, trans, Black, Indigenous, and people of colour). It provides support groups, education, research, advocacy, and community space. Organization history: One of the founding members, Barry Deeprose, wrote a history of the organization in 2002. The foreword begins: "In 1984, a group of directors from the Board of Gays of Ottawa set out to create a non-profit organization with charitable status called Pink Triangle Services, in the hope that such recognition would enable the organization to more easily raise funds. In the spring of that year the corporation was founded and was granted charitable status under the Income Tax Act, a first for Canada. In the ensuing fifteen years Pink Triangle Services (PTS, as it quickly became known) flourished and grew, in many ways fulfilling the vision of the founding members. Those fifteen years also brought tremendous changes to the gay and lesbian community." AIDS decimated a generation of gay men, while at the same time in Canada gays and lesbians won legal rights that an earlier generation scarcely dreamed of. I have had the privilege of being associated with PTS from the beginning, and it remains close to my heart. There is no doubt in my mind that it continues to provide much-needed services to the gay, lesbian, and bisexual community. Until the 2000s, Kind Space was run by a working board of directors and funded primarily by individual donations from those communities. Starting in the 2000s, the organizational structure changed: the board shifted to focus on policy and governance, and the organization hired its first executive director. Another major change in the 2000s was a greater focus on underserved members of the community, including bisexuals, trans people, and those who identify solely as queer. This was also influenced by a decrease in stigma toward gays and lesbians in the broader community, which led to a court ruling making same-sex marriage legal in Ontario in 2003. A bisexual women's group started at Kind Space, which years later evolved into BiAmore, a group for bisexuals of all genders. In 2004, Trans Youth Ottawa started a group for transgender and transsexual youth and young adults that ran until 2009. Gender Quest, a group for all trans adults, started in 2005 and still operates today. In 2005 the board of directors added "trans," "two-spirit," and "queer" to the letters patent to state their official commitment to those communities. Rebranding and restructuring After years of financial troubles, which resulted in the temporary layoff of the organization's executive director, Kind Space began to restructure and repair its reputation. In 2015, Kind Space won a $100,000 rebranding package from Stiff and changed its name from Pink Triangle Services (PTS) to Kind Space. Following the rebranding, the organization partnered with Planned Parenthood Ottawa to share office space. In 2017, the organization celebrated being debt-free for the first time in several years. Programs and services Kind Space offers support, education, and resource programs. Support programs Historically, Kind Space's flagship programs have been peer support groups. Over the years Kind Space has seen many groups come and go based on need. Current support groups include: - QTY – Queer Trans Youth: a peer-led discussion and support group for LGBTTQIA youth, ages 25 and under. - Cafe Q: a low-pressure hangout space where participants enjoy board games, video games, and lounging on couches. - High School Student Alliance: focused on empowering youth to make systemic change within schools that is reflective of their needs as queer and trans youth. - The Men's Group: a peer-led social group for men of all ages, cultures, abilities, and orientations. QPoC-It: A space to celebrate the diversity of our cultures and to amplify the voices and concerns of QTBIPoC communities. Polybilities: An open forum for exploring relationship dynamics beyond standard societal norms. BiAmore: A comfortable space to gather and discuss issues related to being bisexual, pansexual, bi-curious, and questioning. Gender Quest: A diverse, peer-led support group for people at any stage of transition. Ace/Aro Spectrum: For anyone questioning their identity as asexual, aromantic, demisexual, demiromantic, greysexual, or elsewhere on the asexual/aromantic spectrum. Educational programs Historically, Kind Space has focused on education and advocacy work since its founding. Financial troubles previously restricted Kind Space's education program; however, the program has recently been revitalized. Skills for the Revolution is a grassroots, peer-led program that empowers queer and trans communities by building knowledge and providing hands-on experience in fundamental life, advocacy, and community-building skills to help deconstruct oppressive structures. This program centers healing, emotional, and restorative justice. Resource programs Kind Space offers several resources, including the Dr. Kelly McGinnis Library, which contains a large selection of queer- and trans-specific literature. Space Rentals Celebrating Self-Support in Accessing Other Community Resources Referrals Letters of Support – Refugee Hearings Manajiwin – Body-Positive Fitness References LGBT Community Centres LGBT Culture in Ottawa Organizations Based in Ottawa Charities Based in Canada	Kind Space (formerly Pink Triangle Services, or PTS) is an LGBT community centre in Ottawa, Ontario, Canada. It is the oldest registered LGBT-specific charity in Canada, having been registered in nineteen eighty four. The organization serves people of all ages in the National Capital Region, including gay, lesbian, bisexual, transgender, Two-Spirit, non-binary, queer, questioning, intersex, and asexual individuals, as well as QTBIPoC (queer, trans, Black, Indigenous, and people of colour). It provides support groups, education, research, advocacy, and community space. Organization history: One of the founding members, Barry Deeprose, wrote a history of the organization in two thousand two. The foreword begins: "In nineteen eighty four, a group of directors from the Board of Gays of Ottawa set out to create a non-profit organization with charitable status called Pink Triangle Services, in the hope that such recognition would enable the organization to more easily raise funds. In the spring of that year the corporation was founded and was granted charitable status under the Income Tax Act, a first for Canada. In the ensuing fifteen years Pink Triangle Services (PTS, as it quickly became known) flourished and grew, in many ways fulfilling the vision of the founding members. Those fifteen years also brought tremendous changes to the gay and lesbian community." AIDS decimated a generation of gay men, while at the same time in Canada gays and lesbians won legal rights that an earlier generation scarcely dreamed of. I have had the privilege of being associated with PTS from the beginning, and it remains close to my heart. There is no doubt in my mind that it continues to provide much-needed services to the gay, lesbian, and bisexual community. Until the two thousands, Kind Space was run by a working board of directors and funded primarily by individual donations from those communities. Starting in the two thousands, the organizational structure changed: the board shifted to focus on policy and governance, and the organization hired its first executive director. Another major change in the two thousands was a greater focus on underserved members of the community, including bisexuals, trans people, and those who identify solely as queer. This was also influenced by a decrease in stigma toward gays and lesbians in the broader community, which led to a court ruling making same-sex marriage legal in Ontario in two thousand three. A bisexual women's group started at Kind Space, which years later evolved into BiAmore, a group for bisexuals of all genders. In two thousand four, Trans Youth Ottawa started a group for transgender and transsexual youth and young adults that ran until two thousand nine. Gender Quest, a group for all trans adults, started in two thousand five and still operates today. In two thousand five the board of directors added "trans," "two-spirit," and "queer" to the letters patent to state their official commitment to those communities. Rebranding and restructuring After years of financial troubles, which resulted in the temporary layoff of the organization's executive director, Kind Space began to restructure and repair its reputation. In two thousand fifteen, Kind Space won a one hundred thousand dollars rebranding package from Stiff and changed its name from Pink Triangle Services (PTS) to Kind Space. Following the rebranding, the organization partnered with Planned Parenthood Ottawa to share office space. In two thousand seventeen, the organization celebrated being debt-free for the first time in several years. Programs and services Kind Space offers support, education, and resource programs. Support programs Historically, Kind Space's flagship programs have been peer support groups. Over the years Kind Space has seen many groups come and go based on need. Current support groups include: - QTY – Queer Trans Youth: a peer-led discussion and support group for LGBTTQIA youth, ages twenty five and under. - Cafe Q: a low-pressure hangout space where participants enjoy board games, video games, and lounging on couches. - High School Student Alliance: focused on empowering youth to make systemic change within schools that is reflective of their needs as queer and trans youth. - The Men's Group: a peer-led social group for men of all ages, cultures, abilities, and orientations. QPoC-It: A space to celebrate the diversity of our cultures and to amplify the voices and concerns of QTBIPoC communities. Polybilities: An open forum for exploring relationship dynamics beyond standard societal norms. BiAmore: A comfortable space to gather and discuss issues related to being bisexual, pansexual, bi-curious, and questioning. Gender Quest: A diverse, peer-led support group for people at any stage of transition. Ace slash Aro Spectrum: For anyone questioning their identity as asexual, aromantic, demisexual, demiromantic, greysexual, or elsewhere on the asexual slash aromantic spectrum. Educational programs Historically, Kind Space has focused on education and advocacy work since its founding. Financial troubles previously restricted Kind Space's education program; however, the program has recently been revitalized. Skills for the Revolution is a grassroots, peer-led program that empowers queer and trans communities by building knowledge and providing hands-on experience in fundamental life, advocacy, and community-building skills to help deconstruct oppressive structures. This program centers healing, emotional, and restorative justice. Resource programs Kind Space offers several resources, including the Dr. Kelly McGinnis Library, which contains a large selection of queer- and trans-specific literature. Space Rentals Celebrating Self-Support in Accessing Other Community Resources Referrals Letters of Support – Refugee Hearings Manajiwin – Body-Positive Fitness References LGBT Community Centres LGBT Culture in Ottawa Organizations Based in Ottawa Charities Based in Canada.
long_en_345	poet_en	1,042	en	For several years before meeting Hubby, I would spend New Year's Eve at my girlfriend's house with her and her boys. She has five boys, and when they were younger, if I showed up and we made a lot of food, it was a party. So on 12/31/95 I went over in the early evening and we started to party. That entailed drinking some wine, making food, and watching movies. A little after midnight, after we had counted down to the New Year, my girlfriend said, "Let's go to the Hop." The Hop was a disco — yes, disco — that she liked to frequent. Since I was a raging workaholic, I did not go out at all. She had decided that my New Year's resolution should be to get a life, so she convinced me to head to the Hop. I was wearing flat shoes, a sloppy sweater, and my hair was flat — not my typical going-out attire. But we went. (Her oldest son was old enough to stay at home with the other boys, and my girlfriend had a pager; we could be back in 10 minutes if something happened.) When we walked in the door, I looked across the bar and saw a tall, good-looking guy in a white shirt that made his broad shoulders look even broader. Hmmm, I thought, but figured I was not dressed to pick up any hotties. My friend said, "Oh, look. There's Bob!" and headed across the room. This girl knew everyone in that place! Yes, she was a partier, but only when her kids were with their dad. She walked right up to the broad-shouldered hottie and said hi. Oh my God — that was Bob, and she knew him! We ended up talking and dancing until the bar closed. I figured that was it until my girlfriend invited him to follow us to her house to continue our New Year celebration. He agreed and followed us. In the car she told me all about how she had seen him there for the last couple of years and what a nice guy he was. He was divorced with three kids and loved them like crazy. Well, I knew he was not for me in that case. I was not interested in kids and didn't want to date anyone with kids. I actually didn't want to date anyone. After my divorce several years earlier, I had decided I was better off on my own. We got to my girlfriend's house and ended up drinking wine and talking into the early morning. We had sobered up by the time we decided to leave, so he said goodbye, got in his car, and left. I got in my car and left. That might have been the end of it, but my girlfriend decided that he and I should get together, so she started playing matchmaker. It took her a few weeks, but she finally got us out together. I thank her regularly for her perseverance. I'm going to catch up on the happenings around here. On Friday, December 19, I had scheduled a day off to go with my husband to pick up his new leg. The Thursday before, I woke up and felt funny. I don't even know how to describe it. My tummy was mildly upset, but I figured it would go away, so I got ready and went to work. When I got to work, I felt worse. I didn't know if I had a stomach bug or if something I ate had disagreed with me, but I was just feeling worse and worse. I told my boss I might need to leave, and he told me to go if I needed to. Hubby called me around 10, and when I told him how I was feeling, he offered to come get me. I have never asked Hubby to come get me from work. I said, "Yes, please." So he came to pick me up. When I got in the car, he handed me a couple of plastic bags. "In case you need to throw up," he said. Isn't he sweet? I got home and crawled into bed because I felt like I needed to throw up except when I was lying down. So I spent the next 18 hours in bed or in the bathroom. Hubby has never seen me sleep that much. He said he came into the bedroom a few times to make sure I was still alive. LOL! He cannot walk on it for long amounts of time, but he can wear it and stand on it. He starts PT to learn how to walk with it today. But he wore it on Saturday, 12/20, when we had our get-together at his mom's. He walked into her house on his own two feet. He spent most of the day sitting and used his walker to get around when he did. He probably only spent about 30 minutes total walking on it, so not too much. The get-together was nice. Our kids and two grandkids were there. Jacob, our middle grandson, was not there. Chance gave some mumbled excuse. I think maybe he just didn't feel like going to pick Jacob up. Oh well, we just sent his gifts home with Chance. I took some pics but no group shots. Got some of those on MIL's camera, so hopefully she will send us some. Here is a typical pic of Hubby and our kids. Christmas Day was nice and relaxing. We had way too many gifts to open, but they were all fun. Hubby thinks I need lots of presents to open. Don't tell him, but I love it. Oh, don't worry — I am cooking Thanksgiving dinner for Hubby and me. I will not deny us my awesome cooking. LOL! But we are not having anyone over for Thanksgiving. Hubby and I both figured that our kids would do as they had done in previous years and attend the Thanksgiving get-together that Hubby's sister hosts — the sister Hubby has not spoken to in four years.	For several years before meeting Hubby, I would spend New Year's Eve at my girlfriend's house with her and her boys. She has five boys, and when they were younger, if I showed up and we made a lot of food, it was a party. So on twelve thirty one ninety five I went over in the early evening and we started to party. That entailed drinking some wine, making food, and watching movies. A little after midnight, after we had counted down to the New Year, my girlfriend said, "Let's go to the Hop." The Hop was a disco — yes, disco — that she liked to frequent. Since I was a raging workaholic, I did not go out at all. She had decided that my New Year's resolution should be to get a life, so she convinced me to head to the Hop. I was wearing flat shoes, a sloppy sweater, and my hair was flat — not my typical going-out attire. But we went. (Her oldest son was old enough to stay at home with the other boys, and my girlfriend had a pager; we could be back in ten minutes if something happened.) When we walked in the door, I looked across the bar and saw a tall, good-looking guy in a white shirt that made his broad shoulders look even broader. Hmmm, I thought, but figured I was not dressed to pick up any hotties. My friend said, "Oh, look. There's Bob!" and headed across the room. This girl knew everyone in that place! Yes, she was a partier, but only when her kids were with their dad. She walked right up to the broad-shouldered hottie and said hi. Oh my God — that was Bob, and she knew him! We ended up talking and dancing until the bar closed. I figured that was it until my girlfriend invited him to follow us to her house to continue our New Year celebration. He agreed and followed us. In the car she told me all about how she had seen him there for the last couple of years and what a nice guy he was. He was divorced with three kids and loved them like crazy. Well, I knew he was not for me in that case. I was not interested in kids and didn't want to date anyone with kids. I actually didn't want to date anyone. After my divorce several years earlier, I had decided I was better off on my own. We got to my girlfriend's house and ended up drinking wine and talking into the early morning. We had sobered up by the time we decided to leave, so he said goodbye, got in his car, and left. I got in my car and left. That might have been the end of it, but my girlfriend decided that he and I should get together, so she started playing matchmaker. It took her a few weeks, but she finally got us out together. I thank her regularly for her perseverance. I'm going to catch up on the happenings around here. On Friday, December nineteen, I had scheduled a day off to go with my husband to pick up his new leg. The Thursday before, I woke up and felt funny. I don't even know how to describe it. My tummy was mildly upset, but I figured it would go away, so I got ready and went to work. When I got to work, I felt worse. I didn't know if I had a stomach bug or if something I ate had disagreed with me, but I was just feeling worse and worse. I told my boss I might need to leave, and he told me to go if I needed to. Hubby called me around ten, and when I told him how I was feeling, he offered to come get me. I have never asked Hubby to come get me from work. I said, "Yes, please." So he came to pick me up. When I got in the car, he handed me a couple of plastic bags. "In case you need to throw up," he said. Isn't he sweet? I got home and crawled into bed because I felt like I needed to throw up except when I was lying down. So I spent the next eighteen hours in bed or in the bathroom. Hubby has never seen me sleep that much. He said he came into the bedroom a few times to make sure I was still alive. LOL! He cannot walk on it for long amounts of time, but he can wear it and stand on it. He starts PT to learn how to walk with it today. But he wore it on Saturday, twelve slash twenty, when we had our get-together at his mom's. He walked into her house on his own two feet. He spent most of the day sitting and used his walker to get around when he did. He probably only spent about thirty minutes total walking on it, so not too much. The get-together was nice. Our kids and two grandkids were there. Jacob, our middle grandson, was not there. Chance gave some mumbled excuse. I think maybe he just didn't feel like going to pick Jacob up. Oh well, we just sent his gifts home with Chance. I took some pics but no group shots. Got some of those on MIL's camera, so hopefully she will send us some. Here is a typical pic of Hubby and our kids. Christmas Day was nice and relaxing. We had way too many gifts to open, but they were all fun. Hubby thinks I need lots of presents to open. Don't tell him, but I love it. Oh, don't worry — I am cooking Thanksgiving dinner for Hubby and me. I will not deny us my awesome cooking. LOL! But we are not having anyone over for Thanksgiving. Hubby and I both figured that our kids would do as they had done in previous years and attend the Thanksgiving get-together that Hubby's sister hosts — the sister Hubby has not spoken to in four years.
long_en_359	poet_en	840	en	In the following years a few events took place. The Queen was poisoned after she gave birth, and, as the King wished, the child was sent to be raised by Thomas’s family. The grandparents took the child in, but they could not love the child of the man who had killed their son, so the child grew up without knowing love. The people discovered the Queen’s crime and demanded she pay with her life for Thomas’s death and for deceiving the royal family; the rumors had originated with the Loyal Royal Knights. It was a tragic end for a woman who had reigned for over a decade: her father had died, and she had lost the will to live. She kissed her newborn daughter before the child was taken away, never to be seen again, and then the Queen died. Governor Valtac received a modest burial. As the King’s father-in-law, his funeral raised questions, but the King could not give him a grander ceremony. Valtac had turned his daughter into a schemer, and after his death the King discovered he had misgoverned Harval. Once that became known, people stopped praising Valtac and accepted the King’s decision. The Chest of the Gods was guarded for several years. Then the knights were able to return home and retire. A single guard remained to watch it, but he was allowed to rest and pursue hobbies in the meantime. The Crown Prince learned from both Mahnu and his father; he was obstinate at first because of his mother's death. After a month, King Richard could no longer tolerate his son's behavior and spoke only of the late queen's deeds so his son would understand what had happened. By telling him everything, the king hoped the prince would grow up faster and learn from his mother's mistakes. King Richard also explained everything about the gods so his son could understand King Maximus and the nature of sacrifice. The Crown Prince stood by his father's side for all royal duties and was even asked for his opinions on some matters now that he was older. Mahnu refused the position of queen, leaving it vacant until the Crown Prince became king; the current king had decided not to appoint anyone. He already wanted to abdicate, and if Mahnu would not be queen, he did not want anyone else to hold the title. It was also his final mark of respect for the late queen. William did not become leader of the Loyal Royal Knights; instead, he continued to follow Mahnu, guarding her and documenting her past. Instead, the position of Leader of the Loyal Royal Knights was decided as it had been centuries before: by combat. The fight became a major event, drawing many people to Conrella to celebrate the new leader and enjoy the festivities. Conrella had never been so crowded and noisy. In the end a knight was chosen and given the title; he was surprised when he learned about the order's past. For a full day he could not believe what he had heard from King Richard, from William, and even from Mahnu. He had not expected the role to be like this, but as days passed he settled into a routine, obeying the King's orders and faithfully supporting the Crowned Prince. Mahnu came to understand more about finding happiness in a family. As time went on she felt life growing within her and realized what was coming: someone would depend on her completely, and she needed to be ready. Her belly grew, and soon she experienced childbirth. Like many mothers, Mahnu was in pain and told her beloved that she never wanted to give birth again. Later, as her family grew, she described it as very fulfilling and learned patience and gentleness. She had to learn how to teach and make sure she showed her love. It was also very tiring. Sleep came easily, something she had always wanted when she was a god, and now she couldn't get enough of it. Nonetheless, promises were kept. Their first daughter was named Cassandra, and their first son was named Marcus. After a few years Richard abdicated the throne, passing it to his eldest son; he couldn't wait to spend the rest of his life with his beloved Mahnu. Through long, hard days their relationship grew, and neither of them ever felt lonely—right up until the final day when Mahnu passed away. Her life had been full. She loved her family very much, did not regret her existence, and had thought for a while that instead of her life being a cruel fate, it was something she accepted; perhaps fate had led her to seek out what she had not experienced in over a thousand years. In her opinion, her greatest victory was not over the gods nor in accomplishing her goals, but in filling her life with happiness and experience. It had been more than worth it.	In the following years a few events took place. The Queen was poisoned after she gave birth, and, as the King wished, the child was sent to be raised by Thomas’s family. The grandparents took the child in, but they could not love the child of the man who had killed their son, so the child grew up without knowing love. The people discovered the Queen’s crime and demanded she pay with her life for Thomas’s death and for deceiving the royal family; the rumors had originated with the Loyal Royal Knights. It was a tragic end for a woman who had reigned for over a decade: her father had died, and she had lost the will to live. She kissed her newborn daughter before the child was taken away, never to be seen again, and then the Queen died. Governor Valtac received a modest burial. As the King’s father-in-law, his funeral raised questions, but the King could not give him a grander ceremony. Valtac had turned his daughter into a schemer, and after his death the King discovered he had misgoverned Harval. Once that became known, people stopped praising Valtac and accepted the King’s decision. The Chest of the Gods was guarded for several years. Then the knights were able to return home and retire. A single guard remained to watch it, but he was allowed to rest and pursue hobbies in the meantime. The Crown Prince learned from both Mahnu and his father; he was obstinate at first because of his mother's death. After a month, King Richard could no longer tolerate his son's behavior and spoke only of the late queen's deeds so his son would understand what had happened. By telling him everything, the king hoped the prince would grow up faster and learn from his mother's mistakes. King Richard also explained everything about the gods so his son could understand King Maximus and the nature of sacrifice. The Crown Prince stood by his father's side for all royal duties and was even asked for his opinions on some matters now that he was older. Mahnu refused the position of queen, leaving it vacant until the Crown Prince became king; the current king had decided not to appoint anyone. He already wanted to abdicate, and if Mahnu would not be queen, he did not want anyone else to hold the title. It was also his final mark of respect for the late queen. William did not become leader of the Loyal Royal Knights; instead, he continued to follow Mahnu, guarding her and documenting her past. Instead, the position of Leader of the Loyal Royal Knights was decided as it had been centuries before: by combat. The fight became a major event, drawing many people to Conrella to celebrate the new leader and enjoy the festivities. Conrella had never been so crowded and noisy. In the end a knight was chosen and given the title; he was surprised when he learned about the order's past. For a full day he could not believe what he had heard from King Richard, from William, and even from Mahnu. He had not expected the role to be like this, but as days passed he settled into a routine, obeying the King's orders and faithfully supporting the Crowned Prince. Mahnu came to understand more about finding happiness in a family. As time went on she felt life growing within her and realized what was coming: someone would depend on her completely, and she needed to be ready. Her belly grew, and soon she experienced childbirth. Like many mothers, Mahnu was in pain and told her beloved that she never wanted to give birth again. Later, as her family grew, she described it as very fulfilling and learned patience and gentleness. She had to learn how to teach and make sure she showed her love. It was also very tiring. Sleep came easily, something she had always wanted when she was a god, and now she couldn't get enough of it. Nonetheless, promises were kept. Their first daughter was named Cassandra, and their first son was named Marcus. After a few years Richard abdicated the throne, passing it to his eldest son; he couldn't wait to spend the rest of his life with his beloved Mahnu. Through long, hard days their relationship grew, and neither of them ever felt lonely—right up until the final day when Mahnu passed away. Her life had been full. She loved her family very much, did not regret her existence, and had thought for a while that instead of her life being a cruel fate, it was something she accepted; perhaps fate had led her to seek out what she had not experienced in over a thousand years. In her opinion, her greatest victory was not over the gods nor in accomplishing her goals, but in filling her life with happiness and experience. It had been more than worth it.
long_en_361	poet_en	974	en	Find Your Mate - Part 1: An Emergency Story By Donnalouise A friend of mine called me up one night and asked me to come over for dinner. I said, "Sure, I'll come." I drove over to Mike and Amy's and went to the front door and knocked. Mike answered and told me Amy was in the kitchen on the phone and to come on back. We both went to the kitchen and Mike offered me a Coke while Amy finished her conversation. We sat at the breakfast bar and Mike and I talked, catching up on the news since we'd last seen each other. We don't get to see each other as much as we'd like; with Mike's job as a chef at an Italian restaurant and Amy as an intensive care nurse at the hospital, we hadn't seen each other since the last holiday. It was nice to catch up about the kids and what they were doing. Mike and Amy have two children, a boy and a girl. Amy was still on the phone when there was a knock at the door. Mike went to the door just as Amy finished her conversation. "It was work, as usual," she said—there was a problem with a patient Amy had been taking care of earlier that day. Amy fixed herself a glass of iced tea and joined me at the breakfast bar. I asked Amy who Mike and she wanted me to meet. I told her Mike had said there was this really cute guy—then we heard the knock on the door. I reminded Amy that she should remember I really do not like blind dates. She laughed and told me, "You'll like this one!" Amy said she had met a nice guy at the hospital who was a paramedic for the fire department. She thought I could get information from him that might help with my studies and that he was someone I would like; if it went any further, that would be up to the two of us. Mike came back to the kitchen with a really cute guy and introduced us. "Donna, please meet John Gage." "John, this is Donna. We have known you both for some time now and thought it was about time you met each other," Mike added. John and Mike joined us, and then we went dancing at a nice club not far from my place. We had a couple of drinks and danced until ten o'clock. We left the club and John walked me to my car. I followed him to his apartment and we watched TV and talked more. It was so relaxing; we discovered that we had a lot more in common than we had first thought. I was starting to get sleepy and wondered if I would be able to drive home. John asked if I could drive; I told him I thought so. He asked one more time, "Are you sure? If not, you could stay here." I looked at him and asked, "Where would I sleep?" I told John I wasn't ready for the next step, and he replied, "You could sleep in the guest room. I have a really nice guest room." I told him, "No — I'll be OK. I really need to get home. I'll call you when I get home." He told me goodnight, kissed me, and said he would give me about 30 minutes to get to my apartment. If I had not called him in 30 minutes, he would drive to make sure I got there okay. I told John I would be OK and that he needed to get some sleep too, because I knew he was tired and had to get up early. He was probably more tired than I was because he'd had a big fire run just before his shift ended. I said I would call him as soon as I got home. I gave him a kiss, got into my car, and headed home. I drove with my window down and my radio up, hoping I wouldn't fall asleep. It only took me 15 minutes to drive home. I called John as soon as I got into my apartment. I told him I had a really good time, thanked him, and said we could talk tomorrow. I told him he should get to bed. We said goodnight. I told John that I had really enjoyed the date and couldn't remember when I'd had so much fun. I also told him I was glad I had the next day off so I could sleep. John and I agreed he would call after he got off work and we'd plan when to get together and have dinner at my place. I told John we could ride down with his friends and pull the boat. That was fine with me; he should just let me know what time we were starting so I could be ready and tell Amy, so they'd be ready when we pulled up at their place. We had planned to follow one another down to the lake so that if someone had any car trouble we could stop and help. We talked a little longer and made arrangements to meet tomorrow to go over everything and make sure we hadn't left out any details. I asked John to come over for dinner and said that we would have a quiet evening at my apartment. "We don't have to go out to eat or go dancing — we can just stay here, eat, and watch TV." John came over after he got off work, about seven o'clock. We decided to order pizza and watch an old movie on TV.	Find Your Mate - Part one: An Emergency Story By Donnalouise A friend of mine called me up one night and asked me to come over for dinner. I said, "Sure, I'll come." I drove over to Mike and Amy's and went to the front door and knocked. Mike answered and told me Amy was in the kitchen on the phone and to come on back. We both went to the kitchen and Mike offered me a Coke while Amy finished her conversation. We sat at the breakfast bar and Mike and I talked, catching up on the news since we'd last seen each other. We don't get to see each other as much as we'd like; with Mike's job as a chef at an Italian restaurant and Amy as an intensive care nurse at the hospital, we hadn't seen each other since the last holiday. It was nice to catch up about the kids and what they were doing. Mike and Amy have two children, a boy and a girl. Amy was still on the phone when there was a knock at the door. Mike went to the door just as Amy finished her conversation. "It was work, as usual," she said—there was a problem with a patient Amy had been taking care of earlier that day. Amy fixed herself a glass of iced tea and joined me at the breakfast bar. I asked Amy who Mike and she wanted me to meet. I told her Mike had said there was this really cute guy—then we heard the knock on the door. I reminded Amy that she should remember I really do not like blind dates. She laughed and told me, "You'll like this one!" Amy said she had met a nice guy at the hospital who was a paramedic for the fire department. She thought I could get information from him that might help with my studies and that he was someone I would like; if it went any further, that would be up to the two of us. Mike came back to the kitchen with a really cute guy and introduced us. "Donna, please meet John Gage." "John, this is Donna. We have known you both for some time now and thought it was about time you met each other," Mike added. John and Mike joined us, and then we went dancing at a nice club not far from my place. We had a couple of drinks and danced until ten o'clock. We left the club and John walked me to my car. I followed him to his apartment and we watched TV and talked more. It was so relaxing; we discovered that we had a lot more in common than we had first thought. I was starting to get sleepy and wondered if I would be able to drive home. John asked if I could drive; I told him I thought so. He asked one more time, "Are you sure? If not, you could stay here." I looked at him and asked, "Where would I sleep?" I told John I wasn't ready for the next step, and he replied, "You could sleep in the guest room. I have a really nice guest room." I told him, "No — I'll be OK. I really need to get home. I'll call you when I get home." He told me goodnight, kissed me, and said he would give me about thirty minutes to get to my apartment. If I had not called him in thirty minutes, he would drive to make sure I got there okay. I told John I would be okay and that he needed to get some sleep too, because I knew he was tired and had to get up early. He was probably more tired than I was because he'd had a big fire run just before his shift ended. I said I would call him as soon as I got home. I gave him a kiss, got into my car, and headed home. I drove with my window down and my radio up, hoping I wouldn't fall asleep. It only took me fifteen minutes to drive home. I called John as soon as I got into my apartment. I told him I had a really good time, thanked him, and said we could talk tomorrow. I told him he should get to bed. We said goodnight. I told John that I had really enjoyed the date and couldn't remember when I'd had so much fun. I also told him I was glad I had the next day off so I could sleep. John and I agreed he would call after he got off work and we'd plan when to get together and have dinner at my place. I told John we could ride down with his friends and pull the boat. That was fine with me; he should just let me know what time we were starting so I could be ready and tell Amy, so they'd be ready when we pulled up at their place. We had planned to follow one another down to the lake so that if someone had any car trouble we could stop and help. We talked a little longer and made arrangements to meet tomorrow to go over everything and make sure we hadn't left out any details. I asked John to come over for dinner and said that we would have a quiet evening at my apartment. "We don't have to go out to eat or go dancing — we can just stay here, eat, and watch TV." John came over after he got off work, about seven o'clock. We decided to order pizza and watch an old movie on TV.
long_en_366	poet_en	878	en	When the father of Adna’s friend was discharged from the hospital, they had no money to pay. The friend’s older sister was a nun, so Adna asked the head of the hospital business office whether a letter of authority could be accepted for a nun’s father. The head said such letters covered only priests who had been confined. Adna proposed that the bill be sent to the office that handles priests’ accounts for collection one month after discharge, the standard period for settling hospital expenses. The head, who already knew Adna, accepted the proposal. This would allow the nun’s father to leave the hospital without incurring daily costs. Adna then went to the office in charge of letters of authority. She belonged to the same congregation as the patient’s daughter. At first the officer refused because the patient was not a priest but the father of a nun. Adna replied, “I am not a nun, but I care deeply for this man, who is so loved by his daughter, your fellow nun. He is not my obligation and not my family, yet I have left my family and given my time and effort to help them. I know you can do more to help than I can.” After Adna's explanation, the head nun approved that the letter of authority would come from the priests' account, but they would not pay the hospital bill. The assistance was only to give them time—exactly one month after discharge—to find the money to settle their hospital account. Adna thanked the head nun and returned to the hospital to tell her friend's family they could check out without paying immediately, but must not exceed the agreed time to pay. When the bill was processed, Adna was so happy that the neurosurgeon, who had the highest fee, did not charge a single cent. The family was very grateful. Adna's friend cared for her father at home while her older sister, a nun, returned to the congregation house. After a month, the father claimed his early retirement payment, which covered the hospital debt. They then went to the wife's home province, and with fresh air, fruits, and vegetables, the father recovered. Once, he asked his daughter to invite Adna to their house in the province. Considering the distance, which was farther than Uncle Rpm’s ancestral house, Adna said, "Thank you for inviting me, but tell your father to stay healthy; there is no need to thank me." Ten years later Adna’s friend called to say that her father had passed away. Adna told her, "Your father was almost gone, but God, who saw your love for him, extended his life so you could spend more quality time together." She said the same thing had happened when she almost lost her Aunt Felip: she had prayed to God, and He had extended Aunt Felip’s life for more than two years after that Christmas. Adna said God had given the friend and her father more time; she had wished the same for Aunt Felip so she could keep her near. The friend understood that everything Adna had said was true. She was in her forties and had remained single all her life. Last year she met Al and Adna at the market with her companion. She told Al, "We are getting married this coming May, and I want you and Adna to be our principal sponsors." Adna asked, "Is the wedding at the nearby church?" She answered, "In the province." Adna said, "It's too far; we cannot go. Just move your wedding here, and I will cook a lot of food." Adna also did not like wearing gowns or high heels. She disliked heavy makeup and preferred her hair simply styled. Adna said, "If your wedding is here, I can help prepare the food." The day before the wedding, Adna asked her children, Lex, Azia, and Al the Second, "Would you like to travel to the province?" They were all excited and replied, "Yes." Adna told them, "Go and tell your father, Al, that we will attend the wedding tomorrow, but we have to leave after lunch and stay overnight." Al asked Adna, "Are we going to be the principal sponsors?" Adna replied, "No, I'm just going to wear ordinary slippers." After lunch, Lex, Azia, and Al the Second packed their backpacks and a small travel bag. Adna reminded them not to forget their toothbrushes and toothpaste. Then they went to the bus terminal because it would take them more than four hours to get there. Adna asked Lex to contact her friend for directions. Her friend was happy that the whole family would attend the wedding. On the way, no buses were available, so they had to pay extra to hire a van. They arrived late in the afternoon. Another sister of Adna's friend took charge of looking after the family. There was a small, beautiful house that was large enough to accommodate them for the night. The wedding was scheduled for ten o'clock the next morning, and they all went to the local church. After the ceremony, they returned to the bride's house for the reception.	When the father of Adna’s friend was discharged from the hospital, they had no money to pay. The friend’s older sister was a nun, so Adna asked the head of the hospital business office whether a letter of authority could be accepted for a nun’s father. The head said such letters covered only priests who had been confined. Adna proposed that the bill be sent to the office that handles priests’ accounts for collection one month after discharge, the standard period for settling hospital expenses. The head, who already knew Adna, accepted the proposal. This would allow the nun’s father to leave the hospital without incurring daily costs. Adna then went to the office in charge of letters of authority. She belonged to the same congregation as the patient’s daughter. At first the officer refused because the patient was not a priest but the father of a nun. Adna replied, “I am not a nun, but I care deeply for this man, who is so loved by his daughter, your fellow nun. He is not my obligation and not my family, yet I have left my family and given my time and effort to help them. I know you can do more to help than I can.” After Adna's explanation, the head nun approved that the letter of authority would come from the priests' account, but they would not pay the hospital bill. The assistance was only to give them time—exactly one month after discharge—to find the money to settle their hospital account. Adna thanked the head nun and returned to the hospital to tell her friend's family they could check out without paying immediately, but must not exceed the agreed time to pay. When the bill was processed, Adna was so happy that the neurosurgeon, who had the highest fee, did not charge a single cent. The family was very grateful. Adna's friend cared for her father at home while her older sister, a nun, returned to the congregation house. After a month, the father claimed his early retirement payment, which covered the hospital debt. They then went to the wife's home province, and with fresh air, fruits, and vegetables, the father recovered. Once, he asked his daughter to invite Adna to their house in the province. Considering the distance, which was farther than Uncle Rpm’s ancestral house, Adna said, "Thank you for inviting me, but tell your father to stay healthy; there is no need to thank me." Ten years later Adna’s friend called to say that her father had passed away. Adna told her, "Your father was almost gone, but God, who saw your love for him, extended his life so you could spend more quality time together." She said the same thing had happened when she almost lost her Aunt Felip: she had prayed to God, and He had extended Aunt Felip’s life for more than two years after that Christmas. Adna said God had given the friend and her father more time; she had wished the same for Aunt Felip so she could keep her near. The friend understood that everything Adna had said was true. She was in her forties and had remained single all her life. Last year she met Al and Adna at the market with her companion. She told Al, "We are getting married this coming May, and I want you and Adna to be our principal sponsors." Adna asked, "Is the wedding at the nearby church?" She answered, "In the province." Adna said, "It's too far; we cannot go. Just move your wedding here, and I will cook a lot of food." Adna also did not like wearing gowns or high heels. She disliked heavy makeup and preferred her hair simply styled. Adna said, "If your wedding is here, I can help prepare the food." The day before the wedding, Adna asked her children, Lex, Azia, and Al the Second, "Would you like to travel to the province?" They were all excited and replied, "Yes." Adna told them, "Go and tell your father, Al, that we will attend the wedding tomorrow, but we have to leave after lunch and stay overnight." Al asked Adna, "Are we going to be the principal sponsors?" Adna replied, "No, I'm just going to wear ordinary slippers." After lunch, Lex, Azia, and Al the Second packed their backpacks and a small travel bag. Adna reminded them not to forget their toothbrushes and toothpaste. Then they went to the bus terminal because it would take them more than four hours to get there. Adna asked Lex to contact her friend for directions. Her friend was happy that the whole family would attend the wedding. On the way, no buses were available, so they had to pay extra to hire a van. They arrived late in the afternoon. Another sister of Adna's friend took charge of looking after the family. There was a small, beautiful house that was large enough to accommodate them for the night. The wedding was scheduled for ten o'clock the next morning, and they all went to the local church. After the ceremony, they returned to the bride's house for the reception.
long_en_340	poet_en	1,067	en	I open my eyes in an unfamiliar room. My head hurts, and I try to remember how I ended up here, but I can only remember up to when the teleport was almost ready. I hear a sound and look that way. A woman is sitting at a desk. She looks young and wears glasses. After a few seconds she notices my gaze. "Oh, you've finally woken up. I was getting bored." "Where am I? How did I end up here? Is Seelena fine? Is Muradel okay? What happened with the monsters?" "A lot of questions... but all right. You are in Afterlife Management Service room #762. You are here because you died. Seelena is dead. Muradel is dead. The monsters destroyed the world." "Huh? Are you joking with me? I won't forgive you for saying my cute girls are dead, you know!" "I am simply stating the facts, sir." "That's ridiculous. I'm the hero of the prophecies. It's impossible for me to die." "Yes, your report definitely said that. I regret to inform you that, according to our logs, we've concluded there was a bug in the code and you died." "Code? What code? How did I die?" "Your universe's code. An unusual combination of factors occurred that weren't considered in the test cases. I'm sorry — on behalf of the company." "I don't understand." "I asked what killed me." "As I said, there were multiple factors involved… but ultimately you could say that it was a mosquito bite." "Me, the great hero, dying from a mosquito bite… I still can't believe it." "Well, there is nothing I can do about that, Sir Hero. I have only been assigned to your case to explain the situation." "Bring me back to my world. I need to save the Kingdom of Feldem!" "Unfortunately, that won't be possible. Your world has already been completely destroyed. No one could stop a young boy and his minions." "I haven't grasped it all, but isn't this your fault? Do something about it!" "Well, we do have a program to revive people, but it is too late. No one is alive in that world anymore. In the past we had a program that could reverse time; however, it has been incompatible since the company decided to switch operating systems." "I don't understand—are you saying that everyone is dead?" "That's right." Tears started falling from my face. It was unseemly, but I couldn't stop. Everything I knew, everyone I knew—gone. I had failed them. The only thing left was my regret. I should have tried to kill them all without my full power. Then at least everyone else would still be alive. I dwelled in my memories: my father training me to be a knight like him, the first monster I had slain, my father's death caused by a high-level demon king, my anger and revenge, unleashing a power I didn't even know I had, killing them all. The great oracle announced that I was the hero of the prophecies. I left my mother alone with my brother to travel the world, trying to save people from evil—saving people, becoming popular, saving Seelena, traveling together, our first kiss. Then we met Muradel; Seelena and Muradel fought to decide who was worthy of my lips, which ended in a draw. We met other girls, formed a bigger party, and I enjoyed my life. But all that is over. The woman with glasses simply watches me in silence. After a while my tears stop, and I try to understand the situation I'm in. "Why am I here if I have died?" I ask. "Well, your universe has an insurance, and the insurance company is affiliated with us. According to my reports, you saved many lives and had a positive influence on many things. After analyzing this, we've deemed it to be worth 95,894 afterlife points. After some deliberation, and considering that you would have earned more achievements without the bug, we decided to round it to 100,000. With that many points you enter the VIP category. Congratulations, sir." "What does that even mean?" I ask. "People with less than 0 points can either go to hell or choose to be reborn to try to earn more points in their new life. People between 0 and 99,999 may go to heaven and, depending on their points, enjoy different perks. With 100,000 or more you enter the VIP category." "That means a customized universe for yourself, where you keep living the greatest years of your life again and again. Here are the documents—just sign, and you will soon be enjoying your life as a hero again." "So I will be able to save the kingdom of Feldem? I will be able to save everyone?" "Well, no. It's just a replay of your previous life, so things won't change from how they happened. We currently have no programmer available to fix the bug, so when that moment comes we will reset the universe and you will live from the start of your hero days again. Of course, you won't remember any of this, so you'll be as happy as you were in your original life." "What is a programmer? When will one be available?" "A programmer is someone who creates the rules for the universe you live in. From your perspective, you could call them a God. I wouldn't recommend waiting for one to be available. Unfortunately, the last one we had quit after becoming depressed over his newest creation, a planet called Earth. That was many thousands of years ago. Now, if you could sign these documents, please." "I don't want that customized-universe thing; it would be all fake!" "Well, it depends on your definition of fake. After all, everything is still code." "But if everything will happen the same way as before, there's no free will. It would just be a copy of the other life, so to me it's fake." "I am a hero. I would rather save real people than fake ones." "Even if that means never seeing your loved ones again?" That question makes my heart ache again. However, I have already made up my mind.	I open my eyes in an unfamiliar room. My head hurts, and I try to remember how I ended up here, but I can only remember up to when the teleport was almost ready. I hear a sound and look that way. A woman is sitting at a desk. She looks young and wears glasses. After a few seconds she notices my gaze. "Oh, you've finally woken up. I was getting bored." "Where am I? How did I end up here? Is Seelena fine? Is Muradel okay? What happened with the monsters?" "A lot of questions... but all right. You are in Afterlife Management Service room number seven hundred sixty-two. You are here because you died. Seelena is dead. Muradel is dead. The monsters destroyed the world." "Huh? Are you joking with me? I won't forgive you for saying my cute girls are dead, you know!" "I am simply stating the facts, sir." "That's ridiculous. I'm the hero of the prophecies. It's impossible for me to die." "Yes, your report definitely said that. I regret to inform you that, according to our logs, we've concluded there was a bug in the code and you died." "Code? What code? How did I die?" "Your universe's code. An unusual combination of factors occurred that weren't considered in the test cases. I'm sorry — on behalf of the company." "I don't understand." "I asked what killed me." "As I said, there were multiple factors involved… but ultimately you could say that it was a mosquito bite." "Me, the great hero, dying from a mosquito bite… I still can't believe it." "Well, there is nothing I can do about that, Sir Hero. I have only been assigned to your case to explain the situation." "Bring me back to my world. I need to save the Kingdom of Feldem!" "Unfortunately, that won't be possible. Your world has already been completely destroyed. No one could stop a young boy and his minions." "I haven't grasped it all, but isn't this your fault? Do something about it!" "Well, we do have a program to revive people, but it is too late. No one is alive in that world anymore. In the past we had a program that could reverse time; however, it has been incompatible since the company decided to switch operating systems." "I don't understand—are you saying that everyone is dead?" "That's right." Tears started falling from my face. It was unseemly, but I couldn't stop. Everything I knew, everyone I knew—gone. I had failed them. The only thing left was my regret. I should have tried to kill them all without my full power. Then at least everyone else would still be alive. I dwelled in my memories: my father training me to be a knight like him, the first monster I had slain, my father's death caused by a high-level demon king, my anger and revenge, unleashing a power I didn't even know I had, killing them all. The great oracle announced that I was the hero of the prophecies. I left my mother alone with my brother to travel the world, trying to save people from evil—saving people, becoming popular, saving Seelena, traveling together, our first kiss. Then we met Muradel; Seelena and Muradel fought to decide who was worthy of my lips, which ended in a draw. We met other girls, formed a bigger party, and I enjoyed my life. But all that is over. The woman with glasses simply watches me in silence. After a while my tears stop, and I try to understand the situation I'm in. "Why am I here if I have died?" I ask. "Well, your universe has an insurance, and the insurance company is affiliated with us. According to my reports, you saved many lives and had a positive influence on many things. After analyzing this, we've deemed it to be worth ninety-five thousand eight hundred ninety-four afterlife points. After some deliberation, and considering that you would have earned more achievements without the bug, we decided to round it to one hundred thousand. With that many points you enter the VIP category. Congratulations, sir." "What does that even mean?" I ask. "People with less than zero points can either go to hell or choose to be reborn to try to earn more points in their new life. People between zero and ninety-nine thousand nine hundred ninety-nine may go to heaven and, depending on their points, enjoy different perks. With one hundred thousand or more you enter the VIP category." "That means a customized universe for yourself, where you keep living the greatest years of your life again and again. Here are the documents—just sign, and you will soon be enjoying your life as a hero again." "So I will be able to save the kingdom of Feldem? I will be able to save everyone?" "Well, no. It's just a replay of your previous life, so things won't change from how they happened. We currently have no programmer available to fix the bug, so when that moment comes we will reset the universe and you will live from the start of your hero days again. Of course, you won't remember any of this, so you'll be as happy as you were in your original life." "What is a programmer? When will one be available?" "A programmer is someone who creates the rules for the universe you live in. From your perspective, you could call them a God. I wouldn't recommend waiting for one to be available. Unfortunately, the last one we had quit after becoming depressed over his newest creation, a planet called Earth. That was many thousands of years ago. Now, if you could sign these documents, please." "I don't want that customized-universe thing; it would be all fake!" "Well, it depends on your definition of fake. After all, everything is still code." "But if everything will happen the same way as before, there's no free will. It would just be a copy of the other life, so to me it's fake." "I am a hero. I would rather save real people than fake ones." "Even if that means never seeing your loved ones again?" That question makes my heart ache again. However, I have already made up my mind.
long_en_294	wiki_en	1,033	en	Only Visiting This Planet is a Christian rock album recorded by Larry Norman in 1972. The album was selected as the second-best album in CCM Magazine's The 100 Greatest Albums in Christian Music. In April 2014, it was announced as one of 25 sound recordings inducted for 2013 into the Library of Congress National Recording Registry, which preserves "cultural, artistic, and/or historical treasures representing the richness and diversity of the American soundscape," making it the first Christian rock album chosen for the registry. History: On September 8, 1972, Norman began recording his second studio album, Only Visiting This Planet, the first album in a projected trilogy, at AIR Studios in London. Often ranked as Norman's best album, it mixed his Christian message with strong political themes and was meant to reach the flower children disillusioned by the government and the church with its abrasive, urban reality of the gospel. In a 1980 interview, Norman explained its purpose: "Only Visiting This Planet is the first part of the trilogy, and represents the present. On the front cover I find myself standing in the middle of New York City, with buildings and traffic pressed around me and my hand on my head kind of saying, 'What is going on in this life? Is this really Earth?' And the back cover is me visiting the site of a previous civilization with its own monoliths—not skyscrapers, but amazing, architecturally sound structures just the same." The Druids apparently constructed Stonehenge to help them observe or worship the sun; their civilization is now gone, as New York will be someday. I'm just standing there, looking around, wondering what happened to kill off this culture and reduce its entire recorded history to a few standing structures. On January 6, 1973, Norman was one of three named Best New Male Artist of the Year by Cashbox. By February 1973, songs from Only Visiting This Planet had been recommended by Billboard for "heavy Top 40 airplay" and were being played on WVVS-FM, KSHE-FM, and WKTK-FM. In 1990, CCM Magazine voted it "the greatest Christian album ever recorded." Only Visiting This Planet was one of 25 sound recordings inducted in 2013 into the Library of Congress National Recording Registry, which preserves recordings as "cultural, artistic and/or historical treasures, representing the richness and diversity of the American soundscape." A statement by the Library of Congress called the album "the key work in the early history of Christian rock," describing Norman as someone who "commented on the world as he saw it from his position as a passionate, idiosyncratic outsider to mainstream churches." After touring South Africa in June and the UK in July, he released in July the songbook "Why Should the Devil Have All the Good Music?", which featured some of his songs from both Upon This Rock and Only Visiting This Planet. In the song "Reader's Digest," Norman sings the verse: "Dear John, who's more popular now? I've been listening to Paul's records. I think he really is dead." The line "Who's more popular now?" references John Lennon's claim that the Beatles were "more popular than Jesus" (see the "Paul Is Dead" rumor). The album features bassist John Wetton—then a member of the progressive rock band Family—who later played with King Crimson and fronted Asia. A three-LP box set containing the entire trilogy in their originally intended forms, titled The Compleat Trilogy (as mentioned on the insert of the Street Level reissue of Only Visiting This Planet), has never been released. Solid Rock Records has issued multiple re-releases. All tracks were composed by Larry Norman. The original LP release order was as on the Verve album. On the Street Level vinyl reissue in 1977, Norman claimed he had always wanted the album to open with "I've Got to Learn to Live Without You"; subsequent re-releases placed that track first and "Why Don't You Look Into Jesus" third. Side 1: "Why Don't You Look into Jesus" – 4:03 "The Outlaw" – 3:52 "I've Got to Learn to Live Without You" – 3:35 "Righteous Rocker #1" – 3:32 "I Wish We'd All Been Ready" – 4:31 Side 2: "I Am Six O'Clock News" – 6:04 "The Great American Novel" – 4:30 "Pardon Me" – 3:36 "Why Should the Devil Have All the Good Music" – 2:37 "Reader's Digest" – 2:43 "Oh, How I Love You" – 0:42 (brief song snippet – not listed on label or album sleeve) Additional tracks on some subsequent releases: "Peace Pollution Revolution" (1971 single) "Righteous Rocker" (rough mix / Hard Rock Version / Delta Swamp Version) "The Outlaw" (demo / Rock Remake / Peace Mix Remake) "Digest" (rock version) / "Reader's Digest" (Hard Rock Remake / Solid Rock Studio Remake) Maximum Planet (The Anthology Series): "I've Got to Learn to Live Without You" – basic master track "Why Don't You Look into Jesus" – master track "I Wish We'd All Been Ready" – basic master track "I Am the Six O'Clock News" – basic master track "Six O'Clock News" – jet fade-in with stewardess "Six O'Clock News" – jet fade-out jam "The Great American Novel" – demo "Pardon Me" – faint vocals; with vocals and no orchestra "Why Should the Devil Have All the Good Music" – vocals 2.0 "Uncredited, Unidentified Song" – spiral out-groove "The Great American Novel" – warm-up demo "I've Got to Learn to Live Without You" – basic track "The Outlaw" – with electric guitars and guide vocal "I Am the Six O'Clock News" – basic track with guitars and guide vocal "I Wish We'd All Been Ready" – with orchestra and no vocals "Why Should the Devil" – with guide vocal "Why Don't You Look into Jesus" – on stage Personnel Larry Norman – vocals, piano John Wetton – bass Keith Smart – drums Mickey Keen – guitar Rod Edwards – piano and backing vocals Roger Hand – backing vocals Gordon Giltrap – guitar Bob Brady – piano Bill Price – engineer See also Larry Norman discography External links Library of Congress essay on album Categories 1972 albums Larry Norman albums United States National Recording Registry recordings United States National Recording Registry albums	Only Visiting This Planet is a Christian rock album recorded by Larry Norman in nineteen seventy-two. The album was selected as the second-best album in CCM Magazine's The one hundred Greatest Albums in Christian Music. In April two thousand fourteen, it was announced as one of twenty-five sound recordings inducted for two thousand thirteen into the Library of Congress National Recording Registry, which preserves "cultural, artistic, and/or historical treasures representing the richness and diversity of the American soundscape," making it the first Christian rock album chosen for the registry. History: On September eight, nineteen seventy-two, Norman began recording his second studio album, Only Visiting This Planet, the first album in a projected trilogy, at AIR Studios in London. Often ranked as Norman's best album, it mixed his Christian message with strong political themes and was meant to reach the flower children disillusioned by the government and the church with its abrasive, urban reality of the gospel. In a nineteen eighty interview, Norman explained its purpose: "Only Visiting This Planet is the first part of the trilogy, and represents the present." On the front cover I find myself standing in the middle of New York City, with buildings and traffic pressed around me and my hand on my head kind of saying, 'What is going on in this life? Is this really Earth?' And the back cover is me visiting the site of a previous civilization with its own monoliths—not skyscrapers, but amazing, architecturally sound structures just the same." The Druids apparently constructed Stonehenge to help them observe or worship the sun; their civilization is now gone, as New York will be someday. I'm just standing there, looking around, wondering what happened to kill off this culture and reduce its entire recorded history to a few standing structures. On January six, nineteen seventy three, Norman was one of three named Best New Male Artist of the Year by Cashbox. By February nineteen seventy three, songs from Only Visiting This Planet had been recommended by Billboard for "heavy Top forty airplay" and were being played on WVVS-FM, KSHE-FM, and WKTK-FM. In nineteen ninety, CCM Magazine voted it "the greatest Christian album ever recorded." Only Visiting This Planet was one of twenty five sound recordings inducted in two thousand thirteen into the Library of Congress National Recording Registry, which preserves recordings as "cultural, artistic and/or historical treasures, representing the richness and diversity of the American soundscape." A statement by the Library of Congress called the album "the key work in the early history of Christian rock," describing Norman as someone who "commented on the world as he saw it from his position as a passionate, idiosyncratic outsider to mainstream churches." After touring South Africa in June and the UK in July, he released in July the songbook "Why Should the Devil Have All the Good Music?", which featured some of his songs from both Upon This Rock and Only Visiting This Planet. In the song "Reader's Digest," Norman sings the verse: "Dear John, who's more popular now? I've been listening to Paul's records. I think he really is dead." The line "Who's more popular now?" references John Lennon's claim that the Beatles were "more popular than Jesus" (see the "Paul Is Dead" rumor). The album features bassist John Wetton—then a member of the progressive rock band Family—who later played with King Crimson and fronted Asia. A three-LP box set containing the entire trilogy in their originally intended forms, titled The Compleat Trilogy (as mentioned on the insert of the Street Level reissue of Only Visiting This Planet), has never been released. Solid Rock Records has issued multiple re-releases. All tracks were composed by Larry Norman. The original LP release order was as on the Verve album. On the Street Level vinyl reissue in nineteen seventy seven, Norman claimed he had always wanted the album to open with "I've Got to Learn to Live Without You"; subsequent re-releases placed that track first and "Why Don't You Look Into Jesus" third. Side one: "Why Don't You Look into Jesus" – four colon zero three "The Outlaw" – three colon fifty two "I've Got to Learn to Live Without You" – three colon thirty five "Righteous Rocker number one" – three colon thirty two "I Wish We'd All Been Ready" – four colon thirty one Side two: "I Am Six O'Clock News" – six colon zero four "The Great American Novel" – four colon thirty "Pardon Me" – three colon thirty six "Why Should the Devil Have All the Good Music" – two colon thirty seven "Reader's Digest" – two colon forty three "Oh, How I Love You" – zero colon forty two (brief song snippet – not listed on label or album sleeve) Additional tracks on some subsequent releases: "Peace Pollution Revolution" (nineteen seventy one single) "Righteous Rocker" (rough mix / Hard Rock Version / Delta Swamp Version) "The Outlaw" (demo / Rock Remake / Peace Mix Remake) "Digest" (rock version) / "Reader's Digest" (Hard Rock Remake / Solid Rock Studio Remake) Maximum Planet (The Anthology Series): "I've Got to Learn to Live Without You" – basic master track "Why Don't You Look into Jesus" – master track "I Wish We'd All Been Ready" – basic master track "I Am the Six O'Clock News" – basic master track "Six O'Clock News" – jet fade-in with stewardess "Six O'Clock News" – jet fade-out jam "The Great American Novel" – demo "Pardon Me" – faint vocals; with vocals and no orchestra "Why Should the Devil Have All the Good Music" – vocals two point zero "Uncredited, Unidentified Song" – spiral out-groove "The Great American Novel" – warm-up demo "I've Got to Learn to Live Without You" – basic track "The Outlaw" – with electric guitars and guide vocal "I Am the Six O'Clock News" – basic track with guitars and guide vocal "I Wish We'd All Been Ready" – with orchestra and no vocals "Why Should the Devil" – with guide vocal "Why Don't You Look into Jesus" – on stage Personnel Larry Norman – vocals, piano John Wetton – bass Keith Smart – drums Mickey Keen – guitar Rod Edwards – piano and backing vocals Roger Hand – backing vocals Gordon Giltrap – guitar Bob Brady – piano Bill Price – engineer See also Larry Norman discography External links Library of Congress essay on album Categories nineteen seventy two albums Larry Norman albums United States National Recording Registry recordings United States National Recording Registry albums.
long_en_328	poet_en	1,036	en	I haven't written too much about what the kids are up to these days. They've been doing really cute things. Madison absolutely loves it when you jump out at her and scare her. Her whole body flinches and then she squeals with laughter, runs away laughing, then comes back for more. I say "BOO" when she comes around the corner, then "I'm gonna get you" as she's running away. We do it for so long and I laugh so hard I usually end up having an asthma attack. Today baby Max came over for a visit. The kids have some matching outfits, so we attempted a photo shoot. At one point we had all three of them on the couch in their bear suits (Max's was a hand-me-down from Jackson). Max was in the middle and Maddie and Jackson were on either side of him. They wouldn't stop looking at each other and laughing! It was such a crack-up. Also today, while Max slept on my chest (AHHHHH) M and J were playing peek-a-boo around a door frame and laughing hysterically. Madison quite literally doesn't stop walking all day long. She is constantly on the move, clutching a toy or binky in one or both hands and sometimes also in her mouth. Jackson still has no teeth. He really isn't interested in walking or even standing unsupported. Eating has gotten a lot better. They are both willing to eat just about anything I put in front of them. Just today I started putting more than one piece of food on their tray and they did okay. Before, they would put all of it in their mouths, so I had to dole it out in individual bites. One day last week they both decided they were going to use sippy cups for their intended purpose rather than as chew toys. Just like that, they started sucking. Now they love to drink milk out of their sippy cups. Some of our staples at mealtime are eggs, rye bread/toast, plain yogurt, zucchini pancakes (frozen and they LOVE them), black beans, grilled cheese, bananas, strawberries, blueberries, carrots, and sweet potato (although they both boycotted that today). I'm trying to work in whole wheat pasta, but so far they haven't been a fan. I'll keep working. I'm finally entering a photo contest at the Twinkies. Please take the time to go over and vote for Jackson (click on the word Twinkies to go to her blog and vote). I finally figured out where I want to store the rest of my photos that I don't post here. I was having issues finding a place that didn't have limits or fees (I'm very cheap). Then I realized that I can just use my professional proofing site that I already have. It turns out it's much faster to upload to because it doesn't store the huge original files. So if you're looking for pictures of Madison and Jackson in folders by month, click here and bookmark it. You don't need to log in, just click "continue." I'm working on some other albums for friends and family. At 8 p.m. we turned off all the lights and lit candles around the house. It was bathtime for the kids, so I brought candles and a flashlight into the bathroom and they had a nice candlelit bath. It's 10:45 now and we haven't bothered to turn the lights back on. We may even make a habit of this—if nothing else, it will help our electric bill. My grandmother came out from western Mass to visit for Easter. On Monday I picked her up at my parents' house, we took the kids to the park, and then had lunch; afterward she came back to our house. It was cold so we didn't stay long at the park. This wasn't their first time in swings, but it was the first time they really enjoyed it. They were laughing out loud—Jackson kept looking at Maddie and laughing. There was another baby on the swing next to Maddie, and she couldn't take her eyes off him. I dragged Grandma around with us again on Tuesday. We went to lunch again, ran some errands, and visited my friend Melanie, who had her first baby two weeks ago. I got to hold the baby, and she fell asleep on my chest until we had to leave. That left poor Melanie—recovering from a c-section and sleep deprived—to look after my two little terrors who were trying to get into everything. They weren't that bad, but she did have to keep an eye on them for me. Alice is a very sweet, cute baby, but she's been giving her mom and dad a hard time and cries a lot. Our second Easter was quite a bit easier than the first. We started off in Gloucester as usual, with Jackson catching a 20-minute nap on the way (none for Maddie). They were both pretty clingy when we first got there, but after an hour or so they were running all over. Jackson crawled around the living room playing with toys, and Maddie continuously walked back and forth from the kitchen to the living room to grab a bite of someone's cookie. I brought Maddie outside with me while the older kids hunted for eggs; some of them were nice enough to give her one. We were there for four hours without naps for either child, then we headed to my aunt's house in Salem. They both snoozed for about 30 minutes in the car. As soon as we arrived they started tearing the place apart looking for things to get into. There were birthday presents to open and Easter baskets from Grandma. We visited for about an hour and a half, then headed home. We live only about a seven-minute drive from my aunt's house, but the entire way home Maddie was grunting or squealing and Chris kept imitating her.	I haven't written too much about what the kids are up to these days. They've been doing really cute things. Madison absolutely loves it when you jump out at her and scare her. Her whole body flinches and then she squeals with laughter, runs away laughing, then comes back for more. I say "BOO" when she comes around the corner, then "I'm gonna get you" as she's running away. We do it for so long and I laugh so hard I usually end up having an asthma attack. Today baby Max came over for a visit. The kids have some matching outfits, so we attempted a photo shoot. At one point we had all three of them on the couch in their bear suits (Max's was a hand-me-down from Jackson). Max was in the middle and Maddie and Jackson were on either side of him. They wouldn't stop looking at each other and laughing! It was such a crack-up. Also today, while Max slept on my chest (AHHHHH) M and J were playing peek-a-boo around a door frame and laughing hysterically. Madison quite literally doesn't stop walking all day long. She is constantly on the move, clutching a toy or binky in one or both hands and sometimes also in her mouth. Jackson still has no teeth. He really isn't interested in walking or even standing unsupported. Eating has gotten a lot better. They are both willing to eat just about anything I put in front of them. Just today I started putting more than one piece of food on their tray and they did okay. Before, they would put all of it in their mouths, so I had to dole it out in individual bites. One day last week they both decided they were going to use sippy cups for their intended purpose rather than as chew toys. Just like that, they started sucking. Now they love to drink milk out of their sippy cups. Some of our staples at mealtime are eggs, rye bread/toast, plain yogurt, zucchini pancakes (frozen and they LOVE them), black beans, grilled cheese, bananas, strawberries, blueberries, carrots, and sweet potato (although they both boycotted that today). I'm trying to work in whole wheat pasta, but so far they haven't been a fan. I'll keep working. I'm finally entering a photo contest at the Twinkies. Please take the time to go over and vote for Jackson (click on the word Twinkies to go to her blog and vote). I finally figured out where I want to store the rest of my photos that I don't post here. I was having issues finding a place that didn't have limits or fees (I'm very cheap). Then I realized that I can just use my professional proofing site that I already have. It turns out it's much faster to upload to because it doesn't store the huge original files. So if you're looking for pictures of Madison and Jackson in folders by month, click here and bookmark it. You don't need to log in, just click "continue." I'm working on some other albums for friends and family. At eight p.m. we turned off all the lights and lit candles around the house. It was bathtime for the kids, so I brought candles and a flashlight into the bathroom and they had a nice candlelit bath. It's ten:forty five now and we haven't bothered to turn the lights back on. We may even make a habit of this—if nothing else, it will help our electric bill. My grandmother came out from western Mass to visit for Easter. On Monday I picked her up at my parents' house, we took the kids to the park, and then had lunch; afterward she came back to our house. It was cold so we didn't stay long at the park. This wasn't their first time in swings, but it was the first time they really enjoyed it. They were laughing out loud—Jackson kept looking at Maddie and laughing. There was another baby on the swing next to Maddie, and she couldn't take her eyes off him. I dragged Grandma around with us again on Tuesday. We went to lunch again, ran some errands, and visited my friend Melanie, who had her first baby two weeks ago. I got to hold the baby, and she fell asleep on my chest until we had to leave. That left poor Melanie—recovering from a c-section and sleep deprived—to look after my two little terrors who were trying to get into everything. They weren't that bad, but she did have to keep an eye on them for me. Alice is a very sweet, cute baby, but she's been giving her mom and dad a hard time and cries a lot. Our second Easter was quite a bit easier than the first. We started off in Gloucester as usual, with Jackson catching a twenty-minute nap on the way (none for Maddie). They were both pretty clingy when we first got there, but after an hour or so they were running all over. Jackson crawled around the living room playing with toys, and Maddie continuously walked back and forth from the kitchen to the living room to grab a bite of someone's cookie. I brought Maddie outside with me while the older kids hunted for eggs; some of them were nice enough to give her one. We were there for four hours without naps for either child, then we headed to my aunt's house in Salem. They both snoozed for about thirty minutes in the car. As soon as we arrived they started tearing the place apart looking for things to get into. There were birthday presents to open and Easter baskets from Grandma. We visited for about an hour and a half, then headed home. We live only about a seven-minute drive from my aunt's house, but the entire way home Maddie was grunting or squealing and Chris kept imitating her.
long_en_174	paper_en	2,037	en	Some of these approaches are retrieval-based, which augment language models via fetching related documents and including the retrieved results into contexts. Our work is complementary to these works, as our attention mechanism is unmodified during inference. Many works modify multi-head attention to be approximated ones. They alleviate the quadratic complexity of the self-attention computation. For example, Longformer and BigBird use sparse attention to handle long sequences. Other works utilize memory mechanisms as a compression on past inputs, to look up relevant tokens. One limitation of these works is that these compressions have a large gap to full attention, making it infeasible to fine-tune pre-trained LLMs. Although our work also involves an approximation of attention mechanism, it has a similar shape and a small gap to standard attention. This enables fine-tuning pre-trained LLMs on S2-Attn and maintain full attention during inference. Long-context LLMs. LLMs are typically pre-trained with a pre-defined context length, such as 2048 for LLaMA and 4096 for Llama2. Training LLMs with long context from scratch is prohibitively expensive for most researchers. Recently, several works have tried to extend the context length of LLMs via fine-tuning. Position Interpolation modifies rotary position encoding and extends the context length of LLaMA to 32768. Focused Transformer utilizes contrastive learning to train LongLLaMA. Both of them rely on full fine-tuning, which is computationally expensive (128 A100 GPUs / 128 TPUv3 for training). Landmark attention is an efficient approach, but somewhat lossy. It compresses long context inputs into retrieved tokens. Our method saves substantial fine-tuning costs, while preserving the quality of the original attention. Ours maintain full access to the entire input via unmodified attention during inference. Some literature focuses on the position embedding modification of LLMs for long context extension, including Position Interpolation, NTK-aware, Yarn, positional Skipping, and methods based on out-of-distribution analysis. Our method focuses on efficient fine-tuning and retaining the original architecture during inference, which is orthogonal to these position embedding methods. Our models apply the Position Interpolation in experiments. Efficient Fine-tuning. This work is based on LoRA, a classical efficient fine-tuning approach. In addition to LoRA, there are many other parameter-efficient fine-tuning methods, including prompt tuning, prefix tuning, hidden state tuning, bias tuning, and masked weight learning. Input-tuning introduces an adapter to tune input embedding. Although the input embedding layers are also trainable in ours, this is not enough for long context extension. We make a comprehensive analysis on layer types in experiments. Existing work shows sparse masks can effectively save training costs and avoid performance drops. Section 3: LongLoRA Subsection 3.1: Background Transformer. LLMs are typically built with transformers. Taking Llama2 for example, as shown in Figure 2, an LLM model consists of an embedding input layer and a number of decoder layers. Each decoder layer comprises a self-attention module. It maps input features into a set of queries, keys, and values {q, k, v}, via linear projection layers with weight matrices {Wq, Wk, Wv}. Given {q, k, v}, it computes the outputs by taking the softmax of the query-key dot product and multiplying it by the value. The outputs are then projected by a linear layer with a weight matrix Wo. And MLP layers are followed. Before and after self-attention modules, layer normalization is applied. A final normalization is conducted after all decoder layers. For long sequences, self-attention struggles with computation cost, which is quadratic to the sequence length. This dramatically slows down the training procedure and increases GPU memory costs. Low-rank Adaptation. LoRA hypothesizes that the weight updates in pre-trained models have a low intrinsic rank during adaptation. For a pre-trained weight matrix W, it is updated with a low-rank decomposition where the change in weight is represented as the product of two smaller matrices, B and A. The rank r is much smaller than the dimensions of the original matrix. During training, W is frozen with no gradient updates, while A and B are trainable. This is the reason why LoRA training is much more efficient than full fine-tuning. In the Transformer structure, LoRA only adapts the attention weights (Wq, Wk, Wv, Wo) and freezes all other layers, including MLP and normalization layers. This manner is simple and parameter-efficient. However, we empirically show that only low-rank adaptation in attention weights does not work for long context extension. Subsection 3.2: Shifted Sparse Attention Standard self-attention costs are quadratic in sequence length, making LLMs on long sequences high memory cost and slow. To avoid this issue during training, we propose Shifted Sparse Attention (S2-Attn), as shown in Figure 2. In the following, we make a pilot study and explain our design step by step. Pilot Study. We build up a standard baseline that is trained and tested with full attention and fine-tuning, which presents consistently good quality in various context lengths. The first trial is to train with short attention, only pattern 1 in Figure 2. As we know for a long context, the high cost mainly comes from self-attention modules. Thus, in this trial, since the input is long, we split into several groups in self-attention. For example, the model takes 8192 tokens as input in both the training and testing stages, but self-attention is conducted in each group with a 2048 size. The group number is 4, as ablated in Section B.2 in the appendix. This pattern is efficient but still does not work in a very long context. The perplexity becomes larger as the context length increases. The reason behind this is that there is no information exchange between different groups. To introduce communication between groups, we include a shifted pattern, as shown in Figure 2. We shift the group partition by half group size in half attention heads. Taking the overall 8192 context length for example, in pattern 1, the first group conducts self-attention from 1st to 2048th tokens. In Pattern 2, the group partition is shifted by 1024. The first attention group begins from 1025th and ends at 3072th tokens, while the first and the last 1024 tokens belong to the same group. We use patterns 1 and 2 in each half self-attention heads respectively. This manner does not increase additional computation costs but enables the information flow between different groups. We show that it gets close to the standard attention baseline. Consistency to Full Attention. Existing efficient attention designs can also improve the efficiency of long-context LLMs. However, most of them are not suitable for long-context fine-tuning. Because, these transformers, designed for training from scratch, have gaps to the standard full attention, which is used in pre-training. We show that S2-Attn not only enables efficient fine-tuning but also supports full attention testing. Although other attentions can also be used in long context fine-tuning, models must be tested with the attention used during fine-tuning. Shifting prevents models from being over-fitted to specific attention patterns. Easy Implementation. S2-Attn is easy to implement. It involves only two steps: (1) shifting tokens in half attention heads, and (2) transposing features from token dimension to batch dimension. Two lines of code are enough. Subsection 3.3: Improved LoRA for Long Context LoRA is an efficient and popular manner for adapting LLMs to other datasets. It saves much trainable parameters and memory cost, compared to full fine-tuning. However, adapting LLMs from short context length to long is not easy. We empirically observe an obvious gap between LoRA and full fine-tuning. The gap between LoRA and full fine-tuning grows as the target context length becomes larger. And LoRA with larger ranks cannot reduce the gap. To bridge this gap, we open embedding and normalization layers for training. They occupy limited parameters but make effects for long context adaptation. Especially for normalization layers, the parameters are only 0.004 percent in the whole Llama2 7B. We denote this improved version of LoRA as LoRA+ in experiments. Section 4: Experiment Subsection 4.1: Experimental Settings Models We extend the pre-trained 7B, 13B, and 70B Llama2 models. The maximum extended context window sizes are up to 100k for 7B models, 65536 for 13B models, and 32768 for 70B models. The position indices for these models are re-scaled with Position Interpolation. Training Procedure We follow most training hyper-parameters in Position Interpolation, except that our batch size is smaller as we use a single 8 by A100 GPUs machine in some cases. All models are fine-tuned via the next token prediction objective. We use AdamW with beta 1 = 0.9 and beta 2 = 0.95. The learning rate is set to 2 times 10 to the power of negative 5 for 7B and 13B models, and 10 to the power of negative 5 for 70B models. We also use a linear learning rate warmup. The weight decay is zero. We set the per-device batch size as 1 and gradient accumulation steps as 8, which means that the global batch size equals 64, using 8 GPUs. We train our models for 1000 steps. Datasets We use the Redpajama dataset for training. We evaluate the long-sequence language modeling performance of our fine-tuned models on the book corpus dataset PG19 and the cleaned Arxiv Math proof-pile dataset. We use the test split of PG19, consisting of 100 documents. For the proof-pile dataset, we also use the test split of it for evaluation. We follow Position Interpolation for proof-pile data processing. We evaluate perplexity by using a sliding window approach with S=256. Subsection 4.2: Main Results Long-sequence Language Modeling. We report the perplexity for our models and baseline on proof-pile and PG19 datasets. Under certain training context lengths, our models achieve better perplexity with longer context sizes. This indicates the effectiveness of our efficient fine-tuning method. For the same training and evaluation context length cases, the perplexity decreases as the context size increases. By increasing the context window size from 8192 to 32768, for the Llama2 7B model, we observe that the perplexity gets better from 2.72 to 2.50 by -0.22. For Llama2 13B model, we observe that the perplexity reduces by -0.28. We further examine the maximum context length that we can fine-tune on a single 8 by A100 machine. We extend Llama2 7B, 13B, and 70B to 100k, 65536, and 32768 context length respectively. LongLoRA achieves promising results on these extremely large settings. In addition, we find some perplexity degradation on small context sizes for the extended models. This is a known limitation of Position Interpolation. Retrieval-based Evaluation. We conduct experiments on retrieval in long contexts. We compare our model with other open LLMs on the topic retrieval task introduced in LongChat. This task is to retrieve the target topic from a very long conversation, with lengths varying from 3k, 6k, 10k, 13k, to 16k. As some questions in LongChat are longer than 16k, we fine-tuned Llama2 13B with a context length of 18k. The training cost is similar to that for 16k. Our model achieves comparable performance to LongChat-13B, the state-of-the-art model in this task. Unlike LongChat-13B, which is fully fine-tuned on self-collected long context conversation text, our model is efficiently adapted on RedPajama via next-token generation. Our model even slightly outperforms LongChat-13B in the 16k evaluation. We present the passkey retrieval accuracy of our model, following Landmark Attention. This task has also been adopted by other literature. In this task, the models need to find a random passkey hidden in a long document. We show the document format is in Section B.1 in the appendix. We study Llama2 7B and our LongLoRA model which fine-tunes Llama2 7B with 32768 context length. We test the passkey retrieval accuracy from 1k to 34k, with an interval of roughly 1k (as the sentence length can not be precisely controlled). For each document length, we test the model 10 times with different random passkey values.	Some of these approaches are retrieval-based, which augment language models via fetching related documents and including the retrieved results into contexts. Our work is complementary to these works, as our attention mechanism is unmodified during inference. Many works modify multi-head attention to be approximated ones. They alleviate the quadratic complexity of the self-attention computation. For example, Longformer and BigBird use sparse attention to handle long sequences. Other works utilize memory mechanisms as a compression on past inputs, to look up relevant tokens. One limitation of these works is that these compressions have a large gap to full attention, making it infeasible to fine-tune pre-trained LLMs. Although our work also involves an approximation of attention mechanism, it has a similar shape and a small gap to standard attention. This enables fine-tuning pre-trained LLMs on S two Attn and maintain full attention during inference. Long-context LLMs. LLMs are typically pre-trained with a pre-defined context length, such as two thousand forty-eight for LLaMA and four thousand ninety-six for Llama2. Training LLMs with long context from scratch is prohibitively expensive for most researchers. Recently, several works have tried to extend the context length of LLMs via fine-tuning. Position Interpolation modifies rotary position encoding and extends the context length of LLaMA to thirty-two thousand seven hundred sixty-eight. Focused Transformer utilizes contrastive learning to train LongLLaMA. Both of them rely on full fine-tuning, which is computationally expensive (one hundred twenty-eight A one hundred GPUs slash one hundred twenty-eight TPUv three for training). Landmark attention is an efficient approach, but somewhat lossy. It compresses long context inputs into retrieved tokens. Our method saves substantial fine-tuning costs, while preserving the quality of the original attention. Ours maintain full access to the entire input via unmodified attention during inference. Some literature focuses on the position embedding modification of LLMs for long context extension, including Position Interpolation, NTK-aware, Yarn, positional Skipping, and methods based on out-of-distribution analysis. Our method focuses on efficient fine-tuning and retaining the original architecture during inference, which is orthogonal to these position embedding methods. Our models apply the Position Interpolation in experiments. Efficient Fine-tuning. This work is based on LoRA, a classical efficient fine-tuning approach. In addition to LoRA, there are many other parameter-efficient fine-tuning methods, including prompt tuning, prefix tuning, hidden state tuning, bias tuning, and masked weight learning. Input-tuning introduces an adapter to tune input embedding. Although the input embedding layers are also trainable in ours, this is not enough for long context extension. We make a comprehensive analysis on layer types in experiments. Existing work shows sparse masks can effectively save training costs and avoid performance drops. Section three: LongLoRA Subsection three point one: Background Transformer. LLMs are typically built with transformers. Taking Llama2 for example, as shown in Figure two, an LLM model consists of an embedding input layer and a number of decoder layers. Each decoder layer comprises a self-attention module. It maps input features into a set of queries, keys, and values q, k, v, via linear projection layers with weight matrices Wq, Wk, Wv. Given q, k, v, it computes the outputs by taking the softmax of the query-key dot product and multiplying it by the value. The outputs are then projected by a linear layer with a weight matrix Wo. And MLP layers are followed. Before and after self-attention modules, layer normalization is applied. A final normalization is conducted after all decoder layers. For long sequences, self-attention struggles with computation cost, which is quadratic to the sequence length. This dramatically slows down the training procedure and increases GPU memory costs. Low-rank Adaptation. LoRA hypothesizes that the weight updates in pre-trained models have a low intrinsic rank during adaptation. For a pre-trained weight matrix W, it is updated with a low-rank decomposition where the change in weight is represented as the product of two smaller matrices, B and A. The rank r is much smaller than the dimensions of the original matrix. During training, W is frozen with no gradient updates, while A and B are trainable. This is the reason why LoRA training is much more efficient than full fine-tuning. In the Transformer structure, LoRA only adapts the attention weights (W q, W k, W v, W o) and freezes all other layers, including MLP and normalization layers. This manner is simple and parameter-efficient. However, we empirically show that only low-rank adaptation in attention weights does not work for long context extension. Subsection three point two: Shifted Sparse Attention Standard self-attention costs are quadratic in sequence length, making LLMs on long sequences high memory cost and slow. To avoid this issue during training, we propose Shifted Sparse Attention (S two Attn), as shown in Figure two. In the following, we make a pilot study and explain our design step by step. Pilot Study. We build up a standard baseline that is trained and tested with full attention and fine-tuning, which presents consistently good quality in various context lengths. The first trial is to train with short attention, only pattern one in Figure two. As we know for a long context, the high cost mainly comes from self-attention modules. Thus, in this trial, since the input is long, we split into several groups in self-attention. For example, the model takes eight thousand one hundred ninety-two tokens as input in both the training and testing stages, but self-attention is conducted in each group with a two thousand forty-eight size. The group number is four, as ablated in Section B point two in the appendix. This pattern is efficient but still does not work in a very long context. The perplexity becomes larger as the context length increases. The reason behind this is that there is no information exchange between different groups. To introduce communication between groups, we include a shifted pattern, as shown in Figure two. We shift the group partition by half group size in half attention heads. Taking the overall eight thousand one hundred ninety two context length for example, in pattern one, the first group conducts self-attention from first to two thousand forty eighth tokens. In Pattern two, the group partition is shifted by one thousand twenty four. The first attention group begins from one thousand twenty fifth and ends at three thousand seventy second tokens, while the first and the last one thousand twenty four tokens belong to the same group. We use patterns one and two in each half self-attention heads respectively. This manner does not increase additional computation costs but enables the information flow between different groups. We show that it gets close to the standard attention baseline. Consistency to Full Attention. Existing efficient attention designs can also improve the efficiency of long-context LLMs. However, most of them are not suitable for long-context fine-tuning. Because, these transformers, designed for training from scratch, have gaps to the standard full attention, which is used in pre-training. We show that S two Attn not only enables efficient fine-tuning but also supports full attention testing. Although other attentions can also be used in long context fine-tuning, models must be tested with the attention used during fine-tuning. Shifting prevents models from being over-fitted to specific attention patterns. Easy Implementation. S two Attn is easy to implement. It involves only two steps: (one) shifting tokens in half attention heads, and (two) transposing features from token dimension to batch dimension. Two lines of code are enough. Subsection three point three: Improved LoRA for Long Context LoRA is an efficient and popular manner for adapting LLMs to other datasets. It saves much trainable parameters and memory cost, compared to full fine-tuning. However, adapting LLMs from short context length to long is not easy. We empirically observe an obvious gap between LoRA and full fine-tuning. The gap between LoRA and full fine-tuning grows as the target context length becomes larger. And LoRA with larger ranks cannot reduce the gap. To bridge this gap, we open embedding and normalization layers for training. They occupy limited parameters but make effects for long context adaptation. Especially for normalization layers, the parameters are only zero point zero zero four percent in the whole Llama two seven B. We denote this improved version of LoRA as LoRA plus in experiments. Section four: Experiment Subsection four point one: Experimental Settings Models We extend the pre-trained seven B, thirteen B, and seventy B Llama two models. The maximum extended context window sizes are up to one hundred thousand for seven B models, sixty five thousand five hundred thirty six for thirteen B models, and thirty two thousand seven hundred sixty eight for seventy B models. The position indices for these models are re-scaled with Position Interpolation. Training Procedure We follow most training hyper-parameters in Position Interpolation, except that our batch size is smaller as we use a single eight by A one hundred GPUs machine in some cases. All models are fine-tuned via the next token prediction objective. We use AdamW with beta one equals zero point nine and beta two equals zero point nine five. The learning rate is set to two times ten to the power of negative five for seven B and thirteen B models, and ten to the power of negative five for seventy B models. We also use a linear learning rate warmup. The weight decay is zero. We set the per-device batch size as one and gradient accumulation steps as eight, which means that the global batch size equals sixty four, using eight GPUs. We train our models for one thousand steps. Datasets We use the Redpajama dataset for training. We evaluate the long-sequence language modeling performance of our fine-tuned models on the book corpus dataset PG19 and the cleaned Arxiv Math proof-pile dataset. We use the test split of PG19, consisting of one hundred documents. For the proof-pile dataset, we also use the test split of it for evaluation. We follow Position Interpolation for proof-pile data processing. We evaluate perplexity by using a sliding window approach with S equals two hundred fifty six. Subsection four point two: Main Results Long-sequence Language Modeling. We report the perplexity for our models and baseline on proof-pile and PG19 datasets. Under certain training context lengths, our models achieve better perplexity with longer context sizes. This indicates the effectiveness of our efficient fine-tuning method. For the same training and evaluation context length cases, the perplexity decreases as the context size increases. By increasing the context window size from eight thousand one hundred ninety-two to thirty-two thousand seven hundred sixty-eight, for the Llama two seven B model, we observe that the perplexity gets better from two point seven two to two point five zero by negative zero point two two. For Llama two thirteen B model, we observe that the perplexity reduces by negative zero point two eight. We further examine the maximum context length that we can fine-tune on a single eight by A one hundred machine. We extend Llama two seven B, Llama two thirteen B, and Llama two seventy B to one hundred thousand, sixty-five thousand five hundred thirty-six, and thirty-two thousand seven hundred sixty-eight context length respectively. LongLoRA achieves promising results on these extremely large settings. In addition, we find some perplexity degradation on small context sizes for the extended models. This is a known limitation of Position Interpolation. Retrieval-based Evaluation. We conduct experiments on retrieval in long contexts. We compare our model with other open LLMs on the topic retrieval task introduced in LongChat. This task is to retrieve the target topic from a very long conversation, with lengths varying from three thousand, six thousand, ten thousand, thirteen thousand, to sixteen thousand. As some questions in LongChat are longer than sixteen thousand, we fine-tuned Llama two thirteen B with a context length of eighteen thousand. The training cost is similar to that for sixteen thousand. Our model achieves comparable performance to LongChat thirteen B, the state-of-the-art model in this task. Unlike LongChat thirteen B, which is fully fine-tuned on self-collected long context conversation text, our model is efficiently adapted on RedPajama via next-token generation. Our model even slightly outperforms LongChat thirteen B in the sixteen k evaluation. We present the passkey retrieval accuracy of our model, following Landmark Attention. This task has also been adopted by other literature. In this task, the models need to find a random passkey hidden in a long document. We show the document format is in Section B point one in the appendix. We study Llama two seven B and our LongLoRA model which fine-tunes Llama two seven B with thirty two thousand seven hundred sixty eight context length. We test the passkey retrieval accuracy from one k to thirty four k, with an interval of roughly one k (as the sentence length can not be precisely controlled). For each document length, we test the model ten times with different random passkey values.
long_en_317	wiki_en	921	en	Hindus have found support for, or ideas foreshadowing, evolutionary concepts in scriptures, such as the mytheme of the Dashavatara, the incarnations of Vishnu beginning with a fish. Reception in India: In India, there were few references to Darwinism in the 1800s. Although elements of Victorian England opposed Darwinism, some Hindu traditions already contained notions of common ancestry between humans and animals. While the creation–evolution controversy has been widely debated in the United States, the Middle East, and parts of Africa, it is an insignificant issue in India, largely because of its Hindu-majority population. Most Indian scientists accept biological evolution, and it is taught in Indian universities. Spiritual evolution: Many Hindu reformers compare the Samkhya philosophy—specifically the term parinama and the concept of evolutes—with Darwinism. David Lagourie Gosling has suggested that Swami Vivekananda based much of his cosmological and biological thought on Samkhya. Influenced by Western thought and esotericism, Vivekananda and Sri Aurobindo developed a view of reincarnation in which an involution of the Divine into matter takes place and the person must evolve over multiple lives until the Divine recognizes its true nature and liberation is attained. Hindu creationism: Hindu creationism, also known as Vedic creationism, is a form of religious old-earth creationism. Historian of science Ronald Numbers has commented that "Hindu Creationists have insisted on the antiquity of humans, who they believe appeared fully formed as long, perhaps, as trillions of years ago." The views of Hindu creationism are based on the Vedas, which depict an extreme antiquity of the universe and the history of the Earth. The emergence of modern Vedic creationism has been linked to Dayananda Saraswati, the founder of Arya Samaj. In his book Satyarth Prakash, Saraswati promoted anti-evolutionary views and took a literal reading of the Vedas. He argued that God designed the physical bodies of all species 1.96 billion years ago on Earth and on other planets at the beginning of the present cosmic cycle. He stated that God conjoined the bodies with pre-existing souls and that different species were created and assigned to souls in accordance with their karma from the previous cosmic cycle. In a public lecture, Saraswati condemned Darwinian evolution but misunderstood the concept of common descent by asking why monkeys no longer evolve into humans. Vedic creationism holds a view of the world derived largely from the Bhagavad Gita. It was promoted by A. C. Bhaktivedanta Swami Prabhupada, the founder of ISKCON, who referred to Charles Darwin and his followers as "rascals." Vedic creationism was also promoted by ISKCON devotees Michael Cremo and Richard L. Thompson, authors of the 1993 book Forbidden Archeology. They argue that human beings are a distinct species that has existed for billions of years. Vedic creationists are known to search for anomalies and reinterpret the fossil record to make it fit with their metaphysical assumptions. Vanara: The Sanskrit epics of the Hindus mention several exotic creatures, including ape-like humanoids. The Ramayana speaks of the Vanaras, an ape-like species (ape-men) with human intelligence that existed millions of years ago alongside modern humans. Dashavatara: The order of the Dashavatara (the ten principal avatars of the god Vishnu) is interpreted by some to convey Darwinian evolution. British geneticist and evolutionary biologist J. B. S. Haldane opined that they are a true sequential depiction of the great unfolding of evolution. According to this interpretation, like the evolutionary process itself, the first avatar is a fish, Matsya, depicting aquatic life; then the aquatic reptile turtle, Kurma, depicting creatures moving to land; then a mammal, the boar Varaha; then Narasimha, a man-lion being, sometimes taken to mean creatures like the Okapi or Archaeopteryx; then Vamana, the dwarf hominid. Parashurama depicts humans in a caveman stage, followed by Rama, who depicts the rise of civilization and kingdoms. Sometimes, when Balarama is taken into account, he is taken to represent the growth of agriculture. Krishna is taken to symbolize the growth of art and crafts. The 9th avatar, Kalki, is not yet born. Hinduism mentions Kalki will be born at the end of Kali Yuga; the 9th avatar is expected to come during the later part of the Suvarna Kaliyuga. See also: Creation myth; Ichthys; Relationship between religion and science. References / Further reading: Evolution theory - C. Mackenzie Brown, Hindu Perspectives on Evolution: Darwin, Dharma, and Design (Routledge Hindu Studies Series), Routledge, 2012. - C. Mackenzie Brown (ed.), Asian Religious Responses to Darwinism: Evolutionary Theories in Middle Eastern, South Asian, and East Asian Cultural Contexts, Springer Nature, 2020. Creationism - Cavanaugh, Michael A. (1983). A Sociological Account of Scientific Creationism: Science, True Science, Pseudoscience. Unpublished doctoral dissertation, University of Pittsburgh. - Eve, Harold (1990). Creationist Movement in Modern America. Twayne Publishers. - "Vedic creationism" - Cremo, Michael A., and Richard L. Thompson, Forbidden Archeology: The Full Unabridged Edition. Torchlight Publishing; 2nd rev. ed., January 1998. - Cremo, Michael A., Forbidden Archeology's Impact: How a Controversial New Book Shocked the Scientific Community and Became an Underground Classic. Torchlight Publishing, January 1998. - Cremo, Michael A., The Hidden History of the Human Race (The Condensed Edition of Forbidden Archeology). Cremo. Torchlight Publishing, May 15, 1999. ISBN 0892133252. Hindu nationalism. Nanda, Meera. Prophets Facing Backward: Postmodern Critiques of Science and the Making of Hindu Nationalism in India. Rutgers University Press, 2003. External links: Hinduism and Science — Editorial in The Hindu, April 20, 2004. Dharma vs. Darwin? — Swami B.V. Tripurari. Beliefnet article describing Hindu perspectives on evolution. The Perils of Vedic "Science" — Meera Nanda. Beliefnet article on Hindu science and evolutionary theories.	Hindus have found support for, or ideas foreshadowing, evolutionary concepts in scriptures, such as the mytheme of the Dashavatara, the incarnations of Vishnu beginning with a fish. Reception in India: In India, there were few references to Darwinism in the eighteen hundreds. Although elements of Victorian England opposed Darwinism, some Hindu traditions already contained notions of common ancestry between humans and animals. While the creation–evolution controversy has been widely debated in the United States, the Middle East, and parts of Africa, it is an insignificant issue in India, largely because of its Hindu-majority population. Most Indian scientists accept biological evolution, and it is taught in Indian universities. Spiritual evolution: Many Hindu reformers compare the Samkhya philosophy—specifically the term parinama and the concept of evolutes—with Darwinism. David Lagourie Gosling has suggested that Swami Vivekananda based much of his cosmological and biological thought on Samkhya. Influenced by Western thought and esotericism, Vivekananda and Sri Aurobindo developed a view of reincarnation in which an involution of the Divine into matter takes place and the person must evolve over multiple lives until the Divine recognizes its true nature and liberation is attained. Hindu creationism: Hindu creationism, also known as Vedic creationism, is a form of religious old-earth creationism. Historian of science Ronald Numbers has commented that "Hindu Creationists have insisted on the antiquity of humans, who they believe appeared fully formed as long, perhaps, as trillions of years ago." The views of Hindu creationism are based on the Vedas, which depict an extreme antiquity of the universe and the history of the Earth. The emergence of modern Vedic creationism has been linked to Dayananda Saraswati, the founder of Arya Samaj. In his book Satyarth Prakash, Saraswati promoted anti-evolutionary views and took a literal reading of the Vedas. He argued that God designed the physical bodies of all species one point nine six billion years ago on Earth and on other planets at the beginning of the present cosmic cycle. He stated that God conjoined the bodies with pre-existing souls and that different species were created and assigned to souls in accordance with their karma from the previous cosmic cycle. In a public lecture, Saraswati condemned Darwinian evolution but misunderstood the concept of common descent by asking why monkeys no longer evolve into humans. Vedic creationism holds a view of the world derived largely from the Bhagavad Gita. It was promoted by A. C. Bhaktivedanta Swami Prabhupada, the founder of ISKCON, who referred to Charles Darwin and his followers as "rascals." Vedic creationism was also promoted by ISKCON devotees Michael Cremo and Richard L. Thompson, authors of the one thousand nine hundred ninety-three book Forbidden Archeology. They argue that human beings are a distinct species that has existed for billions of years. Vedic creationists are known to search for anomalies and reinterpret the fossil record to make it fit with their metaphysical assumptions. Vanara: The Sanskrit epics of the Hindus mention several exotic creatures, including ape-like humanoids. The Ramayana speaks of the Vanaras, an ape-like species (ape-men) with human intelligence that existed millions of years ago alongside modern humans. Dashavatara: The order of the Dashavatara (the ten principal avatars of the god Vishnu) is interpreted by some to convey Darwinian evolution. British geneticist and evolutionary biologist J. B. S. Haldane opined that they are a true sequential depiction of the great unfolding of evolution. According to this interpretation, like the evolutionary process itself, the first avatar is a fish, Matsya, depicting aquatic life; then the aquatic reptile turtle, Kurma, depicting creatures moving to land; then a mammal, the boar Varaha; then Narasimha, a man-lion being, sometimes taken to mean creatures like the Okapi or Archaeopteryx; then Vamana, the dwarf hominid. Parashurama depicts humans in a caveman stage, followed by Rama, who depicts the rise of civilization and kingdoms. Sometimes, when Balarama is taken into account, he is taken to represent the growth of agriculture. Krishna is taken to symbolize the growth of art and crafts. The ninth avatar, Kalki, is not yet born. Hinduism mentions Kalki will be born at the end of Kali Yuga; the ninth avatar is expected to come during the later part of the Suvarna Kaliyuga. See also: Creation myth; Ichthys; Relationship between religion and science. References / Further reading: Evolution theory - C. Mackenzie Brown, Hindu Perspectives on Evolution: Darwin, Dharma, and Design (Routledge Hindu Studies Series), Routledge, two thousand twelve. - C. Mackenzie Brown (ed.), Asian Religious Responses to Darwinism: Evolutionary Theories in Middle Eastern, South Asian, and East Asian Cultural Contexts, Springer Nature, two thousand twenty. Creationism - Cavanaugh, Michael A. (nineteen eighty-three). A Sociological Account of Scientific Creationism: Science, True Science, Pseudoscience. Unpublished doctoral dissertation, University of Pittsburgh. - Eve, Harold (nineteen ninety). Creationist Movement in Modern America. Twayne Publishers. - "Vedic creationism" - Cremo, Michael A., and Richard L. Thompson, Forbidden Archeology: The Full Unabridged Edition. Torchlight Publishing; second rev. ed., January nineteen ninety-eight. - Cremo, Michael A., Forbidden Archeology's Impact: How a Controversial New Book Shocked the Scientific Community and Became an Underground Classic. Torchlight Publishing, January nineteen ninety-eight. - Cremo, Michael A., The Hidden History of the Human Race (The Condensed Edition of Forbidden Archeology). Cremo. Torchlight Publishing, May fifteen, nineteen ninety-nine. ISBN zero eight nine two one three three two five two. Hindu nationalism. Nanda, Meera. Prophets Facing Backward: Postmodern Critiques of Science and the Making of Hindu Nationalism in India. Rutgers University Press, two thousand three. External links: Hinduism and Science — Editorial in The Hindu, April twenty, two thousand four. Dharma vs. Darwin? — Swami B.V. Tripurari. Beliefnet article describing Hindu perspectives on evolution. The Perils of Vedic "Science" — Meera Nanda. Beliefnet article on Hindu science and evolutionary theories.
long_en_378	poet_en	830	en	A very special day today: Aunt Felip’s 105th birthday. "Happy birthday, Aunt Felip," Adna said. It was just a simple family celebration with the children — Lex, Azia, and Al, especially Al, who had cared for Aunt Felip when she was bedridden for almost three years. There was a birthday cake. People often ask, "What is the point of celebrating birthdays? What makes them important when the celebrant only grows older each year? Why spend money to invite others to a lavish dinner?" Azia heard this question and asked her mother. Adna explained that a birthday celebration is important not only to the person being celebrated but also to those who raised and loved them. It is a way to give thanks for another joyful year and to honor the person's place in the hearts of their loved ones. Few families make this a priority; in some, people are individualistic, rivalry prevails, and parents are so busy with careers and friends that they miss the important, joyful celebrations of each family member. Burning love also sparks and continues to connect us with those who have long since departed in this life. It becomes an unquenchable flame because the bond is so strong in the hearts and minds of those who love them. Joyous memories are never-ending; they endure and are passed down to future generations because this practice has long been observed. This connectedness of heart and mind, called "burning love," is the second secret of life. The first is gratitude for the one who cared for you when you were an infant, then a child, an adolescent, and eventually a parent yourself. Burning love is the fruit of gratitude. Adna remembered a young priest who worked with her at her first and largest free medical clinic. He said, "The child is the father of the man"—who you were as a child shapes who you become. Good people are often those raised by caregivers who gave their best and provided a beautiful upbringing. Life has ups and downs—good times and bad, success and failure, wealth and poverty—but none of that ultimately matters. What matters most is the time spent together, full of laughter despite the odds. It is like picking flowers: discard the thorns and keep the roses. Others say life is like a bed of roses with thorns: God has given us intellect and free will, so never lie down on the thorns. Make time to remove them and keep a bed of roses free of thorns. Adna told Azia that what they had experienced since Azia was young they would continue to enjoy. "Counting the years makes us happy," Adna said. Tomorrow was also the retired judge’s birthday, but Adna could not cook for it; she was too tired after a week of preparing the feast in the chapel near their house. Climbing the mountain after a very long journey had left her unwell and unable to prepare elaborate food. She would, however, still visit the judge to greet him. The retired judge liked Azia and hoped she would become a good judge. He regarded her as logical and kind. Azia, too, was very inquisitive and always willing to learn, just like Adna. When the head priest came to their garden, Adna asked, "Monsignor, what is the difference between a canon lawyer and a canon expert? I know two priests who are canon lawyers, but recently three people have been named canon experts, including you." The head priest explained that a canon expert is someone with deep, specialized knowledge of canon law, much like a medical specialist, while canon lawyers are those who know and apply the law. Azia, whenever she had questions, would ask for answers. Al the Second has a personality similar to Azia's, but he never stops asking questions. He always has many things on his mind. Adna sometimes says, "We'll discuss it next time." Lex, who is very studious, learns about new things through her own research; however, when a matter involves the law she asks her mother, who had been close to three lawyers—Uncle Rom and his two sons—during her university years. Adna's three children share one thing in common: they love her jelly-and-ripe-mango salad. When the others ate two small cups, Al had already eaten more than five. They also enjoy mango float, made with crushed graham crackers, ripe mangoes, and cream. Al the Second, the only son, loves to cook. He wanted to become a chef, but since it's not a standard bachelor's program, Adna told him he could learn through experience. "I just watched Aunt Felip cook beef steak and sweet-and-sour fish, and I learned to reproduce her flavors," she told him. Al became passionate about cooking and often experimented; at first he didn't get the taste right, but by the next attempt it was much better. He also enjoys gardening.	A very special day today: Aunt Felip’s one hundred five th birthday. "Happy birthday, Aunt Felip," Adna said. It was just a simple family celebration with the children — Lex, Azia, and Al, especially Al, who had cared for Aunt Felip when she was bedridden for almost three years. There was a birthday cake. People often ask, "What is the point of celebrating birthdays? What makes them important when the celebrant only grows older each year? Why spend money to invite others to a lavish dinner?" Azia heard this question and asked her mother. Adna explained that a birthday celebration is important not only to the person being celebrated but also to those who raised and loved them. It is a way to give thanks for another joyful year and to honor the person's place in the hearts of their loved ones. Few families make this a priority; in some, people are individualistic, rivalry prevails, and parents are so busy with careers and friends that they miss the important, joyful celebrations of each family member. Burning love also sparks and continues to connect us with those who have long since departed in this life. It becomes an unquenchable flame because the bond is so strong in the hearts and minds of those who love them. Joyous memories are never-ending; they endure and are passed down to future generations because this practice has long been observed. This connectedness of heart and mind, called "burning love," is the second secret of life. The first is gratitude for the one who cared for you when you were an infant, then a child, an adolescent, and eventually a parent yourself. Burning love is the fruit of gratitude. Adna remembered a young priest who worked with her at her first and largest free medical clinic. He said, "The child is the father of the man"—who you were as a child shapes who you become. Good people are often those raised by caregivers who gave their best and provided a beautiful upbringing. Life has ups and downs—good times and bad, success and failure, wealth and poverty—but none of that ultimately matters. What matters most is the time spent together, full of laughter despite the odds. It is like picking flowers: discard the thorns and keep the roses. Others say life is like a bed of roses with thorns: God has given us intellect and free will, so never lie down on the thorns. Make time to remove them and keep a bed of roses free of thorns. Adna told Azia that what they had experienced since Azia was young they would continue to enjoy. "Counting the years makes us happy," Adna said. Tomorrow was also the retired judge’s birthday, but Adna could not cook for it; she was too tired after a week of preparing the feast in the chapel near their house. Climbing the mountain after a very long journey had left her unwell and unable to prepare elaborate food. She would, however, still visit the judge to greet him. The retired judge liked Azia and hoped she would become a good judge. He regarded her as logical and kind. Azia, too, was very inquisitive and always willing to learn, just like Adna. When the head priest came to their garden, Adna asked, "Monsignor, what is the difference between a canon lawyer and a canon expert? I know two priests who are canon lawyers, but recently three people have been named canon experts, including you." The head priest explained that a canon expert is someone with deep, specialized knowledge of canon law, much like a medical specialist, while canon lawyers are those who know and apply the law. Azia, whenever she had questions, would ask for answers. Al the Second has a personality similar to Azia's, but he never stops asking questions. He always has many things on his mind. Adna sometimes says, "We'll discuss it next time." Lex, who is very studious, learns about new things through her own research; however, when a matter involves the law she asks her mother, who had been close to three lawyers—Uncle Rom and his two sons—during her university years. Adna's three children share one thing in common: they love her jelly-and-ripe-mango salad. When the others ate two small cups, Al had already eaten more than five. They also enjoy mango float, made with crushed graham crackers, ripe mangoes, and cream. Al the Second, the only son, loves to cook. He wanted to become a chef, but since it's not a standard bachelor's program, Adna told him he could learn through experience. "I just watched Aunt Felip cook beef steak and sweet-and-sour fish, and I learned to reproduce her flavors," she told him. Al became passionate about cooking and often experimented; at first he didn't get the taste right, but by the next attempt it was much better. He also enjoys gardening.
long_en_375	poet_en	1,033	en	My father is dead. The dashing young man is famous somewhere else. I don't know where. I don't keep up with him. He left me because we were young and confused, and because we knew — both of us knew — that what I wanted wasn't him and what he wanted wasn't me, and leaving me on an island was better than trapping me in a new palace labyrinth in some rich house in Athens. He was doing me a favor. Really, we both had just wanted away from where we were, and running away together had been the natural way to do it at the time. I moved on long ago. I wouldn't even call him my great love. I wouldn't even call him my pretty great love. Honestly, we never even made love. I've never been with a man. You are probably about to be my great love. Look at you, you. You're adorable. I mean it. You're as delicious as hot chocolate in winter. You're a goddess to me. I bought the box in my favorite curio shop, and the owner, whom I was supposed to call Hecate, told me it was a Living Box and all it did was sit there, waiting. It had a fleshy, squishy look to it — all pink and a little fuzzy like a square peach. It felt like a sausage patty with a heartbeat in my palm. I felt its tiny heartbeat. I watched the spidery blue veins beneath its skin pumping blood. I filled this lonely box with small things. Here is a large, brass button I stole from a former lover's jacket, imprinted with a cheerful anchor that always made me melancholy. Minnie lost it in my car, and I had lied to her for weeks before the end came, saying I had looked for it beneath the passenger seat. I never looked until I was looking for a piece of her because she wouldn't return my calls. Did I ever tell you about Minerva? I will. I promise. And soon. There was a fossil seashell collected from the side of the Anthropology building where I had spent days tugging loose my prize from the limestone wall beside my desk. I've kept that with me for years. I have no idea where my diploma went, but I kept the seashell. I haven't been to the ocean since I was a child. We should go, you and I. We should go before we forget what we remember. Finally, a snow globe from childhood assured me that Rock City was awash in holiday cheer and small bits of flowing white plastic. No matter where I put my living box, when the lid was closed my box looked lonely. I gave up trying to put my box somewhere happy. I placed it on a high shelf between a ship-in-a-bottle that my father had given me and a picture of myself from when I was young. This, too, just made me sad. My father has been dead for years, and the smiling girl's face on my canvas has new marks. I will never have a father again. I will never be so beautiful again. Minnie was never superstitious. She was my last serious girlfriend. She was a mechanical engineer. She folded her clothes and ironed them, and even used a little deodorant-stick-like thingy to take out tea stains in white blouses. I couldn't imagine a life more complex than pulling the clothes straight from the dryer—hot, hot, hot—and throwing them on slightly damp, then running to work at my computer in the bedroom. I never knew what to do about stains. Sometimes, in the night, I resorted to emergency club soda and it either worked or it didn't. I miss her. Of course I do. I miss all my lost lovers—even that boy. I'll miss you when you leave me for a younger girl with bright eyes. I'll be so jealous I'll dream of pouring acid on her face in her sleep, or poisoning her with black magic from Jennifer's Curio Shop. Magical omens and mystic portents: the thunder was terrible that day, but it hadn't rained at all. The boiling clouds tumbled all over each other, mustering up the courage to fight the ground. Styrofoam packing and dead leaves and papers danced in a windy bacchanal. Minnie and I were riding bicycles on a Sunday morning to a café, trying to beat the weather because we thought it would rain any second. When we got to the café, we sat in the window. We sipped lattes and read the gigantic newspaper that spilled all over our little table like a cat on a toy piano. We took off immediately. We raced up a hill, and I won, and she shouted at me to turn the corner on 14th. I asked her which direction, but she couldn't hear me shouting, and I couldn't hear her at all anymore with all that wind. I turned left and rode as far ahead as I could. I had guessed correctly. I rode as fast as I could over bumps and rough blacktop. Dense city buildings ended at a railroad bridge, with brown grass taller than trees growing where a sidewalk should have been. The grasses bowed like supplicants to the wind. Minnie called from behind me. I turned my head long enough to see that she had stopped at one of the little businesses there. She was waving at me and shouting. I skidded to a halt and turned my tires back toward her. Minerva's gift to me was a palm reading. I would have preferred Tarot cards, because they look so beautiful and you never know which one will flip over to look back at you. Palms never change. They get older and older, but the lifeline never shrinks like you think it should. See, Minnie was the one who took me there. She never thought it would be dangerous because she wasn't reckless like me.	My father is dead. The dashing young man is famous somewhere else. I don't know where. I don't keep up with him. He left me because we were young and confused, and because we knew — both of us knew — that what I wanted wasn't him and what he wanted wasn't me, and leaving me on an island was better than trapping me in a new palace labyrinth in some rich house in Athens. He was doing me a favor. Really, we both had just wanted away from where we were, and running away together had been the natural way to do it at the time. I moved on long ago. I wouldn't even call him my great love. I wouldn't even call him my pretty great love. Honestly, we never even made love. I've never been with a man. You are probably about to be my great love. Look at you, you. You're adorable. I mean it. You're as delicious as hot chocolate in winter. You're a goddess to me. I bought the box in my favorite curio shop, and the owner, whom I was supposed to call Hecate, told me it was a Living Box and all it did was sit there, waiting. It had a fleshy, squishy look to it — all pink and a little fuzzy like a square peach. It felt like a sausage patty with a heartbeat in my palm. I felt its tiny heartbeat. I watched the spidery blue veins beneath its skin pumping blood. I filled this lonely box with small things. Here is a large, brass button I stole from a former lover's jacket, imprinted with a cheerful anchor that always made me melancholy. Minnie lost it in my car, and I had lied to her for weeks before the end came, saying I had looked for it beneath the passenger seat. I never looked until I was looking for a piece of her because she wouldn't return my calls. Did I ever tell you about Minerva? I will. I promise. And soon. There was a fossil seashell collected from the side of the Anthropology building where I had spent days tugging loose my prize from the limestone wall beside my desk. I've kept that with me for years. I have no idea where my diploma went, but I kept the seashell. I haven't been to the ocean since I was a child. We should go, you and I. We should go before we forget what we remember. Finally, a snow globe from childhood assured me that Rock City was awash in holiday cheer and small bits of flowing white plastic. No matter where I put my living box, when the lid was closed my box looked lonely. I gave up trying to put my box somewhere happy. I placed it on a high shelf between a ship-in-a-bottle that my father had given me and a picture of myself from when I was young. This, too, just made me sad. My father has been dead for years, and the smiling girl's face on my canvas has new marks. I will never have a father again. I will never be so beautiful again. Minnie was never superstitious. She was my last serious girlfriend. She was a mechanical engineer. She folded her clothes and ironed them, and even used a little deodorant-stick-like thingy to take out tea stains in white blouses. I couldn't imagine a life more complex than pulling the clothes straight from the dryer—hot, hot, hot—and throwing them on slightly damp, then running to work at my computer in the bedroom. I never knew what to do about stains. Sometimes, in the night, I resorted to emergency club soda and it either worked or it didn't. I miss her. Of course I do. I miss all my lost lovers—even that boy. I'll miss you when you leave me for a younger girl with bright eyes. I'll be so jealous I'll dream of pouring acid on her face in her sleep, or poisoning her with black magic from Jennifer's Curio Shop. Magical omens and mystic portents: the thunder was terrible that day, but it hadn't rained at all. The boiling clouds tumbled all over each other, mustering up the courage to fight the ground. Styrofoam packing and dead leaves and papers danced in a windy bacchanal. Minnie and I were riding bicycles on a Sunday morning to a café, trying to beat the weather because we thought it would rain any second. When we got to the café, we sat in the window. We sipped lattes and read the gigantic newspaper that spilled all over our little table like a cat on a toy piano. We took off immediately. We raced up a hill, and I won, and she shouted at me to turn the corner on fourteenth. I asked her which direction, but she couldn't hear me shouting, and I couldn't hear her at all anymore with all that wind. I turned left and rode as far ahead as I could. I had guessed correctly. I rode as fast as I could over bumps and rough blacktop. Dense city buildings ended at a railroad bridge, with brown grass taller than trees growing where a sidewalk should have been. The grasses bowed like supplicants to the wind. Minnie called from behind me. I turned my head long enough to see that she had stopped at one of the little businesses there. She was waving at me and shouting. I skidded to a halt and turned my tires back toward her. Minerva's gift to me was a palm reading. I would have preferred Tarot cards, because they look so beautiful and you never know which one will flip over to look back at you. Palms never change. They get older and older, but the lifeline never shrinks like you think it should. See, Minnie was the one who took me there. She never thought it would be dangerous because she wasn't reckless like me.
long_en_364	poet_en	986	en	Tonight I was trying to clean up some email folders and accidentally deleted a bunch of folders I needed. I'm so upset with myself. I was literally saying, "Step away from the computer, Jill, and let things be." Ugh. I should have stuck to cleaning the house! Ping asked me first thing this morning what our plans were. We've been going out daily, so she assumed we had somewhere to go today. I told her we were staying home, and she was a little bummed. Just as well, since it poured for almost the entire day. Rain, drizzle, and mist are okay, but this was pouring — who wants to be out in that? I made a ton of phone calls and got a lot of work done at my desk. I'm really pleased because Ping and I cleaned out my armoire; it needed major tidying. Ping was thrilled to help, and we straightened all of her hair bows — that took an hour! Late this afternoon I asked Ping if I could have my ottoman so I could sit comfortably and read for a little while; I figured I'd read for 15 minutes until Bill got home. She was unwilling to part with it. She has a bean bag for TV watching, but in the last few weeks she has taken my ottoman and brought it up to the TV. Rather than insist, I went to the family room, but Miss Sophie was sleeping comfortably on the couch. I get no respect. Needless to say, I begged Ping to sit with me so we could snuggle. She went for it. I do love snuggling with her. Bill came home shortly after, and the two of them went out shopping. I watched a show that was featuring some surgery I wanted to see, and then I sat down at the computer to do a few things — and that is when I started deleting. I am such a deleter! When will I learn? Well, that is all for now. I hope you have a splendid weekend. Mama out. Yesterday was supposed to be a very hot and humid day. Well, it ended up being pretty hot, but there wasn't any sun. That was disappointing because I took Ping to Donna's house for a swim. It's an hour there and an hour back. The least I could ask for was some sun. Ping didn't care, though. After Michael and Brad cooked some hamburgers and hot dogs, we ate and then hopped into the pool. Donna said it was too cold for her. I stayed in the pool for a good hour or so. I finally had to drag the prune-toed child out of the pool. Donna promised her some ice cream, so I was able to get her to dry off and change since she had something else that was exciting. We got home around 5:00 p.m. or so. We didn't do much for the rest of the night. I did do a lot of research on the computer for some upcoming things I would like to do. I finally crawled into bed around 2:00 a.m. Sleeping last night wasn't going to come easily. There was a fire call, so I had to listen to that entire scene unfold. Bill didn't have to go because he is on the ambulance, but we still have to listen to it. For some reason I was all twisted up in what I was wearing. Nothing was going right and I was really itchy from the awful bites. I'm not sure if it was from swimming, but my neck really hurt too. I finally fell asleep after 3:00 a.m. or so and woke up at the crack of dawn. I also woke when Bill got ready for work; Ping woke with him. She has been sleeping in lately, but not today, so sleeping this morning wasn't going to happen either. It was fine — we had plans, so I needed to be up anyway. I called Meri to confirm our plans, jumped in the shower, headed to her house, and we went to lunch. We then drove to Playtown Express off Route 495, but when we arrived they told us they were closing in 15 minutes — during the summer they close at 3:00 p.m. We had no idea and felt bad because the kids were so excited, so we jumped back into our cars, went home to West Boylston, and took the kids to our local playground. I took Ping so she could play with a new friend named Ben. I met Ben's mom, Susan, through my CASA training. I had never been to this playground, so I met Susan at a local store and followed her there. It was bright and sunny; we sat in the sun, then moved to a bench in the shade, but the sun found us after a while and we both ended up sunburned. Hopefully Ben is okay; Ping was fine — she's so lucky she tans. Neither of us remembered sunscreen. However, Susan brought plenty of juice boxes and sandwiches for the kids — clearly a thoughtful mom. I brought a drink for Ping, one for myself, and some Goldfish and fruit snacks. Ping didn’t touch the Goldfish or fruit snacks, but she loved the juice boxes and enjoyed a sandwich. Ben and Susan gave Ping an adorable pink necklace; she loved it. On the way home she kept saying it was her favorite necklace. Ping and I had a lot of fun. Next time we'll pick a shadier playground or bring sunscreen. Ping spent the rest of the afternoon watching videos while I caught up on business calls. At one point she looked at me and said, "Mama, why are you so red?"	Tonight I was trying to clean up some email folders and accidentally deleted a bunch of folders I needed. I'm so upset with myself. I was literally saying, "Step away from the computer, Jill, and let things be." Ugh. I should have stuck to cleaning the house! Ping asked me first thing this morning what our plans were. We've been going out daily, so she assumed we had somewhere to go today. I told her we were staying home, and she was a little bummed. Just as well, since it poured for almost the entire day. Rain, drizzle, and mist are okay, but this was pouring — who wants to be out in that? I made a ton of phone calls and got a lot of work done at my desk. I'm really pleased because Ping and I cleaned out my armoire; it needed major tidying. Ping was thrilled to help, and we straightened all of her hair bows — that took an hour! Late this afternoon I asked Ping if I could have my ottoman so I could sit comfortably and read for a little while; I figured I'd read for fifteen minutes until Bill got home. She was unwilling to part with it. She has a bean bag for TV watching, but in the last few weeks she has taken my ottoman and brought it up to the TV. Rather than insist, I went to the family room, but Miss Sophie was sleeping comfortably on the couch. I get no respect. Needless to say, I begged Ping to sit with me so we could snuggle. She went for it. I do love snuggling with her. Bill came home shortly after, and the two of them went out shopping. I watched a show that was featuring some surgery I wanted to see, and then I sat down at the computer to do a few things — and that is when I started deleting. I am such a deleter! When will I learn? Well, that is all for now. I hope you have a splendid weekend. Mama out. Yesterday was supposed to be a very hot and humid day. Well, it ended up being pretty hot, but there wasn't any sun. That was disappointing because I took Ping to Donna's house for a swim. It's an hour there and an hour back. The least I could ask for was some sun. Ping didn't care, though. After Michael and Brad cooked some hamburgers and hot dogs, we ate and then hopped into the pool. Donna said it was too cold for her. I stayed in the pool for a good hour or so. I finally had to drag the prune toed child out of the pool. Donna promised her some ice cream, so I was able to get her to dry off and change since she had something else that was exciting. We got home around five p m or so. We didn't do much for the rest of the night. I did do a lot of research on the computer for some upcoming things I would like to do. I finally crawled into bed around two a m. Sleeping last night wasn't going to come easily. There was a fire call, so I had to listen to that entire scene unfold. Bill didn't have to go because he is on the ambulance, but we still have to listen to it. For some reason I was all twisted up in what I was wearing. Nothing was going right and I was really itchy from the awful bites. I'm not sure if it was from swimming, but my neck really hurt too. I finally fell asleep after three colon zero zero a.m. or so and woke up at the crack of dawn. I also woke when Bill got ready for work; Ping woke with him. She has been sleeping in lately, but not today, so sleeping this morning wasn't going to happen either. It was fine — we had plans, so I needed to be up anyway. I called Meri to confirm our plans, jumped in the shower, headed to her house, and we went to lunch. We then drove to Playtown Express off Route four hundred ninety five, but when we arrived they told us they were closing in fifteen minutes — during the summer they close at three colon zero zero p.m. We had no idea and felt bad because the kids were so excited, so we jumped back into our cars, went home to West Boylston, and took the kids to our local playground. I took Ping so she could play with a new friend named Ben. I met Ben's mom, Susan, through my CASA training. I had never been to this playground, so I met Susan at a local store and followed her there. It was bright and sunny; we sat in the sun, then moved to a bench in the shade, but the sun found us after a while and we both ended up sunburned. Hopefully Ben is okay; Ping was fine — she's so lucky she tans. Neither of us remembered sunscreen. However, Susan brought plenty of juice boxes and sandwiches for the kids — clearly a thoughtful mom. I brought a drink for Ping, one for myself, and some Goldfish and fruit snacks. Ping didn’t touch the Goldfish or fruit snacks, but she loved the juice boxes and enjoyed a sandwich. Ben and Susan gave Ping an adorable pink necklace; she loved it. On the way home she kept saying it was her favorite necklace. Ping and I had a lot of fun. Next time we'll pick a shadier playground or bring sunscreen. Ping spent the rest of the afternoon watching videos while I caught up on business calls. At one point she looked at me and said, "Mama, why are you so red?".
long_en_339	poet_en	1,024	en	For as long as I can remember — sometimes 50-some years ago and sometimes only a minute ago — I have found amazement in the ability for gifts to create themselves. My children will roll their eyes at this, but they will also shake their heads: yes. Oh yes. Here she goes. Grab a Coke and a bag of chips and get comfortable... Something inside me believes as strongly as I believe in truth that everything around us and among us has some thread of life and spirit of giving. When these characteristics make themselves known, it feels like a gift to me — I mean not to ME, but to me. I can recall, many years ago, thinking it pretty brave that the grass grew back after being mowed time and time again, especially when the hardworking little green blades knew darn well they'd be cut back again anyway. If the sun shone on a day I really needed to be out in it, well, you may as well have put a bow on sunrise that day. Talk about gifts. I was thrilled that my children chose to be born, as if they could choose not to be. Well, maybe they could have; they are each pretty determined creatures. I'm just glad that they swam to the light. My children used to give me coupons for holidays like Mother's Day or (puke) birthday. I still have the Ovaltine coupon jar in the attic. I didn't use the coupons much, not nearly as much as I could have, but just having the thoughts and ideas from my own children—well, it was overwhelming. I mean, come on—who doesn't find it incredibly invaluable to have a 10-minute "playing with hair" while waiting? When my car starts, I am thankful. It might just be damn tired of being a multi-use vehicle and decide not to run anymore. Such a gift. Cooking—messy cheffing—when the food gets eaten, I am thankful for those who took the chance. Yep, even now. The fact that my counselor continues to support me and sees me weekly is unbelievable. I continue to try to prepare myself for the "You're fired, Melanie" words, but he reassures me that we will keep talking and we have lots of work to do. Don't I know it. The stories and experiences that have entered his space over these years have filled every nook, cranny, and corner with silken, interwoven webs from my past and present. Still, "See you next week" comes out at the end of our time and I am always surprised. Always. Isn't it a gift to find a safe place where you can completely bear your soul and all that lies enmeshed in it? Gift. Connecting with and forming a trusting and delightful friendship with a friend's daughter—gift. The gift of watching her become... is such a present. I enjoy working in the dirt. My back doesn't love it so much, so my effective working time is limited, but I find calm and hopefulness in gardening. When bulbs come up—when vegetables produce, flowers bloom, cuttings take root—how are those not huge, huge gifts? I am putting them in the ground to sleep. No one told them they had to get up. Call me quirky. I don't care. I appreciate the simple gifts of nature and friends, and that's about all I can take, too. For good reason, but old reasoning, I have a visceral reaction to being given gifts outside that simple range. I'm working on it (see above... weekly talk time with one who knows and is wise). When the work church good elves and good fairies leave me treats and presents, I am stunned. I've been here over five years now and still have the same reaction. Home church surrounds me with more gifting than I can fit into this post even if I changed to a 3-point font. When my boss reviews me and doesn't say, "It's been nice having you, but really—let's be real: you just aren't cut out for this job," I walk away in a daze. That's a big gift. Big. A couple of years ago, I rode my neat, old, worn bike to work. Before the day was out, the bike was being driven by a regular visitor to work church who needed transportation to find work. Even though I knew this visitor well enough to know he would probably sell it and purchase things I didn't want to know about, I thought giving him a dose of trust might be a good thing, so I sent him off on my bike, where he was spotted half an hour later by a work friend who saw him driving like a bat... We didn't read any headlines about robbers on bikes the next day, so we let it go. I started looking for a replacement but I wasn't really sure what to get. Classifieds read: "Middle-aged body seeking comfortable two-wheeler..." What to get? A friend, who is framily now, called and took me looking. He and his best-in-the-world girlfriend and I had fun shopping. Actually, I think they had fun shopping while I was having heart attacks at how much new bikes cost if you get them with a seat and wheels, which is really more useful than a lone frame. The seats were mostly built to hold one cheek of the average adult's bottom, too. What's that about? The day ended and I had such a nice time. The next day, I came home from church and on my back stoop sat a bike. A bike with wheels and handlebars and a seat that might just hold even my aged a... I couldn't believe it. How serendipitous! The very next day after we'd been looking and shopping! I called my friends and yelled into the phone, "Someone bought me a bike? Did you buy me a bike?"	For as long as I can remember — sometimes fifty-some years ago and sometimes only a minute ago — I have found amazement in the ability for gifts to create themselves. My children will roll their eyes at this, but they will also shake their heads: yes. Oh yes. Here she goes. Grab a Coke and a bag of chips and get comfortable... Something inside me believes as strongly as I believe in truth that everything around us and among us has some thread of life and spirit of giving. When these characteristics make themselves known, it feels like a gift to me — I mean not to ME, but to me. I can recall, many years ago, thinking it pretty brave that the grass grew back after being mowed time and time again, especially when the hardworking little green blades knew darn well they'd be cut back again anyway. If the sun shone on a day I really needed to be out in it, well, you may as well have put a bow on sunrise that day. Talk about gifts. I was thrilled that my children chose to be born, as if they could choose not to be. Well, maybe they could have; they are each pretty determined creatures. I'm just glad that they swam to the light. My children used to give me coupons for holidays like Mother's Day or (puke) birthday. I still have the Ovaltine coupon jar in the attic. I didn't use the coupons much, not nearly as much as I could have, but just having the thoughts and ideas from my own children—well, it was overwhelming. I mean, come on—who doesn't find it incredibly invaluable to have a ten-minute "playing with hair" while waiting? When my car starts, I am thankful. It might just be damn tired of being a multi-use vehicle and decide not to run anymore. Such a gift. Cooking—messy cheffing—when the food gets eaten, I am thankful for those who took the chance. Yep, even now. The fact that my counselor continues to support me and sees me weekly is unbelievable. I continue to try to prepare myself for the "You're fired, Melanie" words, but he reassures me that we will keep talking and we have lots of work to do. Don't I know it. The stories and experiences that have entered his space over these years have filled every nook, cranny, and corner with silken, interwoven webs from my past and present. Still, "See you next week" comes out at the end of our time and I am always surprised. Always. Isn't it a gift to find a safe place where you can completely bear your soul and all that lies enmeshed in it? Gift. Connecting with and forming a trusting and delightful friendship with a friend's daughter—gift. The gift of watching her become... is such a present. I enjoy working in the dirt. My back doesn't love it so much, so my effective working time is limited, but I find calm and hopefulness in gardening. When bulbs come up—when vegetables produce, flowers bloom, cuttings take root—how are those not huge, huge gifts? I am putting them in the ground to sleep. No one told them they had to get up. Call me quirky. I don't care. I appreciate the simple gifts of nature and friends, and that's about all I can take, too. For good reason, but old reasoning, I have a visceral reaction to being given gifts outside that simple range. I'm working on it (see above... weekly talk time with one who knows and is wise). When the work church good elves and good fairies leave me treats and presents, I am stunned. I've been here over five years now and still have the same reaction. Home church surrounds me with more gifting than I can fit into this post even if I changed to a three-point font. When my boss reviews me and doesn't say, "It's been nice having you, but really—let's be real: you just aren't cut out for this job," I walk away in a daze. That's a big gift. Big. A couple of years ago, I rode my neat, old, worn bike to work. Before the day was out, the bike was being driven by a regular visitor to work church who needed transportation to find work. Even though I knew this visitor well enough to know he would probably sell it and purchase things I didn't want to know about, I thought giving him a dose of trust might be a good thing, so I sent him off on my bike, where he was spotted half an hour later by a work friend who saw him driving like a bat... We didn't read any headlines about robbers on bikes the next day, so we let it go. I started looking for a replacement but I wasn't really sure what to get. Classifieds read: "Middle-aged body seeking comfortable two-wheeler..." What to get? A friend, who is framily now, called and took me looking. He and his best-in-the-world girlfriend and I had fun shopping. Actually, I think they had fun shopping while I was having heart attacks at how much new bikes cost if you get them with a seat and wheels, which is really more useful than a lone frame. The seats were mostly built to hold one cheek of the average adult's bottom, too. What's that about? The day ended and I had such a nice time. The next day, I came home from church and on my back stoop sat a bike. A bike with wheels and handlebars and a seat that might just hold even my aged a... I couldn't believe it. How serendipitous! The very next day after we'd been looking and shopping! I called my friends and yelled into the phone, "Someone bought me a bike? Did you buy me a bike?".
long_en_355	poet_en	987	en	A couple of years ago, Shane and I were working at Fruth one evening before Halloween. I like to demo costumes, and I decided to do a little Halloween makeup. I did my face to look like I was in a fight or car wreck or something. It was very realistic, and my customers were freaking out. As the evening evolved, so did our story. We told people that we had an argument and Shane beat me up. One of my regular customers came in, and when he saw me he asked, "Who did that to you?!" I was in full joking mode and told him that Shane did it. Then I saw the storm clouds gathering on the guy's face. He was furious. He said, "He's going to take care of him!" He added, "He doesn't know what I do for a living! I WILL take care of this." When I saw how mad he was, I started to deny the story, and I could see he didn't believe me. He thought I was a typical female covering for her man or something. The more I denied it, the more he got worked up. Finally I had the presence of mind to smear the makeup to prove that I was not hurt. I did thank the man for his concern, and it was nice that one of my customers actually cared enough to come to my defense. I guess I should become a Hollywood makeup artist or something. Shane had no idea how close he came to getting his behind kicked that night. I did not tell anybody else that story either. I have had a lot of good Halloweens. My last really good one was my dad's last Halloween. I picked him up at the nursing home and took him to dinner at the Waffle House. He loved it because he got to eat biscuits and gravy and other food that he wasn't supposed to have. Then I took him to the house. He sat on the porch all evening giving candy out to trick-or-treaters. They would get two or three pieces and he would get one. He was a diabetic, but he knew that I wouldn't say anything because of the holiday. He loved giving candy to the kids. He had his Yorkie, Katie, with him. She was dressed up in her devil costume, which said "Killer" on the shirt. Katie and Nikita (our other dog) would absolutely go nuts every time a trick-or-treater came on the porch. Daddy loved hollering at her not to bite anybody and keeping her under control. He knew that she was prone to sneaking up behind people and biting them on the back of their ankles. We laughed a lot that night, both at the dogs and the kids. I took him back to the nursing home at the end of the night. He had just as good a time as I did. He died four days later. When I was at Ohio State, our dorm was having a Halloween party. It was a typical college party full of fun and drunken debauchery and such. We were all in our costumes, cruising from room to room and floor to floor just having a good time in general. I was a ghost. I took the sheet off my bed and cut a hole in it for my head, then I took my pillowcase, cut out two eyeholes, and circled them with a black marker. If I had been smarter I would have cut out a mouth hole to make drinking easier! A friend and I went into a suite belonging to some other friends. They weren't there, so we decided to visit a room we didn't know—a party, you know. We walked in and it was full of Black men. We tried to laugh and talk with them, but they weren't friendly; it started to hurt my feelings, even though I was drunk. Finally one of them said, "What in the hell are you supposed to be, anyway?!" I replied in my West Virginia accent that I was a ghost. They all looked at each other as if to say, "Yeah, right." It was obviously not a friendly atmosphere, so my friend and I left. Once we were out, my friend pointed out that they had thought I was dressed as a Ku Klux Klan member. My Southern accent only made things worse. That possibility hadn't even occurred to me, but after looking in the mirror I could see why they would think that. I felt terrible and wanted to go back and tell them I was definitely not a Klan member, but my friend convinced me it would be best not to make a big deal of it. It just goes to show that appearances can be deceiving and you should never assume anything about someone. When I was in junior high, we were in the Smoky Mountains for a family vacation. We were at a go-kart track in Pigeon Forge, a tourist town at the base of the hills. My brother, Billy, my sister, Rhonda, and I were having a ball, driving around the go-kart track at full speed like we did when we were kids. We didn't know what slow meant on anything! I had used up my time and was sitting in the van with my mom and dad, watching Billy and Rhonda go around the track when my sister wrecked. I remember seeing the whole thing and sitting frozen, even though I wanted to help her. One second she was flying around the track and the next she was spinning in circles (not rolling, thank God). She was lying on her back in the go-kart. When it stopped spinning, she did not get up.	A couple of years ago, Shane and I were working at Fruth one evening before Halloween. I like to demo costumes, and I decided to do a little Halloween makeup. I did my face to look like I was in a fight or car wreck or something. It was very realistic, and my customers were freaking out. As the evening evolved, so did our story. We told people that we had an argument and Shane beat me up. One of my regular customers came in, and when he saw me he asked, "Who did that to you?!" I was in full joking mode and told him that Shane did it. Then I saw the storm clouds gathering on the guy's face. He was furious. He said, "He's going to take care of him!" He added, "He doesn't know what I do for a living! I WILL take care of this." When I saw how mad he was, I started to deny the story, and I could see he didn't believe me. He thought I was a typical female covering for her man or something. The more I denied it, the more he got worked up. Finally I had the presence of mind to smear the makeup to prove that I was not hurt. I did thank the man for his concern, and it was nice that one of my customers actually cared enough to come to my defense. I guess I should become a Hollywood makeup artist or something. Shane had no idea how close he came to getting his behind kicked that night. I did not tell anybody else that story either. I have had a lot of good Halloweens. My last really good one was my dad's last Halloween. I picked him up at the nursing home and took him to dinner at the Waffle House. He loved it because he got to eat biscuits and gravy and other food that he wasn't supposed to have. Then I took him to the house. He sat on the porch all evening giving candy out to trick-or-treaters. They would get two or three pieces and he would get one. He was a diabetic, but he knew that I wouldn't say anything because of the holiday. He loved giving candy to the kids. He had his Yorkie, Katie, with him. She was dressed up in her devil costume, which said "Killer" on the shirt. Katie and Nikita (our other dog) would absolutely go nuts every time a trick-or-treater came on the porch. Daddy loved hollering at her not to bite anybody and keeping her under control. He knew that she was prone to sneaking up behind people and biting them on the back of their ankles. We laughed a lot that night, both at the dogs and the kids. I took him back to the nursing home at the end of the night. He had just as good a time as I did. He died four days later. When I was at Ohio State, our dorm was having a Halloween party. It was a typical college party full of fun and drunken debauchery and such. We were all in our costumes, cruising from room to room and floor to floor just having a good time in general. I was a ghost. I took the sheet off my bed and cut a hole in it for my head, then I took my pillowcase, cut out two eyeholes, and circled them with a black marker. If I had been smarter I would have cut out a mouth hole to make drinking easier! A friend and I went into a suite belonging to some other friends. They weren't there, so we decided to visit a room we didn't know—a party, you know. We walked in and it was full of Black men. We tried to laugh and talk with them, but they weren't friendly; it started to hurt my feelings, even though I was drunk. Finally one of them said, "What in the hell are you supposed to be, anyway?!" I replied in my West Virginia accent that I was a ghost. They all looked at each other as if to say, "Yeah, right." It was obviously not a friendly atmosphere, so my friend and I left. Once we were out, my friend pointed out that they had thought I was dressed as a Ku Klux Klan member. My Southern accent only made things worse. That possibility hadn't even occurred to me, but after looking in the mirror I could see why they would think that. I felt terrible and wanted to go back and tell them I was definitely not a Klan member, but my friend convinced me it would be best not to make a big deal of it. It just goes to show that appearances can be deceiving and you should never assume anything about someone. When I was in junior high, we were in the Smoky Mountains for a family vacation. We were at a go-kart track in Pigeon Forge, a tourist town at the base of the hills. My brother, Billy, my sister, Rhonda, and I were having a ball, driving around the go-kart track at full speed like we did when we were kids. We didn't know what slow meant on anything! I had used up my time and was sitting in the van with my mom and dad, watching Billy and Rhonda go around the track when my sister wrecked. I remember seeing the whole thing and sitting frozen, even though I wanted to help her. One second she was flying around the track and the next she was spinning in circles (not rolling, thank God). She was lying on her back in the go-kart. When it stopped spinning, she did not get up.
long_en_362	poet_en	658	en	"I need the prince of Himekaizen." What kind of request was that? Ryu shook his head in amusement. "You really don't know what makes girls tick, Urufu. Right now you're the king of Himekaizen, and the senior girls probably think you're more interesting than I am anyway." He wasn't sure how he knew, but he did. Urufu's personality and the way he handled the festival were certain to attract more than a few third years. They were preparing to leave high school behind, and to a degree they acted more adult than they would as first- or second-year university students. At least, if what his father had said was true. That only meant he had to become a more grown-up prince. It was a strange thought, and one he wouldn't have had just half a year earlier. He wondered if he was growing up as well. The way Urufu and Kuri had affected his life scared him, but he admitted they had. Easy for them — they had their wingmen, after all. Then he accepted why that thought lacked dignity: they had earned their wingmen. Sometimes he wished he could be more like Yukio, or Kyoko for that matter. Ryu turned left after the vending machines and walked almost to the end of the corridor. From here on, it was mostly a matter of knocking on doors. Urufu wanted a dozen third-year girls to do two shopping rounds each. Ryu would get them; otherwise he wasn't worthy of being called the prince of Himekaizen. "Princess of Scandinavia" had once been her second name, but in this world she was the princess of Himekaizen; maybe "empress" was a better description. She was neither petite nor cute in the preferred way of a Japanese high school princess. Yeah—empress it is, she thought, when reminded of the task ahead. What Ulf asked was far beyond what could decently be asked of a high school girl, but that wasn't what he had done, was it? For once he didn't look at her with love; she understood how he addressed the Billion Dollar Empress instead of his "little Ina." It hurt a little, and it made her immensely proud of him. As long as he understood he couldn't use the rest of them this way, she thought, but it was already too late. Sooner or later some of them would break, and they would resent him because they were unable to live up to his expectations. He didn't really care; his reputation would suffer because they'd think he was after personal glory. It had to be done, and it was all because of her. She shouldn't have started the Uniclo circus. She silently pleaded, "Ulf, please don't hate me for it." "Hangers six and seven further back and to the left," someone called. Christina watched two more metal frames move closer to the exit. "Try it on." Fifteen minutes of fame—more like fifteen seconds, but adding an ad-hoc fashion show on the football field should secure a lot of guests in one place when the school got crammed beyond capacity. Five hundred, maybe six hundred, she thought. Perhaps the best of the crazy ideas he'd come up with, and it still wouldn't be enough. She smiled to herself. She tried so hard even when she knew it wouldn't suffice. There were rumours about him. They called him the magician, and 'they' included teachers as well. "Okay, crew: team one, stay here and prepare the planned show. Team two, follow me!" If you want me to run two shows, I'll make them both better than the one we planned. Today would put her to the test. Without any professionals by her side, she would try to surpass Ulf's insane expectations. She owed him that much. She loved him that much.	"I need the prince of Himekaizen." What kind of request was that? Ryu shook his head in amusement. "You really don't know what makes girls tick, Urufu. Right now you're the king of Himekaizen, and the senior girls probably think you're more interesting than I am anyway." He wasn't sure how he knew, but he did. Urufu's personality and the way he handled the festival were certain to attract more than a few third years. They were preparing to leave high school behind, and to a degree they acted more adult than they would as first- or second-year university students. At least, if what his father had said was true. That only meant he had to become a more grown-up prince. It was a strange thought, and one he wouldn't have had just half a year earlier. He wondered if he was growing up as well. The way Urufu and Kuri had affected his life scared him, but he admitted they had. Easy for them — they had their wingmen, after all. Then he accepted why that thought lacked dignity: they had earned their wingmen. Sometimes he wished he could be more like Yukio, or Kyoko for that matter. Ryu turned left after the vending machines and walked almost to the end of the corridor. From here on, it was mostly a matter of knocking on doors. Urufu wanted a dozen third-year girls to do two shopping rounds each. Ryu would get them; otherwise he wasn't worthy of being called the prince of Himekaizen. "Princess of Scandinavia" had once been her second name, but in this world she was the princess of Himekaizen; maybe "empress" was a better description. She was neither petite nor cute in the preferred way of a Japanese high school princess. Yeah—empress it is, she thought, when reminded of the task ahead. What Ulf asked was far beyond what could decently be asked of a high school girl, but that wasn't what he had done, was it? For once he didn't look at her with love; she understood how he addressed the Billion Dollar Empress instead of his "little Ina." It hurt a little, and it made her immensely proud of him. As long as he understood he couldn't use the rest of them this way, she thought, but it was already too late. Sooner or later some of them would break, and they would resent him because they were unable to live up to his expectations. He didn't really care; his reputation would suffer because they'd think he was after personal glory. It had to be done, and it was all because of her. She shouldn't have started the Uniclo circus. She silently pleaded, "Ulf, please don't hate me for it." "Hangers six and seven further back and to the left," someone called. Christina watched two more metal frames move closer to the exit. "Try it on." Fifteen minutes of fame—more like fifteen seconds, but adding an ad-hoc fashion show on the football field should secure a lot of guests in one place when the school got crammed beyond capacity. Five hundred, maybe six hundred, she thought. Perhaps the best of the crazy ideas he'd come up with, and it still wouldn't be enough. She smiled to herself. She tried so hard even when she knew it wouldn't suffice. There were rumours about him. They called him the magician, and 'they' included teachers as well. "Okay, crew: team one, stay here and prepare the planned show. Team two, follow me!" If you want me to run two shows, I'll make them both better than the one we planned. Today would put her to the test. Without any professionals by her side, she would try to surpass Ulf's insane expectations. She owed him that much. She loved him that much.
long_en_338	poet_en	1,032	en	Thirteen Christmas Traditions My Family Has (or Had in the Past!) 1. Our whole family loves to sing, so we always go Christmas caroling with the church. Love it, love it. It's so nice to spread Christmas cheer to those who aren't able to get out and enjoy the season. 2. For as long as I can remember I've been a part of the Christmas play or cantata at my church. I even wrote a few of the plays. This is something most of my family takes part in as well. It's fun to see it all come together after all the practice. 3. On Christmas Eve we all go to my Grandma and Grandpa Cockman's house. These are my mom's parents. The family on that side is huge: Mom has 2 brothers and 4 sisters, plus all of their kids, which comes to 16 grandkids in all, plus the grandkids' children, which stands at 3 right now, plus all the spouses, boyfriends, and girlfriends. Just put it this way, there are a LOT of people there. But we always have a fun time catching up with each other. 4. Sometime on Christmas Eve my brother and I get together with our cousins Scott, Tracy, Will, and Chris to do our gift exchange and to watch "One More Sleep," better known to the rest of the world as "The Muppet Christmas Carol." 5. When we were younger and still lived at home, my sister and I shared a room. My brother had his own room because he was a boy (bum!), but on Christmas Eve Mom always made him sleep in our room because he was sneaky and would try to get a peek at the presents. When we'd wake up on Christmas morning we'd all sit there and talk about what we thought we'd gotten. The rule was we had to wait until at least five to wake our parents. As soon as the clock struck five we would all yell together, "Mama, Daddy, wake up. It's Christmas!!!" Mom always passed out the gifts because she was the only one who knew who was supposed to get what. See, she'd write names on the packages in tiny print where no one could see it. We spent the weeks before Christmas trying to find our names written on the packages. Sometimes even she couldn't find where she had written them. Heather Smith Thirteen Reasons Santa Might Have Picked Elves to Help Him Instead of Humans 1. He was obviously on a power trip. He wanted to stand head and shoulders above his entire workforce. 2. He felt sorry for the elves who made the shoes and wanted to offer them something more fun and less stinky to do. 3. He figured elves wouldn't eat as much, so there would be more food for him to eat. After all, you can't stay the big, jolly guy if you are hungry all the time. 4. He knew that the elves wouldn't expect health insurance or a retirement plan. 5. He figured no human would want the thankless job of cleaning up after the reindeer. 6. He knew that if he hired humans they'd be playing with the toys instead of building them (especially if he hired men!). 7. He figured he'd have a lifetime supply of cookies since the Keebler elves were the experts in cookie making. He didn't want anybody the same size as him in the North Pole because they might try to overthrow him. 9. He figured it would add more mystery to his existence and make more people question, "Does he exist or not?" 10. He decided he'd never get in trouble over affirmative action because he had hired a complete minority. 11. He had tons of extra doll clothes and needed to hire someone who could wear them. 12. He wanted to feel important since he was the only one tall enough to reach the top of the Christmas tree. 13. Too many of the humans he had tried to hire were hunters in disguise trying to kill the reindeer. Merry Christmas, everybody! Get the Thursday Thirteen code here! View more Thursday Thirteen participants. Posted by Thirteen Things. Tracy and I Did to Decorate Our House for Christmas 1. Do you remember a few weeks back where I told you about the giant Buddy the Elf cutout that was standing in my bay window? Well, he is now taped to my front door trimmed in pretty green garland. (This was Tracy's work.) 2. One of my dad's deer heads that was left hanging above the mantel now has a green garland and red Christmas bow, bow-tie thing going on. (Also Tracy's work.) 3. There are candy dishes full of candy. (Tracy again.) 4. There is a Christmas tree in front of the bay window with about 2,500 white lights on it. (This one was me.) 5. There are Christmas knick-knacks on the mantel below the deer. 1. (This was half me, half Tracy.) One of these knickknacks is a picture of our cousin Scott last year at Christmastime wearing my nephew Trace's "My First Christmas" Santa Claus hat. It's a really funny picture in a Christmas frame. 2. Garland is hung from the mantel with lights strung through it (my doing). 3. There are lots of snowman and Santa candles placed around the house (half Tracy, half me). 4. A glass nativity scene is set up on the bar with garland and lights surrounding it. 5. A Christmas village is on our entertainment center on the shelf above our TV (thanks to Tracy's mom for the loan of the village). 6. Wreaths are hung on several walls, making the place look festive. 7. There is another nativity scene on my dresser and another on a bookshelf. After all, Jesus is the reason for the season, so you can never have too many nativity scenes. 8. Lights are strung from the canopy on my bed. This is one of my favorite things.	Thirteen Christmas Traditions My Family Has (or Had in the Past!) one. Our whole family loves to sing, so we always go Christmas caroling with the church. Love it, love it. It's so nice to spread Christmas cheer to those who aren't able to get out and enjoy the season. two. For as long as I can remember I've been a part of the Christmas play or cantata at my church. I even wrote a few of the plays. This is something most of my family takes part in as well. It's fun to see it all come together after all the practice. three. On Christmas Eve we all go to my Grandma and Grandpa Cockman's house. These are my mom's parents. The family on that side is huge: Mom has two brothers and four sisters, plus all of their kids, which comes to sixteen grandkids in all, plus the grandkids' children, which stands at three right now, plus all the spouses, boyfriends, and girlfriends. Just put it this way, there are a LOT of people there. But we always have a fun time catching up with each other. four. Sometime on Christmas Eve my brother and I get together with our cousins Scott, Tracy, Will, and Chris to do our gift exchange and to watch "One More Sleep," better known to the rest of the world as "The Muppet Christmas Carol." five. When we were younger and still lived at home, my sister and I shared a room. My brother had his own room because he was a boy (bum!), but on Christmas Eve Mom always made him sleep in our room because he was sneaky and would try to get a peek at the presents. When we'd wake up on Christmas morning we'd all sit there and talk about what we thought we'd gotten. The rule was we had to wait until at least five to wake our parents. As soon as the clock struck five we would all yell together, "Mama, Daddy, wake up. It's Christmas!!!" Mom always passed out the gifts because she was the only one who knew who was supposed to get what. See, she'd write names on the packages in tiny print where no one could see it. We spent the weeks before Christmas trying to find our names written on the packages. Sometimes even she couldn't find where she had written them. Heather Smith Thirteen Reasons Santa Might Have Picked Elves to Help Him Instead of Humans one. He was obviously on a power trip. He wanted to stand head and shoulders above his entire workforce. two. He felt sorry for the elves who made the shoes and wanted to offer them something more fun and less stinky to do. three. He figured elves wouldn't eat as much, so there would be more food for him to eat. After all, you can't stay the big, jolly guy if you are hungry all the time. four. He knew that the elves wouldn't expect health insurance or a retirement plan. five. He figured no human would want the thankless job of cleaning up after the reindeer. six. He knew that if he hired humans they'd be playing with the toys instead of building them (especially if he hired men!). seven. He figured he'd have a lifetime supply of cookies since the Keebler elves were the experts in cookie making. He didn't want anybody the same size as him in the North Pole because they might try to overthrow him. nine. He figured it would add more mystery to his existence and make more people question, "Does he exist or not?" ten. He decided he'd never get in trouble over affirmative action because he had hired a complete minority. eleven. He had tons of extra doll clothes and needed to hire someone who could wear them. twelve. He wanted to feel important since he was the only one tall enough to reach the top of the Christmas tree. thirteen. Too many of the humans he had tried to hire were hunters in disguise trying to kill the reindeer. Merry Christmas, everybody! Get the Thursday Thirteen code here! View more Thursday Thirteen participants. Posted by Thirteen Things. Tracy and I Did to Decorate Our House for Christmas one. Do you remember a few weeks back where I told you about the giant Buddy the Elf cutout that was standing in my bay window? Well, he is now taped to my front door trimmed in pretty green garland. (This was Tracy's work.) two. One of my dad's deer heads that was left hanging above the mantel now has a green garland and red Christmas bow, bow-tie thing going on. (Also Tracy's work.) three. There are candy dishes full of candy. (Tracy again.) four. There is a Christmas tree in front of the bay window with about two thousand five hundred white lights on it. (This one was me.) five. There are Christmas knick-knacks on the mantel below the deer. one. (This was half me, half Tracy.) One of these knickknacks is a picture of our cousin Scott last year at Christmastime wearing my nephew Trace's "My First Christmas" Santa Claus hat. It's a really funny picture in a Christmas frame. two. Garland is hung from the mantel with lights strung through it (my doing). three. There are lots of snowman and Santa candles placed around the house (half Tracy, half me). four. A glass nativity scene is set up on the bar with garland and lights surrounding it. five. A Christmas village is on our entertainment center on the shelf above our TV (thanks to Tracy's mom for the loan of the village). six. Wreaths are hung on several walls, making the place look festive. seven. There is another nativity scene on my dresser and another on a bookshelf. After all, Jesus is the reason for the season, so you can never have too many nativity scenes. eight. Lights are strung from the canopy on my bed. This is one of my favorite things.
long_en_350	poet_en	953	en	"My name is Corwin Hale. It is a pleasure to meet all of you," Corwin says, bowing to be as polite as Nell. He even manages to come off as feminine as she does, despite being clearly a man. Fenrir worries for Corwin when he sees Marija licking her lips while looking at him. It's clear her strike zone includes men and women alike as long as they are polite, feminine, and either look like princesses or princes who want to be treated like princesses. Next is Rainbow Afro. As soon as he opens his mouth to introduce himself, Marija cuts in, "That should do it for introductions." "Ouch," King Cat says. "You really hate his accent, don't you?" "It isn't cute at all." "You know, for somebody who loves to hang skeletons all over the place, you sure do love cute things. Kind of contradictory, don't you think?" "How is that contradictory? Skeletons are cute. You should see the charm I have on my phone." "Come on, don't get started on this in front of everybody. You're supposed to be some sort of tough chieftess, not a goth squealing over skeletons." "You just reminded me. I got tickets for a concert but won't be able to make it. Want them?" "Are they some goth metal band like the last one with tons of screaming?" "Yes." "I guess I'll drop by and take them off you for my sis. I—" "Maybe some other time?" Rainbow Afro interrupts. King Cat and Marija clear their throats and lean back in their seats, trying to get back on topic. "Sorry about that," Marija says. "It's all right. I'm guessing you two are friends in real life?" Fenrir asks. "Coworkers," Marija replies. "And friends," King Cat adds. "But mainly coworkers." "You never want to just admit it, you stubborn woman." "You always want to be too personal, you insubordinate cat‑lover." Rainbow Afro coughs, ending their back-and-forth. King Cat apologizes this time. "Sorry about that." "Let's get to the point before we get distracted again," Marija says. "You believe an army is going to attack?" "I'm willing to bet it will, but right now all I can say for a fact is that an army exists," Fenrir explains. "And you agree with him?" Marija asks King Cat. "That I do," he answers. "Call it a gut feeling. Or something else. Either way, I can tell something big is going to happen—call it a cat's instincts." "Why don't you get some cat ears and a tail like the dog over there if you're so obsessed with cats?" "Because those look better on girls than guys." "You could always become a girl." "Not after what you put me through last time." Fenrir and Nell sit there, blinking at the conversation. Corwin is keeping to himself and tuning out any conversations that are clearly not meant for him to hear, and Tabitha is still looking at Marija’s muscles as if they are the most beautiful wonders she has ever seen. “The Shoebill is going to get jealous if you stare too hard,” Fenrir whispers to Tabitha. Tabitha elbows him in the side before returning her attention to Marija. “C’mon, y’all. I was busy practicing my new skills before this. Bicker like a pair o' lovers after the rest of us are gone,” Rainbow Afro says. Despite being the most immersion-breaking man in the room, he’s the one who seems to be taking the meeting the most seriously, even as he combs a rainbow-colored brush through his afro. “They are quite the characters, my hero,” Nell whispers. “Yeah, but they seem to get along pretty well. I like them,” Fenrir whispers back. “I should also mention that I will be expecting quite the pampering later.” “Thanks, and you’ll get pampered as much as you want.” “I am tempted to mimic Serra and ask if there is something else I can get as much as I want, but alas, I would feel guilty stealing her job from her.” “You know, just saying you want to say that is basically the same as saying it.” “Saying what, my hero? I have only said that I wish to be pampered.” “Sure you did. Anyway, if you want that later then I can give you as much of—” King Cat coughs this time. That's when Fenrir and Nell realize they have gone from whispering to speaking at a normal volume. They both blush and try to brush it off with light chuckling, but even Tabitha is staring at them with an unimpressed expression. "Anyway," Fenrir says, "I think we have two options. We can either wait until the army attacks on its own, or we can lure it to a place where we're prepared." "Is seeking the army out and destroying it at its location not an option?" Marija asks. "It is—unless we have a good way of destroying an army at the bottom of the ocean. Tabs here could probably make some depth charges if she really wanted to, but I have a feeling those wouldn't be enough. Though they might help thin the numbers. Tabs, could you do that?" "Already workin' on plans for some. The deer suggested that," Tabitha answers. "Leave it to Olly." "Is it safe to assume you want to lure the army to a specified location, then?" Marija asks. "Yeah. I think it'd be the best option to deal with it without putting any of the towns at risk." "Spike Port is the most defensible town in this region."	"My name is Corwin Hale. It is a pleasure to meet all of you," Corwin says, bowing to be as polite as Nell. He even manages to come off as feminine as she does, despite being clearly a man. Fenrir worries for Corwin when he sees Marija licking her lips while looking at him. It's clear her strike zone includes men and women alike as long as they are polite, feminine, and either look like princesses or princes who want to be treated like princesses. Next is Rainbow Afro. As soon as he opens his mouth to introduce himself, Marija cuts in, "That should do it for introductions." "Ouch," King Cat says. "You really hate his accent, don't you?" "It isn't cute at all." "You know, for somebody who loves to hang skeletons all over the place, you sure do love cute things. Kind of contradictory, don't you think?" "How is that contradictory? Skeletons are cute. You should see the charm I have on my phone." "Come on, don't get started on this in front of everybody. You're supposed to be some sort of tough chieftess, not a goth squealing over skeletons." "You just reminded me. I got tickets for a concert but won't be able to make it. Want them?" "Are they some goth metal band like the last one with tons of screaming?" "Yes." "I guess I'll drop by and take them off you for my sis. I—" "Maybe some other time?" Rainbow Afro interrupts. King Cat and Marija clear their throats and lean back in their seats, trying to get back on topic. "Sorry about that," Marija says. "It's all right. I'm guessing you two are friends in real life?" Fenrir asks. "Coworkers," Marija replies. "And friends," King Cat adds. "But mainly coworkers." "You never want to just admit it, you stubborn woman." "You always want to be too personal, you insubordinate cat‑lover." Rainbow Afro coughs, ending their back-and-forth. King Cat apologizes this time. "Sorry about that." "Let's get to the point before we get distracted again," Marija says. "You believe an army is going to attack?" "I'm willing to bet it will, but right now all I can say for a fact is that an army exists," Fenrir explains. "And you agree with him?" Marija asks King Cat. "That I do," he answers. "Call it a gut feeling. Or something else. Either way, I can tell something big is going to happen—call it a cat's instincts." "Why don't you get some cat ears and a tail like the dog over there if you're so obsessed with cats?" "Because those look better on girls than guys." "You could always become a girl." "Not after what you put me through last time." Fenrir and Nell sit there, blinking at the conversation. Corwin is keeping to himself and tuning out any conversations that are clearly not meant for him to hear, and Tabitha is still looking at Marija’s muscles as if they are the most beautiful wonders she has ever seen. “The Shoebill is going to get jealous if you stare too hard,” Fenrir whispers to Tabitha. Tabitha elbows him in the side before returning her attention to Marija. “C’mon, y’all. I was busy practicing my new skills before this. Bicker like a pair o' lovers after the rest of us are gone,” Rainbow Afro says. Despite being the most immersion-breaking man in the room, he’s the one who seems to be taking the meeting the most seriously, even as he combs a rainbow-colored brush through his afro. “They are quite the characters, my hero,” Nell whispers. “Yeah, but they seem to get along pretty well. I like them,” Fenrir whispers back. “I should also mention that I will be expecting quite the pampering later.” “Thanks, and you’ll get pampered as much as you want.” “I am tempted to mimic Serra and ask if there is something else I can get as much as I want, but alas, I would feel guilty stealing her job from her.” “You know, just saying you want to say that is basically the same as saying it.” “Saying what, my hero? I have only said that I wish to be pampered.” “Sure you did. Anyway, if you want that later then I can give you as much of—” King Cat coughs this time. That's when Fenrir and Nell realize they have gone from whispering to speaking at a normal volume. They both blush and try to brush it off with light chuckling, but even Tabitha is staring at them with an unimpressed expression. "Anyway," Fenrir says, "I think we have two options. We can either wait until the army attacks on its own, or we can lure it to a place where we're prepared." "Is seeking the army out and destroying it at its location not an option?" Marija asks. "It is—unless we have a good way of destroying an army at the bottom of the ocean. Tabs here could probably make some depth charges if she really wanted to, but I have a feeling those wouldn't be enough. Though they might help thin the numbers. Tabs, could you do that?" "Already workin' on plans for some. The deer suggested that," Tabitha answers. "Leave it to Olly." "Is it safe to assume you want to lure the army to a specified location, then?" Marija asks. "Yeah. I think it'd be the best option to deal with it without putting any of the towns at risk." "Spike Port is the most defensible town in this region.".
long_en_300	wiki_en	873	en	The behavior of quantum dots (QDs) in solution and their interaction with surfaces is important for biological and industrial applications such as optical displays, animal tagging, anti-counterfeiting dyes and paints, chemical sensing, and fluorescent labeling. Unmodified QDs are typically hydrophobic, which prevents their use in stable water-based colloids. Because their surface-area-to-volume ratio is much higher than that of larger particles, surface dangling bonds raise the free energy and can disrupt quantum confinement of excitons. Encapsulation in micelles with hydrophobic interiors or hydrophilic exteriors allows QDs to be solubilized and introduced into aqueous media, where they can be incorporated into hydrogel networks for applications such as medical imaging and thermal destruction of malignant tumors. Quantum dots are nanoscale semiconductor particles about 2-10 nm in diameter. They have electrical properties between those of bulk semiconductors and individual molecules, and optical properties that make them useful for fluorescence-based applications. Most QDs synthesized for medical imaging are CdSe/ZnS core-shell particles. CdSe quantum dots (QDs) have optical properties superior to those of organic dyes. Coating them with a ZnS shell has a twofold effect: it passivates dangling bonds that would otherwise cause particle aggregation and loss of optical resolution, and it preserves quantum confinement, thereby increasing particle fluorescence. Despite their potential as contrast agents for medical imaging, in vivo use is hindered by cadmium cytotoxicity. To address this, QDs can be encapsulated in biologically inert polymers to reduce toxicity and enable use in living tissue. Although cadmium-free QDs are commercially available, they are not yet suitable substitutes for organic contrast agents. Another challenge is the inherent hydrophobicity of CdSe(ZnS) nanoparticles, which limits their dispersibility in aqueous media such as blood or spinal fluid; hydrophilic polymers are often used to render the dots water-soluble. One notable encapsulation approach employs a fluoroalkyl-terminated polyethylene glycol (Rf-PEG) surfactant that spontaneously forms micellar structures above its critical micelle concentration (CMC); the CMC depends on the PEG chain length. This molecule consists of a hydrophilic PEG backbone with two hydrophobic terminal groups (CnF2n+1-CH2CH2O) attached via isophorone diurethane. Synthesis involves dehydrating a solution of 1,3-dimethyl-5-fluorouracil and PEG in the presence of heavy water (D2O) and combining them by sonication. Micellization: at the appropriate Krafft temperature and above the critical micelle concentration, these molecules form individual tear-drop loops in which the hydrophobic ends associate with one another, with neighboring molecules, and with similarly hydrophobic quantum dots (QDs). This produces a loaded micelle with a hydrophilic outer shell and a hydrophobic core. When encapsulating hydrophobic particles, it is important to ensure the particle size is compatible with the PEG backbone: the number of PEG repeat units (typically corresponding to PEG molecular weights of ~6 kDa or ~10 kDa) determines the maximum particle diameter that can be stably contained in the micelle core. To determine the average diameter, D, of the QDs, an empirical equation relating D to λ, the wavelength of the first absorption peak, is used. Role of the ZnS shell: during encapsulation the ZnS shell helps prevent agglomeration of bare CdSe particles by passivating surface bonds; however, secondary aggregation can still occur due to hydrophobic interactions. This can result in multiple particles within each micelle, which may negatively impact overall resolution. For this reason, multiple combinations of PEG chain length and particle diameter are necessary to achieve optimal imaging properties. Hydrogel network After initial encapsulation, the remaining molecules form connections between the individual micelles to create a network within the aqueous medium called a hydrogel, producing a diffuse and relatively constant concentration of the encapsulated particles within the gel. Hydrogel formation is also observed in superabsorbent polymers, or "slush powders," in which the polymer, often in powder form, absorbs water, becoming up to 99% liquid and swelling 30–60 times in size. Stokes–Einstein equation The diffusivity of spherical particles in a suspension is approximated by the Stokes–Einstein equation: D = k_B T / (6 π η r) where D is the diffusivity, T is the temperature, r is the particle radius, k_B is the Boltzmann constant, and η is the hydrogel viscosity. Typical Rf-PEG hydrogel diffusivities for 2 nm quantum dots are on the order of 10^−16 m^2/s, so suspensions of quantum dots tend to be very stable. Hydrogel viscosity can be determined by rheological techniques. Micelle rheology When encapsulating hydrophobic or potentially toxic materials, it is important that the encapsulant remain intact while inside the body. Studying the rheological properties of micelles permits selection of the polymer most appropriate for long-term biological applications. Rf-PEG exhibits superior rheological properties in vivo. Polymer length influences material behavior and encapsulant retention. The correct chain length prevents release of QDs and other toxic particles that could cause unintended cell necrosis. Polymer length is controlled by two factors: the weight of the PEG backbone (represented as #K, thousands of Daltons) and the length of the hydrophobic terminal groups (denoted C#, number of carbon atoms). Increasing PEG length increases solubility, but if the PEG chain is too long the micelle becomes unstable; stable hydrogels form only with PEG backbones of about 6–10 kDa. Increasing the length of the hydrophobic terminal groups decreases aqueous solubility.	The behavior of quantum dots (QDs) in solution and their interaction with surfaces is important for biological and industrial applications such as optical displays, animal tagging, anti counterfeiting dyes and paints, chemical sensing, and fluorescent labeling. Unmodified QDs are typically hydrophobic, which prevents their use in stable water based colloids. Because their surface area to volume ratio is much higher than that of larger particles, surface dangling bonds raise the free energy and can disrupt quantum confinement of excitons. Encapsulation in micelles with hydrophobic interiors or hydrophilic exteriors allows QDs to be solubilized and introduced into aqueous media, where they can be incorporated into hydrogel networks for applications such as medical imaging and thermal destruction of malignant tumors. Quantum dots are nanoscale semiconductor particles about two to ten nanometers in diameter. They have electrical properties between those of bulk semiconductors and individual molecules, and optical properties that make them useful for fluorescence based applications. Most QDs synthesized for medical imaging are CdSe slash ZnS core shell particles. CdSe quantum dots (QDs) have optical properties superior to those of organic dyes. Coating them with a ZnS shell has a twofold effect: it passivates dangling bonds that would otherwise cause particle aggregation and loss of optical resolution, and it preserves quantum confinement, thereby increasing particle fluorescence. Despite their potential as contrast agents for medical imaging, in vivo use is hindered by cadmium cytotoxicity. To address this, QDs can be encapsulated in biologically inert polymers to reduce toxicity and enable use in living tissue. Although cadmium-free QDs are commercially available, they are not yet suitable substitutes for organic contrast agents. Another challenge is the inherent hydrophobicity of CdSe(ZnS) nanoparticles, which limits their dispersibility in aqueous media such as blood or spinal fluid; hydrophilic polymers are often used to render the dots water-soluble. One notable encapsulation approach employs a fluoroalkyl-terminated polyethylene glycol (Rf-PEG) surfactant that spontaneously forms micellar structures above its critical micelle concentration (CMC); the CMC depends on the PEG chain length. This molecule consists of a hydrophilic PEG backbone with two hydrophobic terminal groups (C n F two n plus one-CH two CH two O) attached via isophorone diurethane. Synthesis involves dehydrating a solution of one,three-dimethyl-five-fluorouracil and PEG in the presence of heavy water (D two O) and combining them by sonication. Micellization: at the appropriate Krafft temperature and above the critical micelle concentration, these molecules form individual tear-drop loops in which the hydrophobic ends associate with one another, with neighboring molecules, and with similarly hydrophobic quantum dots (QDs). This produces a loaded micelle with a hydrophilic outer shell and a hydrophobic core. When encapsulating hydrophobic particles, it is important to ensure the particle size is compatible with the PEG backbone: the number of PEG repeat units (typically corresponding to PEG molecular weights of approximately six kDa or approximately ten kDa) determines the maximum particle diameter that can be stably contained in the micelle core. To determine the average diameter, D, of the QDs, an empirical equation relating D to lambda, the wavelength of the first absorption peak, is used. Role of the ZnS shell: during encapsulation the ZnS shell helps prevent agglomeration of bare CdSe particles by passivating surface bonds; however, secondary aggregation can still occur due to hydrophobic interactions. This can result in multiple particles within each micelle, which may negatively impact overall resolution. For this reason, multiple combinations of PEG chain length and particle diameter are necessary to achieve optimal imaging properties. Hydrogel network After initial encapsulation, the remaining molecules form connections between the individual micelles to create a network within the aqueous medium called a hydrogel, producing a diffuse and relatively constant concentration of the encapsulated particles within the gel. Hydrogel formation is also observed in superabsorbent polymers, or "slush powders," in which the polymer, often in powder form, absorbs water, becoming up to ninety nine percent liquid and swelling thirty to sixty times in size. Stokes–Einstein equation The diffusivity of spherical particles in a suspension is approximated by the Stokes–Einstein equation: D equals k B T over (six pi eta r) where D is the diffusivity, T is the temperature, r is the particle radius, k B is the Boltzmann constant, and eta is the hydrogel viscosity. Typical Rf-PEG hydrogel diffusivities for two nanometer quantum dots are on the order of ten to the minus sixteen meters squared per second, so suspensions of quantum dots tend to be very stable. Hydrogel viscosity can be determined by rheological techniques. Micelle rheology When encapsulating hydrophobic or potentially toxic materials, it is important that the encapsulant remain intact while inside the body. Studying the rheological properties of micelles permits selection of the polymer most appropriate for long-term biological applications. Rf-PEG exhibits superior rheological properties in vivo. Polymer length influences material behavior and encapsulant retention. The correct chain length prevents release of QDs and other toxic particles that could cause unintended cell necrosis. Polymer length is controlled by two factors: the weight of the PEG backbone (represented as number K, thousands of Daltons) and the length of the hydrophobic terminal groups (denoted C number, number of carbon atoms). Increasing PEG length increases solubility, but if the PEG chain is too long the micelle becomes unstable; stable hydrogels form only with PEG backbones of about six to ten kDa. Increasing the length of the hydrophobic terminal groups decreases aqueous solubility.
long_en_308	wiki_en	1,044	en	Tickford Vehicle Engineering (TVE) was responsible for numerous automotive projects and upgrades for Ford Australia between 1991 and 2002. In 1999 TVE set up Ford Tickford Experience (FTE) as a competitor to Holden Special Vehicles (HSV). In 2002 the operations transitioned to Ford Performance Vehicles (FPV), coinciding with Tickford's global operations being bought out by Prodrive. Overview In 1990, twelve years after the last Falcon Cobra rolled off the line, Ford Australia began a worldwide search for an engineering firm to establish a new performance arm in the same mould as Holden Special Vehicles (HSV). The aim was to create a line-up of in-house performance cars that had been missing from the Falcon range since the 1970s. Ford Australia's previous effort in this space had been the 1982 XE-series "European Sports Pack" Falcon. Throughout the late 1980s and early 1990s this performance void was partly filled by third-party tuning companies such as AVO, Special Vehicle Operations (SVO) and Dick Johnson Racing (DJR). Both AVO and DJR experimented with turbocharging to boost output from the Falcon's 4.0-litre six-cylinder, with DJR creating the Grand Prix Turbo based on the XE. DJR had intended his company to function as a performance arm similar to Peter Brock's Holden Dealer Team (HDT), but Ford was wary of such modifications because of warranty concerns, so sales of both AVO and DJR Falcons remained very limited. SVO had greater success creating a range of cars, beginning with the Falcon EA‑series SVO, a model that formed the template for the subsequent XR6. However, Ford still did not provide factory backing. Ultimately, Ford Australia joined forces with Tickford; in 1991 Tickford Vehicle Engineering (TVE) was established as a joint venture between Ford Australia and Tickford. TVE helped create the Falcon XR range, which emerged in 1992 with the EB series and the return of the Falcon GT. TVE was also responsible for other higher‑specification Fords such as the Capri Clubsprint and for fitting optional equipment such as LPG systems and sunroofs. In 1999, in addition to enhancing the Falcon line with the sporty XR range, TVE established Ford Tickford Experience (FTE), comprising a three‑tier, sedan‑only T series based on the new AU series. Managing Director David Flint summed up the brand’s intentions: "Tickford have helped add further refinement, safety and confidence to the T Series range." It is very easy to build a car that just goes fast, but one that handles, performs and lends itself to outstanding driving dynamics is what we have aimed for. In so doing, while the brand was an obvious attempt to combat the successful HSV products, FTE did not want to get into a power war, focusing instead on providing a more sophisticated, high-performance product and making refinement its hallmark. Aside from Falcon-based products, over the years FTE was also responsible for a right-hand-drive (RHD) conversion of the fourth-generation Mustang Cobra and for promoting the European Cougar. In 1999, FTE also bought out Glen Seton Racing to create Ford Tickford Racing—Ford's first factory-backed team in over 30 years. With the purchase of Tickford by Prodrive in 2001, FTE was replaced by the new owner's Ford Performance Vehicles (badged as "FPV") with the introduction of the BA-series Falcon range in 2002. Nevertheless, Tickford was still involved in the early development stages of some models, including the BA Falcon XR6 Turbo, with early parts bearing the "Tickford" badge. The race team Ford Tickford Racing was renamed Ford Performance Racing in 2003. In 2016, the Prodrive Racing Australia division reformed Tickford to offer high-performance upgrades for the Australian-imported Ford Mustang, Ranger and Everest. This followed Ford Australia shutting down local production, ending the Falcon model, and discontinuing the Ford Performance Vehicles brand. Ford Performance Racing was renamed Tickford Racing in 2017. Falcon XR range. This range of models — not to be confused with the 1966–1968 XR series — has been a fixture of the Falcon range since 1991. From its second release in 1993, it has been characterised by a signature quad-headlamp front styling that distinguishes it from standard models. In addition, all factory LPG systems for the EF and EL Falcon range were installed by Tickford and badged "Tickford." EB series. In July 1991, just prior to the formation of TVE, Ford Australia launched the EB Falcon S-XR8 after deciding to offer a V8 engine option again for the first time since the 1982 XE series. In 1992, TVE became responsible for the whole range. The S-XR6 was identified by a red rocker cover with the "Tickford wings" badge; a revised cylinder head and camshaft increased its power. The S-XR8 was visually similar but was powered by the standard 5.0-litre "Windsor" V8 engine. With the EB II, TVE launched an enhanced XR8 model. Both the six- and eight-cylinder models were fitted with ABS, a limited-slip differential (LSD), and alloy wheels; a Momo steering wheel was standard on the S-XR6 and optional on the S-XR8. ED series. With this series, TVE dropped the "S" from its range and introduced the Falcon XR's signature four-headlamp treatment, inspired by the European Ford Escort RS Cosworth. The model range comprised the XR6, the XR8, and — from September 1993 — the XR8 Sprint. The latter featured a more powerful version of the 5.0-litre Windsor V8 engine, courtesy of the performance upgrades fitted to the previous year's Falcon GT 25th Anniversary. The same GT also donated an improved suspension and brake package, and featured unique 16-inch wheels. There was very little body alteration to identify it, other than a black (instead of red) side body stripe, a pair of subtle front wheel arch moulds, "Sprint" badges on the boot and a front chin spoiler. EF series: The XR6 and XR8 models continued with the heavily re-engineered EF series Falcon, with more improvements to the suspension and driveline. The XR range now received a one-piece nose cone including quad headlamp surrounds, bumper and a blanked-out grille similar to the base GLi model. "Cats scratch" vents were applied to the bonnet.	Tickford Vehicle Engineering (TVE) was responsible for numerous automotive projects and upgrades for Ford Australia between nineteen ninety-one and two thousand two. In nineteen ninety-nine TVE set up Ford Tickford Experience (FTE) as a competitor to Holden Special Vehicles (HSV). In two thousand two the operations transitioned to Ford Performance Vehicles (FPV), coinciding with Tickford's global operations being bought out by Prodrive. Overview In nineteen ninety, twelve years after the last Falcon Cobra rolled off the line, Ford Australia began a worldwide search for an engineering firm to establish a new performance arm in the same mould as Holden Special Vehicles (HSV). The aim was to create a line-up of in-house performance cars that had been missing from the Falcon range since the nineteen seventies. Ford Australia's previous effort in this space had been the nineteen eighty-two XE-series "European Sports Pack" Falcon. Throughout the late nineteen eighties and early nineteen nineties this performance void was partly filled by third-party tuning companies such as AVO, Special Vehicle Operations (SVO) and Dick Johnson Racing (DJR). Both AVO and DJR experimented with turbocharging to boost output from the Falcon's four point zero-litre six-cylinder, with DJR creating the Grand Prix Turbo based on the XE. DJR had intended his company to function as a performance arm similar to Peter Brock's Holden Dealer Team (HDT), but Ford was wary of such modifications because of warranty concerns, so sales of both AVO and DJR Falcons remained very limited. SVO had greater success creating a range of cars, beginning with the Falcon EA series SVO, a model that formed the template for the subsequent XR six. However, Ford still did not provide factory backing. Ultimately, Ford Australia joined forces with Tickford; in nineteen ninety-one Tickford Vehicle Engineering (TVE) was established as a joint venture between Ford Australia and Tickford. TVE helped create the Falcon XR range, which emerged in nineteen ninety-two with the EB series and the return of the Falcon GT. TVE was also responsible for other higher-specification Fords such as the Capri Clubsprint and for fitting optional equipment such as LPG systems and sunroofs. In nineteen ninety-nine, in addition to enhancing the Falcon line with the sporty XR range, TVE established Ford Tickford Experience (FTE), comprising a three tier, sedan only T series based on the new AU series. Managing Director David Flint summed up the brand’s intentions: "Tickford have helped add further refinement, safety and confidence to the T Series range." It is very easy to build a car that just goes fast, but one that handles, performs and lends itself to outstanding driving dynamics is what we have aimed for. In so doing, while the brand was an obvious attempt to combat the successful HSV products, FTE did not want to get into a power war, focusing instead on providing a more sophisticated, high-performance product and making refinement its hallmark. Aside from Falcon-based products, over the years FTE was also responsible for a right hand drive (RHD) conversion of the fourth generation Mustang Cobra and for promoting the European Cougar. In nineteen ninety nine, FTE also bought out Glen Seton Racing to create Ford Tickford Racing—Ford's first factory backed team in over thirty years. With the purchase of Tickford by Prodrive in two thousand one, FTE was replaced by the new owner's Ford Performance Vehicles (badged as "FPV") with the introduction of the BA series Falcon range in two thousand two. Nevertheless, Tickford was still involved in the early development stages of some models, including the BA Falcon XR six Turbo, with early parts bearing the "Tickford" badge. The race team Ford Tickford Racing was renamed Ford Performance Racing in two thousand three. In two thousand sixteen, the Prodrive Racing Australia division reformed Tickford to offer high performance upgrades for the Australian imported Ford Mustang, Ranger and Everest. This followed Ford Australia shutting down local production, ending the Falcon model, and discontinuing the Ford Performance Vehicles brand. Ford Performance Racing was renamed Tickford Racing in two thousand seventeen. Falcon XR range. This range of models — not to be confused with the nineteen sixty six–nineteen sixty eight XR series — has been a fixture of the Falcon range since nineteen ninety one. From its second release in nineteen ninety three, it has been characterised by a signature quad headlamp front styling that distinguishes it from standard models. In addition, all factory LPG systems for the EF and EL Falcon range were installed by Tickford and badged "Tickford." EB series. In July nineteen ninety-one, just prior to the formation of TVE, Ford Australia launched the EB Falcon S XR eight after deciding to offer a V eight engine option again for the first time since the nineteen eighty-two XE series. In nineteen ninety-two, TVE became responsible for the whole range. The S XR six was identified by a red rocker cover with the "Tickford wings" badge; a revised cylinder head and camshaft increased its power. The S XR eight was visually similar but was powered by the standard five point zero-litre "Windsor" V eight engine. With the EB two, TVE launched an enhanced XR eight model. Both the six- and eight-cylinder models were fitted with ABS, a limited-slip differential (LSD), and alloy wheels; a Momo steering wheel was standard on the S XR six and optional on the S XR eight. ED series. With this series, TVE dropped the "S" from its range and introduced the Falcon XR's signature four-headlamp treatment, inspired by the European Ford Escort RS Cosworth. The model range comprised the XR six, the XR eight, and — from September nineteen ninety-three — the XR eight Sprint. The latter featured a more powerful version of the five point zero-litre Windsor V eight engine, courtesy of the performance upgrades fitted to the previous year's Falcon GT twenty-fifth Anniversary. The same GT also donated an improved suspension and brake package, and featured unique sixteen-inch wheels. There was very little body alteration to identify it, other than a black (instead of red) side body stripe, a pair of subtle front wheel arch moulds, "Sprint" badges on the boot and a front chin spoiler. EF series: The XR six and XR eight models continued with the heavily re-engineered EF series Falcon, with more improvements to the suspension and driveline. The XR range now received a one-piece nose cone including quad headlamp surrounds, bumper and a blanked-out grille similar to the base GLi model. "Cats scratch" vents were applied to the bonnet.
long_en_370	poet_en	1,027	en	Saturday afternoon, Winter Break 2016 — Week 1. I woke up with a slight headache. White bedsheets covered me. Wait—was Yuka here too? I checked my right side and saw Yuka’s head on the bed; she held a book in her left hand. “Oh, Yuki, you’re awake. What time is it?” Yuka asked. “Twelve ten,” I said, checking my cellphone. “I should be getting home. Do you need a ride, Yuki?” Yuka asked. “I think Tess will drive me back. Thanks for the offer, though,” I said. “Call me whenever you’re free, Yuki!” she said, departing. I walked to the bathroom and splashed cold water on my face. The pounding in my head gradually lessened and I headed downstairs. The living room was empty and spotless—no remnants of food or any sign we'd been there. They were probably in the dining room. “Good morning—rather, good afternoon, Yuki!” Felicity greeted. She seemed normal; I no longer detected the hostility or malice in her voice. She had a plate of eggs and bacon in front of her. I looked around but didn’t see Tess. “Felicity, any idea where Tess is?” I asked. “She’s probably finishing up cleaning the basement. She’ll be here soon. I would give you a ride home, but Memoria and I still have much to do,” Felicity said. Ten minutes later, Tess walked in. She greeted me and sat down at the table. I finished the food Felicity had offered and watched her. Did she recall the events of the party? I couldn’t tell from her bright, uplifting smile. "Tomo, you ready to go?" Tess asked. "Yeah, I'm ready," I replied. "I hope you had fun, Yuki! Let's do it again soon," Felicity said as we left. I turned to face the emotionless woman once we were in the car. She glanced at me as she placed her keys in the ignition. "Tess, how much do you remember from the party?" I asked, looking her straight in the eyes. "Everything. But you want to talk about Felicity, don't you?" Tess said, backing out of the parking lot. "Yeah. How do you feel about it?" I asked, staring at the large house. "There's not much to say. Felicity got pretty drunk and said a few things that were out of line. That's about it," Tess said. "Really? I feel like her emotions are pretty unstable. I'll take your word that she only said that because of the alcohol. How much does she really know about the heroes thing? I just want to know how much access she's had to you and everyone else," I asked. "Felicity knows of our existence and observes us when we train. She's not privy to the important matters. You'll learn about that in time," Tess answered as she drove through the large black gate. "Tess, what are your feelings about everyone as heroes?" I asked. "In terms of what, Tomo?" Tess asked, her voice very cold. "What do you mean by that?" I asked, confused. "State it more concretely, Tomo. As friends? As fighters? Or are you perhaps even including romantic feelings?" Tess clarified. "I meant as friends," I said. "Wait, do you actually like anyone, Tess?" I asked, her question catching my attention. "Not in terms of romance. But if you're speaking on the friendship level, I believe everyone has faults, like any human. I'm not as critical about their private lives as Fel is. As long as they perform when needed, I'm fine. I'm sure the others understand my feelings on that," Tess replied, stopping at a four-way stop. "Oh, I thought you'd be more critical, like you said. But I understand," I said, nodding. "And what about you, Tomo? Has anyone caught your eye?" Tess asked, her face straight as she looked in her rearview mirror. "What?" I stammered, caught off guard. "It's only fair that I reciprocate your question, isn't it?" Tess said. "You have a point. Say, Tess, you seem a lot more—how do I put this—cooler and calmer than usual," I said. "Don't worry about it. It's just the side effects of the alcohol. You still haven't answered my question," Tess reminded me. "I don't especially like anyone. It's hard when you barely know them. The party was my first real chance to see who they are as people. You too, Tess. When we're talking about the whole heroes thing, everyone puts up a kind of facade—not fake, just more reserved. I can understand; we're still getting to know each other," I answered, thinking about what I'd seen at the party. "That's a fair answer," Tess said. "Although Fel does need to control herself more and choose her words with care," Tess noted. "Do they not like Felicity? I mean, they're at least outwardly friendly with her, but there's tension," I asked. "I'm sure they're indifferent. Fel is the only one who really cares that much, and in some ways that's her weakness. Although Darryl may feel differently," Tess answered as she made a left turn. "You're pretty sharp, Tess, even with social interactions. It feels like you know too much — not that it's bad, it's just mysterious," I said, searching for the right words. "It's just what I do, Tomo. There's nothing more to it. By the way, you may want to check your underwear when you get home," Tess said suddenly as she neared my street. "What? Tess, what did you do?" I asked, surprised. "It's more about what Fel did than me. Anyway, you've arrived safely back at your house," Tess said as she pulled into my driveway and ended the conversation. "Tess, you gotta tell me. It's freaking me out the way you said it," I pleaded. Tess remained silent and waved goodbye. Damn — I'd have to find out what she meant when I got back to my room.	Saturday afternoon, Winter Break two thousand sixteen — Week one. I woke up with a slight headache. White bedsheets covered me. Wait—was Yuka here too? I checked my right side and saw Yuka’s head on the bed; she held a book in her left hand. “Oh, Yuki, you’re awake. What time is it?” Yuka asked. “Twelve ten,” I said, checking my cellphone. “I should be getting home. Do you need a ride, Yuki?” Yuka asked. “I think Tess will drive me back. Thanks for the offer, though,” I said. “Call me whenever you’re free, Yuki!” she said, departing. I walked to the bathroom and splashed cold water on my face. The pounding in my head gradually lessened and I headed downstairs. The living room was empty and spotless—no remnants of food or any sign we'd been there. They were probably in the dining room. “Good morning—rather, good afternoon, Yuki!” Felicity greeted. She seemed normal; I no longer detected the hostility or malice in her voice. She had a plate of eggs and bacon in front of her. I looked around but didn’t see Tess. “Felicity, any idea where Tess is?” I asked. “She’s probably finishing up cleaning the basement. She’ll be here soon. I would give you a ride home, but Memoria and I still have much to do,” Felicity said. Ten minutes later, Tess walked in. She greeted me and sat down at the table. I finished the food Felicity had offered and watched her. Did she recall the events of the party? I couldn’t tell from her bright, uplifting smile. "Tomo, you ready to go?" Tess asked. "Yeah, I'm ready," I replied. "I hope you had fun, Yuki! Let's do it again soon," Felicity said as we left. I turned to face the emotionless woman once we were in the car. She glanced at me as she placed her keys in the ignition. "Tess, how much do you remember from the party?" I asked, looking her straight in the eyes. "Everything. But you want to talk about Felicity, don't you?" Tess said, backing out of the parking lot. "Yeah. How do you feel about it?" I asked, staring at the large house. "There's not much to say. Felicity got pretty drunk and said a few things that were out of line. That's about it," Tess said. "Really? I feel like her emotions are pretty unstable. I'll take your word that she only said that because of the alcohol. How much does she really know about the heroes thing? I just want to know how much access she's had to you and everyone else," I asked. "Felicity knows of our existence and observes us when we train. She's not privy to the important matters. You'll learn about that in time," Tess answered as she drove through the large black gate. "Tess, what are your feelings about everyone as heroes?" I asked. "In terms of what, Tomo?" Tess asked, her voice very cold. "What do you mean by that?" I asked, confused. "State it more concretely, Tomo. As friends? As fighters? Or are you perhaps even including romantic feelings?" Tess clarified. "I meant as friends," I said. "Wait, do you actually like anyone, Tess?" I asked, her question catching my attention. "Not in terms of romance. But if you're speaking on the friendship level, I believe everyone has faults, like any human. I'm not as critical about their private lives as Fel is. As long as they perform when needed, I'm fine. I'm sure the others understand my feelings on that," Tess replied, stopping at a four-way stop. "Oh, I thought you'd be more critical, like you said. But I understand," I said, nodding. "And what about you, Tomo? Has anyone caught your eye?" Tess asked, her face straight as she looked in her rearview mirror. "What?" I stammered, caught off guard. "It's only fair that I reciprocate your question, isn't it?" Tess said. "You have a point. Say, Tess, you seem a lot more—how do I put this—cooler and calmer than usual," I said. "Don't worry about it. It's just the side effects of the alcohol. You still haven't answered my question," Tess reminded me. "I don't especially like anyone. It's hard when you barely know them. The party was my first real chance to see who they are as people. You too, Tess. When we're talking about the whole heroes thing, everyone puts up a kind of facade—not fake, just more reserved. I can understand; we're still getting to know each other," I answered, thinking about what I'd seen at the party. "That's a fair answer," Tess said. "Although Fel does need to control herself more and choose her words with care," Tess noted. "Do they not like Felicity? I mean, they're at least outwardly friendly with her, but there's tension," I asked. "I'm sure they're indifferent. Fel is the only one who really cares that much, and in some ways that's her weakness. Although Darryl may feel differently," Tess answered as she made a left turn. "You're pretty sharp, Tess, even with social interactions. It feels like you know too much — not that it's bad, it's just mysterious," I said, searching for the right words. "It's just what I do, Tomo. There's nothing more to it. By the way, you may want to check your underwear when you get home," Tess said suddenly as she neared my street. "What? Tess, what did you do?" I asked, surprised. "It's more about what Fel did than me. Anyway, you've arrived safely back at your house," Tess said as she pulled into my driveway and ended the conversation. "Tess, you gotta tell me. It's freaking me out the way you said it," I pleaded. Tess remained silent and waved goodbye. Damn — I'd have to find out what she meant when I got back to my room.
long_en_360	poet_en	989	en	You know those days that change your life forever? Well, that's exactly what happened two years ago today. On July 31, 2012, Charlie Carter came into this world a little after noon, weighing 8 pounds, 9 ounces. I sat in my bed last night with my sweet new baby girl asleep in her rocker right next to me, thinking about how two years ago I fell asleep in a hospital bed getting ready for his arrival the next day. The nurse called it our last "date night" for a while. Ha. My blood pressure was high. It was hot outside. I was rocking a wicked headache. I remember being so excited and so scared at the same time. I had no idea what to expect from the whole childbirth process, and even though I thought I was prepared to be a mom, it was clear to me the first time they put him in my arms after our long ten-month journey together and we finally met — I had no idea what I was in for. I couldn't help but start to cry thinking about the whole thing last night: my first experience with motherhood. All those memories are so wonderful — so scary, but so wonderful. The first night we had Charlie at home we put him in the Pack 'n Play next to our bed, and I cried, telling Jimmy, "What happens if we can't take care of him? What if we don't know what to do?!" But then I see you smile and laugh and tell us how the train says "choo choo" over and over again. You insist on milk — wait, you change your mind: you'll take water, and then you change your mind again. You point out every body part on yourself and others. You’re curious about differences between boys and girls—that’s something I’ll explain later, sweet boy. Your hugs are the best, but your kisses are even better, and you're especially gentle and loving with your little sister. It all makes me realize that even if your dad and I don’t always do everything "right" or know what we’re doing, as long as we love you—which we do so much—we are, in fact, doing this parenting thing the right way. We love you so much, sweet boy. The love I felt the first time I held you just grows stronger each day. You are funny, fun-loving, sweet, social, and hilarious. Your curly hair is adorable and your toothy smile is even cuter. You’re not the best eater and you love to snack. You never sit still and always want to be doing something. You’re the best dancer—really, your dance moves are amazing. So here we are: your birthday is coming to a close. Next year at this time you’ll be three, and I can’t even bear to think about that, but I know it will be here before we know it. I pray that you continue to live this next year to the fullest. Keep smiling and laughing, hugging and kissing. Keep saying your prayers at night and brushing your teeth. Keep loving me, your dad, and your sister. And don't kick the dogs! While so much has changed in the last year, I'm sure there will be more changes next year—moving to a big-boy bed and no diapers (hopefully!)—so brace yourself, young man, because we're in this together. Thank you, Charlie, for making me a mom. I love being a "boy mom" so much, and especially being your mom. When I hear you call for "mommy" and want to snuggle and kiss me, there is no better feeling in this world. This is real life: trying to get a picture with a toddler. Then he screamed his head off the whole way home. Watch out, terrible twos! So that called for a celebration—and celebrate we did this weekend! As his party got closer, it was way more fun than last year because he actually understood what was going on and talked all about his party. Needless to say, as planning began, I knew I didn't want it at our house. While we do have a much bigger house than last year, factor in that we have a three-week-old (if she had come on her actual due date, she would be two weeks old), I didn't want to deal with the cleaning beforehand, the extensive setup, and the cleanup. We had taken Charlie there a few times this past year, and he loved it. It's a huge indoor playground—a win for burning off two-year-old energy—and it's in a contained area, so you don't have to worry about them running off like at a park (wait—your child doesn't do that?). Factor in that my kid was born in mid-July, which usually means hot temperatures — but it's indoors, so no one had to sweat. Add that the whole event was an hour and a half — even better, because I could fit feeding my little baby girl in, too! I booked it and never looked back. I could have taken advantage of all the services they offered — food, drinks, etc. — but that's not how I roll. I'm more of a semi-homemade type of person, and it worked great. Seriously, I keep thinking back to the smile on his cute face as he sprinted between slides and can't help but smile. It was so much fun for everyone. Our package allowed 12 kids to play (with one other party group) for an hour, followed by 30 minutes in the party room. The catch was that you only had 10 minutes to set up because another party was in the room before you.	You know those days that change your life forever? Well, that's exactly what happened two years ago today. On July thirty one, two thousand twelve, Charlie Carter came into this world a little after noon, weighing eight pounds, nine ounces. I sat in my bed last night with my sweet new baby girl asleep in her rocker right next to me, thinking about how two years ago I fell asleep in a hospital bed getting ready for his arrival the next day. The nurse called it our last "date night" for a while. Ha. My blood pressure was high. It was hot outside. I was rocking a wicked headache. I remember being so excited and so scared at the same time. I had no idea what to expect from the whole childbirth process, and even though I thought I was prepared to be a mom, it was clear to me the first time they put him in my arms after our long ten-month journey together and we finally met — I had no idea what I was in for. I couldn't help but start to cry thinking about the whole thing last night: my first experience with motherhood. All those memories are so wonderful — so scary, but so wonderful. The first night we had Charlie at home we put him in the Pack 'n Play next to our bed, and I cried, telling Jimmy, "What happens if we can't take care of him? What if we don't know what to do?!" But then I see you smile and laugh and tell us how the train says "choo choo" over and over again. You insist on milk — wait, you change your mind: you'll take water, and then you change your mind again. You point out every body part on yourself and others. You’re curious about differences between boys and girls—that’s something I’ll explain later, sweet boy. Your hugs are the best, but your kisses are even better, and you're especially gentle and loving with your little sister. It all makes me realize that even if your dad and I don’t always do everything "right" or know what we’re doing, as long as we love you—which we do so much—we are, in fact, doing this parenting thing the right way. We love you so much, sweet boy. The love I felt the first time I held you just grows stronger each day. You are funny, fun-loving, sweet, social, and hilarious. Your curly hair is adorable and your toothy smile is even cuter. You’re not the best eater and you love to snack. You never sit still and always want to be doing something. You’re the best dancer—really, your dance moves are amazing. So here we are: your birthday is coming to a close. Next year at this time you’ll be three, and I can’t even bear to think about that, but I know it will be here before we know it. I pray that you continue to live this next year to the fullest. Keep smiling and laughing, hugging and kissing. Keep saying your prayers at night and brushing your teeth. Keep loving me, your dad, and your sister. And don't kick the dogs! While so much has changed in the last year, I'm sure there will be more changes next year—moving to a big-boy bed and no diapers (hopefully!)—so brace yourself, young man, because we're in this together. Thank you, Charlie, for making me a mom. I love being a "boy mom" so much, and especially being your mom. When I hear you call for "mommy" and want to snuggle and kiss me, there is no better feeling in this world. This is real life: trying to get a picture with a toddler. Then he screamed his head off the whole way home. Watch out, terrible twos! So that called for a celebration—and celebrate we did this weekend! As his party got closer, it was way more fun than last year because he actually understood what was going on and talked all about his party. Needless to say, as planning began, I knew I didn't want it at our house. While we do have a much bigger house than last year, factor in that we have a three-week-old (if she had come on her actual due date, she would be two weeks old), I didn't want to deal with the cleaning beforehand, the extensive setup, and the cleanup. We had taken Charlie there a few times this past year, and he loved it. It's a huge indoor playground—a win for burning off two-year-old energy—and it's in a contained area, so you don't have to worry about them running off like at a park (wait—your child doesn't do that?). Factor in that my kid was born in mid-July, which usually means hot temperatures — but it's indoors, so no one had to sweat. Add that the whole event was an hour and a half — even better, because I could fit feeding my little baby girl in, too! I booked it and never looked back. I could have taken advantage of all the services they offered — food, drinks, etc. — but that's not how I roll. I'm more of a semi-homemade type of person, and it worked great. Seriously, I keep thinking back to the smile on his cute face as he sprinted between slides and can't help but smile. It was so much fun for everyone. Our package allowed twelve kids to play (with one other party group) for an hour, followed by thirty minutes in the party room. The catch was that you only had ten minutes to set up because another party was in the room before you.
long_en_371	poet_en	970	en	Keaton's birthday was this week. He is now officially double-digits. It started the night before, when we made cookies for his class at school. Apparently he had been preparing his classmates for "my mom's famous chocolate chip cookies... which are the best you've ever had," so we made them together, and I just hoped they would live up to his hype. The morning of his birthday, I sent him to school with 30 cookies to share, which everyone seemed to like (thank goodness). We had spaghetti and meatballs for his birthday dinner, then took our family picture, which is tradition whenever we celebrate someone's birthday—first normal, then crazy. After that I realized that I hadn't wrapped any of Keaton's birthday gifts, so I sent him on a scavenger hunt around the house to find his three gifts, which apparently was the best part of his birthday when I asked him later what his favorite was. He ended up with a basketball, a soccer ball, a birthstone necklace (which he's been asking for for months), and a pair of rollerblades so he can race Kolby around the block. Keaton is a very helpful and tender boy. He loves scouts, riding his bike, and helping others. He has developed an interest in reading lately, and thus he reads every night in his bed before he goes to sleep. He likes to watch "Dino Dan" on TV, rollerblade and ride his bike around the neighborhood, and is intent on making sure that everyone at school has a friend to play with. I love my sweet Keaton. Happy 10th birthday! Based on Kimble's latest obsession with coloring, I have another lefty on my hands. That makes four out of five children in our household lefties. Which one is the odd man out? Take your guesses. At yesterday's market outside a local elementary school, one boy bought one of my bowties; then all his friends bought bowties. I sold about 14 bowties, all to a group of fifth-grade boys who wore them proudly around their necks. It was a sight to see. Last week I had planned to take Mr. Kimble to the zoo. We were going to meet up with two friends and their mommies and have a fun morning before nap time. That morning the husband took the day off work and, after discovering our plans, joined us, since he had never been to our zoo before. I've never seen Mr. Moose do anything but sleep; that day was no exception. It was a perfect day, after all, to stroll around and see the sights. We may have to suggest this to Mr. Moose the next time we see him. I'm so glad Heath came with us. He doesn't often get to spend so much daytime with Kimble. I think they both had a great time. A few weeks ago we had a massive Wednesday night storm that didn't let up until the next day. It was such a forceful rainstorm that it caused flash floods throughout our area. Right about dinnertime, Kolby shouted in a panicked voice, "There is water coming into our basement through the window!" We went downstairs and noticed that water had accumulated in one of the basement window wells and was about one-third of the way up the window. My husband immediately took the boys outside in the torrential downpour and plopped them into the window well. Armed with pitchers, they started dumping the water out of the well as if their canoe were capsizing. Kennedy and I stayed inside, soaking up water with towels and trying to prevent more from coming in. We watched Keaton and Kolby valiantly try to dump out as much as they could. Keaton, a little taller than Kolby, could barely get his pitcher up and over the side. Kolby, on the other hand, would scoop up a pitcher full of water and fling it toward the edge; most of the time half of it fell back in because he couldn't quite get it over. Despite worrying about the basement flooding, I found myself laughing while watching the boys. As an interesting note, the first window well to overflow was the one that housed a black widow spider, which we morbidly watched day after day as it grew. The skeleton of her dead lover remained on display beside her web, a trophy of her power. After about a week, my husband torched her with our BBQ lighter, and she was finally free to join her beloved in the afterlife. As the boys were knee-deep in water, emptying the well, I was happy to remember that the black widow spider was no longer extending her death bites to anyone or anything that crossed her path. Mr. Kimble loves to play with lightsabers now. He giggled and giggled when we were playing, and it was so fun to watch. Missing socks, missing pants, blurry photos because he was moving so much—it's all good. We had a fun time together today. 1. Kimble isn't such a little baby anymore. He is getting bigger seemingly every day. I keep trimming his hair regularly, but you wouldn't know it. With half of his hair curly and the other half straight, it's hard to keep him looking nicely kept, but we are trying. He has also matured so quickly that he is sporting a mustache now. 2. Kimble must be looking at the developmental papers that I have to fill out for his regular checkups. "Can your child make a stack of three blocks?" Check. Yes, he can, although he prefers to stack cans from my pantry. 3. Kimble's list of words is growing.	Keaton's birthday was this week. He is now officially double-digits. It started the night before, when we made cookies for his class at school. Apparently he had been preparing his classmates for "my mom's famous chocolate chip cookies... which are the best you've ever had," so we made them together, and I just hoped they would live up to his hype. The morning of his birthday, I sent him to school with thirty cookies to share, which everyone seemed to like (thank goodness). We had spaghetti and meatballs for his birthday dinner, then took our family picture, which is tradition whenever we celebrate someone's birthday—first normal, then crazy. After that I realized that I hadn't wrapped any of Keaton's birthday gifts, so I sent him on a scavenger hunt around the house to find his three gifts, which apparently was the best part of his birthday when I asked him later what his favorite was. He ended up with a basketball, a soccer ball, a birthstone necklace (which he's been asking for for months), and a pair of rollerblades so he can race Kolby around the block. Keaton is a very helpful and tender boy. He loves scouts, riding his bike, and helping others. He has developed an interest in reading lately, and thus he reads every night in his bed before he goes to sleep. He likes to watch "Dino Dan" on TV, rollerblade and ride his bike around the neighborhood, and is intent on making sure that everyone at school has a friend to play with. I love my sweet Keaton. Happy ten th birthday! Based on Kimble's latest obsession with coloring, I have another lefty on my hands. That makes four out of five children in our household lefties. Which one is the odd man out? Take your guesses. At yesterday's market outside a local elementary school, one boy bought one of my bowties; then all his friends bought bowties. I sold about fourteen bowties, all to a group of fifth-grade boys who wore them proudly around their necks. It was a sight to see. Last week I had planned to take Mr. Kimble to the zoo. We were going to meet up with two friends and their mommies and have a fun morning before nap time. That morning the husband took the day off work and, after discovering our plans, joined us, since he had never been to our zoo before. I've never seen Mr. Moose do anything but sleep; that day was no exception. It was a perfect day, after all, to stroll around and see the sights. We may have to suggest this to Mr. Moose the next time we see him. I'm so glad Heath came with us. He doesn't often get to spend so much daytime with Kimble. I think they both had a great time. A few weeks ago we had a massive Wednesday night storm that didn't let up until the next day. It was such a forceful rainstorm that it caused flash floods throughout our area. Right about dinnertime, Kolby shouted in a panicked voice, "There is water coming into our basement through the window!" We went downstairs and noticed that water had accumulated in one of the basement window wells and was about one-third of the way up the window. My husband immediately took the boys outside in the torrential downpour and plopped them into the window well. Armed with pitchers, they started dumping the water out of the well as if their canoe were capsizing. Kennedy and I stayed inside, soaking up water with towels and trying to prevent more from coming in. We watched Keaton and Kolby valiantly try to dump out as much as they could. Keaton, a little taller than Kolby, could barely get his pitcher up and over the side. Kolby, on the other hand, would scoop up a pitcher full of water and fling it toward the edge; most of the time half of it fell back in because he couldn't quite get it over. Despite worrying about the basement flooding, I found myself laughing while watching the boys. As an interesting note, the first window well to overflow was the one that housed a black widow spider, which we morbidly watched day after day as it grew. The skeleton of her dead lover remained on display beside her web, a trophy of her power. After about a week, my husband torched her with our BBQ lighter, and she was finally free to join her beloved in the afterlife. As the boys were knee-deep in water, emptying the well, I was happy to remember that the black widow spider was no longer extending her death bites to anyone or anything that crossed her path. Mr. Kimble loves to play with lightsabers now. He giggled and giggled when we were playing, and it was so fun to watch. Missing socks, missing pants, blurry photos because he was moving so much—it's all good. We had a fun time together today. one. Kimble isn't such a little baby anymore. He is getting bigger seemingly every day. I keep trimming his hair regularly, but you wouldn't know it. With half of his hair curly and the other half straight, it's hard to keep him looking nicely kept, but we are trying. He has also matured so quickly that he is sporting a mustache now. two. Kimble must be looking at the developmental papers that I have to fill out for his regular checkups. "Can your child make a stack of three blocks?" Check. Yes, he can, although he prefers to stack cans from my pantry. three. Kimble's list of words is growing.
long_en_270	wiki_en	1,111	en	In functional analysis and related areas, a complete topological vector space (TVS) is a topological vector space in which every Cauchy net or Cauchy filter converges to a point of the space. The notion of "points that get progressively closer" is made rigorous by Cauchy nets or Cauchy filters, which generalize Cauchy sequences, while "a point towards which they all get closer" means that such a net or filter converges to that point. Completeness for TVSs is formulated using the theory of uniform spaces to generalize completeness from metric spaces; unlike metric completeness, TVS-completeness does not depend on any specific metric and applies to TVSs that may not be metrizable or Hausdorff. For normed spaces and metrizable TVSs, the usual completeness defined by a particular norm or metric coincides with this notion of TVS-completeness, which is independent of any particular choice of norm or metric. A metrizable topological vector space with a translation-invariant metric is complete as a TVS if and only if it is a complete metric space, which by definition means that every Cauchy sequence converges to some point in the space. Prominent examples of complete TVSs that are also metrizable include all F-spaces and, consequently, all Fréchet spaces, Banach spaces, and Hilbert spaces. Prominent examples of complete TVSs that are typically non-metrizable include strict LF-spaces such as the space of test functions with its canonical LF-topology, the strong dual of any non-normable Fréchet space, and many polar topologies on continuous duals or other topologies on spaces of linear maps. Explicitly, a topological vector space (TVS) is complete if every net, or equivalently every filter, that is Cauchy with respect to the space's canonical uniformity converges to some point. Equivalently, a TVS is complete if its canonical uniformity is a complete uniformity. The canonical uniformity on a TVS is the unique translation-invariant uniformity that induces the given topology. This notion of TVS-completeness depends on vector subtraction and the topology of the TVS and therefore applies to all TVSs, including those whose topologies cannot be defined in terms of metrics or pseudometrics. A first-countable TVS is complete if and only if every Cauchy sequence (or equivalently every elementary Cauchy filter) converges. Every topological vector space, even if it is not metrizable or Hausdorff, has a completion, which by definition is a complete TVS into which it can be TVS-embedded as a dense vector subspace. Moreover, when the TVS is Hausdorff the completion is unique up to TVS-isomorphism. In contrast, non-Hausdorff TVSs have infinitely many non-Hausdorff completions that are TVS-isomorphic to one another. Definitions. This section summarizes the definition of a complete topological vector space (TVS) in terms of nets and prefilters. Information about convergence of nets and filters can be found in the article on filters in topology. Every TVS is a commutative topological group under addition, and its canonical uniformity is defined in terms of subtraction (hence addition); scalar multiplication is not involved. Canonical uniformity. Let X be a TVS. For any subset U of X define the entourage V_U = {(x,y) in X×X : x - y is in U}. The canonical uniformity on X is generated by the entourages V_U as U runs over neighborhoods of 0. If U is symmetric (U = -U) then V_U is symmetric; moreover the composition V_U composed with V_U is contained in V_{U+U}, so the family of such entourages satisfies the usual uniformity axioms. If B is any neighborhood basis at the origin, then {V_U : U in B} is a base of entourages (a prefilter on X×X). Equivalently, if N is the neighborhood filter at the origin, then {V_U : U in N} forms a base of entourages for the canonical uniform structure on X. Explicitly, the canonical uniformity on a topological vector space X is the filter on X × X generated by the prefilter of entourages of the diagonal of the form {(x,y) : x − y ∈ U}, where U ranges over the neighborhoods of the origin 0 in X. The same canonical uniformity results if one uses any neighborhood basis at the origin instead of the full filter of all neighborhoods: if B is any neighborhood basis at 0 in X, then the filter on X × X generated by {{(x,y) : x − y ∈ U} : U ∈ B} equals the canonical uniformity induced by the topology of X. The general theory of uniform spaces has definitions of a "Cauchy prefilter" and a "Cauchy net". For the canonical uniformity on a TVS these definitions reduce to the following concrete description. Suppose (x_i)_{i∈I} is a net in X and (y_j)_{j∈J} is a net in X. The product I × J becomes a directed set by declaring (i,j) ≤ (i',j') iff i ≤ i' and j ≤ j'. The product net (x_i, y_j)_{(i,j)∈I×J} is the Cartesian product net. Its image under vector addition is the (I × J)-indexed net x_i + y_j, and its image under vector subtraction is the (I × J)-indexed net x_i − y_j. In particular, the notation x_{i,j} denotes the (I × J)-indexed net and not an I-indexed net. A net (x_i) in a TVS is called a Cauchy net if for every neighborhood U of 0 in X there exists i_0 ∈ I such that for all i, i' ≥ i_0 we have x_i − x_{i'} ∈ U. It suffices to check this condition for any given neighborhood basis at 0. A Cauchy sequence is a sequence that is also a Cauchy net. Continuity of the vector subtraction map implies that if xα → x and yα → y then xα − yα → x − y, so every convergent net is a Cauchy net. By definition, a space is complete if the converse always holds: a topological vector space is complete if and only if every Cauchy net in the space converges to some point of the space. The same characterization of completeness holds when using filters or prefilters instead of nets. A series is called a Cauchy series (respectively, a convergent series) if the sequence of partial sums is a Cauchy sequence (respectively, a convergent sequence). Every convergent series is a Cauchy series, and in a complete topological vector space every Cauchy series converges. A prefilter on a topological vector space is called a Cauchy prefilter if for every neighborhood U of 0 there exists a member A of the prefilter with A − A ⊂ U; equivalently, the family {A − A : A in the prefilter} forms a prefilter that "converges" to 0.	In functional analysis and related areas, a complete topological vector space (TVS) is a topological vector space in which every Cauchy net or Cauchy filter converges to a point of the space. The notion of "points that get progressively closer" is made rigorous by Cauchy nets or Cauchy filters, which generalize Cauchy sequences, while "a point towards which they all get closer" means that such a net or filter converges to that point. Completeness for TVSs is formulated using the theory of uniform spaces to generalize completeness from metric spaces; unlike metric completeness, TVS-completeness does not depend on any specific metric and applies to TVSs that may not be metrizable or Hausdorff. For normed spaces and metrizable TVSs, the usual completeness defined by a particular norm or metric coincides with this notion of TVS-completeness, which is independent of any particular choice of norm or metric. A metrizable topological vector space with a translation-invariant metric is complete as a TVS if and only if it is a complete metric space, which by definition means that every Cauchy sequence converges to some point in the space. Prominent examples of complete TVSs that are also metrizable include all F-spaces and, consequently, all Fréchet spaces, Banach spaces, and Hilbert spaces. Prominent examples of complete TVSs that are typically non metrizable include strict LF spaces such as the space of test functions with its canonical LF topology, the strong dual of any non normable Fréchet space, and many polar topologies on continuous duals or other topologies on spaces of linear maps. Explicitly, a topological vector space (TVS) is complete if every net, or equivalently every filter, that is Cauchy with respect to the space's canonical uniformity converges to some point. Equivalently, a TVS is complete if its canonical uniformity is a complete uniformity. The canonical uniformity on a TVS is the unique translation invariant uniformity that induces the given topology. This notion of TVS completeness depends on vector subtraction and the topology of the TVS and therefore applies to all TVSs, including those whose topologies cannot be defined in terms of metrics or pseudometrics. A first countable TVS is complete if and only if every Cauchy sequence (or equivalently every elementary Cauchy filter) converges. Every topological vector space, even if it is not metrizable or Hausdorff, has a completion, which by definition is a complete TVS into which it can be TVS embedded as a dense vector subspace. Moreover, when the TVS is Hausdorff the completion is unique up to TVS isomorphism. In contrast, non Hausdorff TVSs have infinitely many non Hausdorff completions that are TVS isomorphic to one another. Definitions. This section summarizes the definition of a complete topological vector space (TVS) in terms of nets and prefilters. Information about convergence of nets and filters can be found in the article on filters in topology. Every TVS is a commutative topological group under addition, and its canonical uniformity is defined in terms of subtraction (hence addition); scalar multiplication is not involved. Canonical uniformity. Let X be a TVS. For any subset U of X define the entourage V U equals {(x,y) in X times X : x minus y is in U}. The canonical uniformity on X is generated by the entourages V U as U runs over neighborhoods of zero. If U is symmetric (U equals negative U) then V U is symmetric; moreover the composition V U composed with V U is contained in V U plus U, so the family of such entourages satisfies the usual uniformity axioms. If B is any neighborhood basis at the origin, then {V U : U in B} is a base of entourages (a prefilter on X times X). Equivalently, if N is the neighborhood filter at the origin, then {V U : U in N} forms a base of entourages for the canonical uniform structure on X. Explicitly, the canonical uniformity on a topological vector space X is the filter on X times X generated by the prefilter of entourages of the diagonal of the form {(x,y) : x minus y is in U}, where U ranges over the neighborhoods of the origin zero in X. The same canonical uniformity results if one uses any neighborhood basis at the origin instead of the full filter of all neighborhoods: if B is any neighborhood basis at zero in X, then the filter on X times X generated by {{(x,y) : x minus y in U} : U in B} equals the canonical uniformity induced by the topology of X. The general theory of uniform spaces has definitions of a "Cauchy prefilter" and a "Cauchy net". For the canonical uniformity on a TVS these definitions reduce to the following concrete description. Suppose (x i)_{i in I} is a net in X and (y j)_{j in J} is a net in X. The product I times J becomes a directed set by declaring (i,j) is less than or equal to (i prime,j prime) if and only if i is less than or equal to i prime and j is less than or equal to j prime. The product net (x i, y j)_{(i,j) in I times J} is the Cartesian product net. Its image under vector addition is the (I times J)-indexed net x i plus y j, and its image under vector subtraction is the (I times J)-indexed net x i minus y j. In particular, the notation x i,j denotes the (I times J)-indexed net and not an I-indexed net. A net (x i) in a TVS is called a Cauchy net if for every neighborhood U of zero in X there exists i zero in I such that for all i, i prime is greater than or equal to i zero we have x i minus x i prime in U. It suffices to check this condition for any given neighborhood basis at zero. A Cauchy sequence is a sequence that is also a Cauchy net. Continuity of the vector subtraction map implies that if x alpha converges to x and y alpha converges to y then x alpha minus y alpha converges to x minus y, so every convergent net is a Cauchy net. By definition, a space is complete if the converse always holds: a topological vector space is complete if and only if every Cauchy net in the space converges to some point of the space. The same characterization of completeness holds when using filters or prefilters instead of nets. A series is called a Cauchy series (respectively, a convergent series) if the sequence of partial sums is a Cauchy sequence (respectively, a convergent sequence). Every convergent series is a Cauchy series, and in a complete topological vector space every Cauchy series converges. A prefilter on a topological vector space is called a Cauchy prefilter if for every neighborhood U of zero there exists a member A of the prefilter with A minus A is contained in U; equivalently, the family A minus A : A in the prefilter forms a prefilter that "converges" to zero.
long_en_358	poet_en	878	en	It has been twenty-three years since Uncle Mike passed away on May 12, but his memory—especially the good times he shared with Adna—remains alive. Aunt Felip passed away on January 12, and Aunt Feliz on September 12. Every day, Adna prays for them, including Papa Mar, Mom, Rona, and Uncle Rom. They are always remembered by Adna and remain in the hearts of her and her whole family. Uncle Mike was a simple man, yet he always thought of Adna when he harvested mangoes and jackfruits. A high school classmate told Adna that during the month there would be various venues for their class reunion, including events for classmates coming home from other countries. When she learned that one of the activities was island hopping, Adna said, “I can’t go across the sea; I don’t know how to swim.” She never forgets the time she almost drowned and how Uncle Mike, standing on the beach within reach, saved her. “Thanks to Uncle Mike,” she always tells everyone. On Mother’s Day—the second Sunday of the month—mothers are greeted with “Happy Mother’s Day.” Malls display many flowers; they are often expensive, but many people still buy them for their mothers. Adna wished Mom Rona were still alive so she could buy her a beautifully arranged bouquet. Adna greeted Mom Rona during her morning prayers. She also remembered Al’s mother. Lex, Azia, and Al II had something special planned for their mother, Adna. Early in the morning, Al, Adna, Azia, and Al II went to the place for the nuga session. It was Al II’s first time experiencing the sensation of warmth while lying down. There was also a dancing exercise, which Adna and Al II joined. Azia enjoyed watching Al II and, since Lex wasn’t with them, took a video of him dancing with others, including Adna. They stayed there for two hours and then had lunch. Lex, who had been left at home, was called by Adna but refused to go to the mall because she was busy with paperwork, so Azia and Al II went home to be with their sister. Al and Adna stayed at the mall, did some window shopping, and ate again because Adna likes to try small dishes at the food court. Afterwards they went to another mall to look for materials Al II needed for his grape plants. When they couldn’t find them, they went to a different mall. Adna remembered how Al’s mother used to go with the family from one mall to another. Al's younger sister married a foreigner and moved to his country. She petitioned for their mother, who then immigrated to live with them. Arrangements were made for her medicines to be sent in packages; Adna was responsible for sending them. The mother had diabetes and hypertension and needed daily medication. Since both the younger sister and her husband were too busy with work, she was often alone in their house. It took the couple a long time to have children. Al's mother, who had been a high school teacher before retiring, was used to having many students; when she was alone she missed her native land. When Al's older sister was on a business trip to the country where their mother had migrated, the mother told her, "I'm going home with you." The younger sister did not expect this decision, as she knew their mother favored her because she was the youngest. The younger sister could not stop her mother, and she had no influence over their older sister, who was domineering. The older sister bought a plane ticket for their mother, and they flew home immediately. At a regular checkup with the family doctor it was found that she already had complications. Every time she was hospitalized, Adna took care of her. Her memory began to fail, but she always remembered Adna's name. Once, a male nurse brought medicine to Al's mother, but she refused to take it. "I want only Adna—she is the only one I trust," she told the nurse. When Adna, who had just returned from the hospital pharmacy, came back, the woman said, "Your smile is so familiar; it looks like Adna's." Adna replied, "It's me, Adna!" The woman laughed, "Oh good, you're here. A man came and tried to force something on me—I thought it was poison. I told him I only trust you." The male nurse came in again and asked, "Who is Adna?" "I am," Adna replied. The nurse handed the tablet to Adna and said, "She only trusts you." Adna received the tablet and gave it to her mother-in-law. It was easy for Adna to get her mother-in-law to take the tablet. The male nurse was amazed at what he had witnessed. After Al's mother was discharged from the hospital, Al's younger sister was very grateful to Adna. She asked Adna to accompany her mother to the beauty salon to fix her hair and have it colored regularly. That Christmas, she sent money so they could go mall-hopping with the children Lex, Azia, and Al the Second.	It has been twenty-three years since Uncle Mike passed away on May twelve, but his memory—especially the good times he shared with Adna—remains alive. Aunt Felip passed away on January twelve, and Aunt Feliz on September twelve. Every day, Adna prays for them, including Papa Mar, Mom, Rona, and Uncle Rom. They are always remembered by Adna and remain in the hearts of her and her whole family. Uncle Mike was a simple man, yet he always thought of Adna when he harvested mangoes and jackfruits. A high school classmate told Adna that during the month there would be various venues for their class reunion, including events for classmates coming home from other countries. When she learned that one of the activities was island hopping, Adna said, “I can’t go across the sea; I don’t know how to swim.” She never forgets the time she almost drowned and how Uncle Mike, standing on the beach within reach, saved her. “Thanks to Uncle Mike,” she always tells everyone. On Mother’s Day—the second Sunday of the month—mothers are greeted with “Happy Mother’s Day.” Malls display many flowers; they are often expensive, but many people still buy them for their mothers. Adna wished Mom Rona were still alive so she could buy her a beautifully arranged bouquet. Adna greeted Mom Rona during her morning prayers. She also remembered Al’s mother. Lex, Azia, and Al two had something special planned for their mother, Adna. Early in the morning, Al, Adna, Azia, and Al two went to the place for the nuga session. It was Al two’s first time experiencing the sensation of warmth while lying down. There was also a dancing exercise, which Adna and Al two joined. Azia enjoyed watching Al two and, since Lex wasn’t with them, took a video of him dancing with others, including Adna. They stayed there for two hours and then had lunch. Lex, who had been left at home, was called by Adna but refused to go to the mall because she was busy with paperwork, so Azia and Al two went home to be with their sister. Al and Adna stayed at the mall, did some window shopping, and ate again because Adna likes to try small dishes at the food court. Afterwards they went to another mall to look for materials Al two needed for his grape plants. When they couldn’t find them, they went to a different mall. Adna remembered how Al’s mother used to go with the family from one mall to another. Al's younger sister married a foreigner and moved to his country. She petitioned for their mother, who then immigrated to live with them. Arrangements were made for her medicines to be sent in packages; Adna was responsible for sending them. The mother had diabetes and hypertension and needed daily medication. Since both the younger sister and her husband were too busy with work, she was often alone in their house. It took the couple a long time to have children. Al's mother, who had been a high school teacher before retiring, was used to having many students; when she was alone she missed her native land. When Al's older sister was on a business trip to the country where their mother had migrated, the mother told her, "I'm going home with you." The younger sister did not expect this decision, as she knew their mother favored her because she was the youngest. The younger sister could not stop her mother, and she had no influence over their older sister, who was domineering. The older sister bought a plane ticket for their mother, and they flew home immediately. At a regular checkup with the family doctor it was found that she already had complications. Every time she was hospitalized, Adna took care of her. Her memory began to fail, but she always remembered Adna's name. Once, a male nurse brought medicine to Al's mother, but she refused to take it. "I want only Adna—she is the only one I trust," she told the nurse. When Adna, who had just returned from the hospital pharmacy, came back, the woman said, "Your smile is so familiar; it looks like Adna's." Adna replied, "It's me, Adna!" The woman laughed, "Oh good, you're here. A man came and tried to force something on me—I thought it was poison. I told him I only trust you." The male nurse came in again and asked, "Who is Adna?" "I am," Adna replied. The nurse handed the tablet to Adna and said, "She only trusts you." Adna received the tablet and gave it to her mother-in-law. It was easy for Adna to get her mother-in-law to take the tablet. The male nurse was amazed at what he had witnessed. After Al's mother was discharged from the hospital, Al's younger sister was very grateful to Adna. She asked Adna to accompany her mother to the beauty salon to fix her hair and have it colored regularly. That Christmas, she sent money so they could go mall-hopping with the children Lex, Azia, and Al the Second.
long_en_376	poet_en	1,027	en	Last night The Boss Lady and I were treated to a night out. One of The Boss Lady's childhood friends has moved into our neighborhood. She watched the kids for us and sent us to one of her favorite restaurants, The Melting Pot. She also handed us the keys to her car, a really nice Lexus. After gorging on a Cheddar and Emmentaler Swiss cheese veggie fondue appetizer and the Center Cut Entree (Filet Mignon, Black Tiger Shrimp, Chicken Breast, Peppercorn Pork Tenderloin, and Teriyaki Sirloin) in a Coq au Vin fondue, we decided on the Flaming Turtle Chocolate Fondue for dessert. They should serve the dessert fondue with a chaser of insulin. I thought I might turn diabetic after eating chocolate-dipped marshmallows that had been coated in crushed Oreos. But I have never been one to waste food, so I ate them all. The Boss Lady was a little more health conscious. She dipped fruit and cheesecake in the chocolate and caramel. There was a chef's salad in there somewhere before the Flaming Turtle, but as Homer Simpson said, "You don't win friends with salad," so who cares. Everything was awesome, even the salad. But if you go to The Melting Pot, don't forget to sign the loan documents at your bank beforehand. And take me—I'll teach you how to really get those marshmallows covered. After dinner we were really creative. We drove out to the construction site of a school that The Boss Lady would like to run next year. Then we drove home. But since we were in a nice car, we did take the long way home. The kids had a ball. They ate at Chick-fil-A and went to a carnival at church. They were even in bed when we got home! What a nice night. I still don't know exactly what we did to deserve it, but thanks, DamnYell. The first words out of The Talker's mouth this morning were "Daddy, I wanna play Play-Doh." No real stress for Mommy there. The second round of talking would have caused problems for Mommy, though. "I was dreaming about babies. Mommy has a new baby." Dang, now I feel guilty for getting a great night's sleep. All while The Boss Lady was off birthing a baby... And finally, why is AtHomeDaddy always the last to know? It is pretty amazing what one can do with a helpful three-year-old, some swim noodles, and Daddy's work clothes. Who cares if he looks as if he has scoliosis and bow legs? The Talker is very proud of his creation. The Princess deserves some credit, too. She slept late yesterday, making this whole thing possible. But not as late as she slept today. It is 9:10 and the girl is still out. Yeah for Daddy! Just for the record, 9:50. One hour and 50 minutes later than usual. This morning, after running some errands, the kids and I headed over for some pre-lunch park time. The Talker recognized a girl that he has only played with once. I couldn't remember her name, but the boy has great recall of names. Last time she was at the park with her grandmother; this time she was with her other granny. Today's grandma asked if I was "the stay-at-home dad her daughter told her about." Since Other Dad lives right across the street from her daughter, I assumed she was mistaking me for him. Then she explained that she had already met him. I was even more confused by her comment, but The Princess fell hard and cried, so I never got around to asking what she meant. Plus, the kids were having fun and she wasn't rushing out of the park with her granddaughter, so I didn't worry about it. Later, though, she hit me with a doozy. The Princess was trying to play with the buckles on the little girl's stroller. She had also figured out where the snacks were stored—under the stroller—so my human grackle was not moving, not until someone shared that snack. Granny told me that she is a retired gifted-and-talented high school teacher, and that it is "obvious that your girl will need gifted services in school, especially if you do a good job getting her ready for school." What? Huh? Who the heck are you? You're such a great high school teacher that you can assign labels to 18-month-olds? Pretty pretentious, don't you think, lady? I'm glad those smart kids in Dallas got to benefit from your insight and wisdom. Bitch. I have no doubt that my kids will be smarter than me; it would be hard for them not to be. But this is the same as someone telling me the boy is a good candidate for a Ritalin and Focalin cocktail every morning with his breakfast. I have never held gifted/talented teachers in very high regard. Probably goes back to being left with the other kids while the geniuses benefited from the best learning methods. Or maybe I just think kids need to be educated with their peers. After we shopped, we ate lunch at the snack bar: cheese pizza for the kids to share and a hot dog for Dad. Then we all scarfed down ice cream — a good outing and an easy lunch. While we were eating the snack tables were full, so we ended up sitting at the last table, big enough for eight people. Three strangers joined us, and did they ever get an earful from the boy. After he introduced everyone, he let them know exactly what our plans were for the day. He is not a shy child; a shy child would not include bathroom breaks when sharing an itinerary with complete strangers. By the time the others left, they knew that The Princess had blown out a diaper in the car and that our dog was in trouble for eating a tree. This afternoon was our neighborhood costume parade.	Last night The Boss Lady and I were treated to a night out. One of The Boss Lady's childhood friends has moved into our neighborhood. She watched the kids for us and sent us to one of her favorite restaurants, The Melting Pot. She also handed us the keys to her car, a really nice Lexus. After gorging on a Cheddar and Emmentaler Swiss cheese veggie fondue appetizer and the Center Cut Entree (Filet Mignon, Black Tiger Shrimp, Chicken Breast, Peppercorn Pork Tenderloin, and Teriyaki Sirloin) in a Coq au Vin fondue, we decided on the Flaming Turtle Chocolate Fondue for dessert. They should serve the dessert fondue with a chaser of insulin. I thought I might turn diabetic after eating chocolate-dipped marshmallows that had been coated in crushed Oreos. But I have never been one to waste food, so I ate them all. The Boss Lady was a little more health conscious. She dipped fruit and cheesecake in the chocolate and caramel. There was a chef's salad in there somewhere before the Flaming Turtle, but as Homer Simpson said, "You don't win friends with salad," so who cares. Everything was awesome, even the salad. But if you go to The Melting Pot, don't forget to sign the loan documents at your bank beforehand. And take me—I'll teach you how to really get those marshmallows covered. After dinner we were really creative. We drove out to the construction site of a school that The Boss Lady would like to run next year. Then we drove home. But since we were in a nice car, we did take the long way home. The kids had a ball. They ate at Chick-fil-A and went to a carnival at church. They were even in bed when we got home! What a nice night. I still don't know exactly what we did to deserve it, but thanks, DamnYell. The first words out of The Talker's mouth this morning were "Daddy, I wanna play Play-Doh." No real stress for Mommy there. The second round of talking would have caused problems for Mommy, though. "I was dreaming about babies. Mommy has a new baby." Dang, now I feel guilty for getting a great night's sleep. All while The Boss Lady was off birthing a baby... And finally, why is AtHomeDaddy always the last to know? It is pretty amazing what one can do with a helpful three-year-old, some swim noodles, and Daddy's work clothes. Who cares if he looks as if he has scoliosis and bow legs? The Talker is very proud of his creation. The Princess deserves some credit, too. She slept late yesterday, making this whole thing possible. But not as late as she slept today. It is nine colon ten and the girl is still out. Yeah for Daddy! Just for the record, nine colon fifty. One hour and fifty minutes later than usual. This morning, after running some errands, the kids and I headed over for some pre-lunch park time. The Talker recognized a girl that he has only played with once. I couldn't remember her name, but the boy has great recall of names. Last time she was at the park with her grandmother; this time she was with her other granny. Today's grandma asked if I was "the stay-at-home dad her daughter told her about." Since Other Dad lives right across the street from her daughter, I assumed she was mistaking me for him. Then she explained that she had already met him. I was even more confused by her comment, but The Princess fell hard and cried, so I never got around to asking what she meant. Plus, the kids were having fun and she wasn't rushing out of the park with her granddaughter, so I didn't worry about it. Later, though, she hit me with a doozy. The Princess was trying to play with the buckles on the little girl's stroller. She had also figured out where the snacks were stored—under the stroller—so my human grackle was not moving, not until someone shared that snack. Granny told me that she is a retired gifted-and-talented high school teacher, and that it is "obvious that your girl will need gifted services in school, especially if you do a good job getting her ready for school." What? Huh? Who the heck are you? You're such a great high school teacher that you can assign labels to eighteen-month-olds? Pretty pretentious, don't you think, lady? I'm glad those smart kids in Dallas got to benefit from your insight and wisdom. Bitch. I have no doubt that my kids will be smarter than me; it would be hard for them not to be. But this is the same as someone telling me the boy is a good candidate for a Ritalin and Focalin cocktail every morning with his breakfast. I have never held gifted/talented teachers in very high regard. Probably goes back to being left with the other kids while the geniuses benefited from the best learning methods. Or maybe I just think kids need to be educated with their peers. After we shopped, we ate lunch at the snack bar: cheese pizza for the kids to share and a hot dog for Dad. Then we all scarfed down ice cream — a good outing and an easy lunch. While we were eating the snack tables were full, so we ended up sitting at the last table, big enough for eight people. Three strangers joined us, and did they ever get an earful from the boy. After he introduced everyone, he let them know exactly what our plans were for the day. He is not a shy child; a shy child would not include bathroom breaks when sharing an itinerary with complete strangers. By the time the others left, they knew that The Princess had blown out a diaper in the car and that our dog was in trouble for eating a tree. This afternoon was our neighborhood costume parade.
long_en_377	poet_en	982	en	The serpent drops Fenrir off on the beach near the others before collapsing. Fenrir, still confused after being carried by her — and it seems the others are too — gets straight to the point. "Can you heal her, Cass?" "Sure," Cassiel says, cautiously approaching the serpent. It's the first time she's been this close to a monster that size that doesn't want to kill her, and the first time she's tried healing a non-player who isn't Rock. The serpent shifts as Cassiel stands beside her, fixing Cassiel with one eye larger than Cassiel's head. Cassiel looks back at Fenrir. "She's safe," he says. "O-okay," Cassiel replies, taking a deep breath and raising her hands. Mystical, golden light flows from her palms into the serpent, which relaxes and closes her eyes at the touch. "Hey there, princess," a familiar voice calls from behind Fenrir. He turns and sees the same odd man as before — only now he's trying to flirt with Oleander. The man hops off his mount and kneels before her. "Would you like to fly with me? I could show you the world on my trusty steed," he proposes, taking Oleander's hand. Fenrir cringes, but when he watches Oleander's reaction he can see the sadistic gears turning in her head. “You’re so sweet!” Oleander says, raising his hands over his heart and speaking in an extra-feminine tone. “And you’re such a cutie, too.” While those two flirt, Fenrir notices something else making the day even crazier: Shogun and Rock are staring at each other—really staring, and looking far too interested. Rock’s tail is wagging, and even Shogun’s feathery tail twitches. “They grow up so fast, Onii-chan,” Saya says. “I am not letting my dog hook up with some oversized fox,” Fenrir thinks. “She’s like a daughter to me!” “So you’re telling me that you not only tongued your own dog, but you did that with a dog you see as a daughter?” Cassiel asks. “I’m going to uninstall you,” Fenrir replies. “You’d have to uninstall your brain to do that, silly Onii-chan,” Cassiel says. Shogun walks up to Rock, rolls onto his back in front of her, and Rock closes the distance, sniffs him all over, then licks his face. “I’ve healed her as much as I can,” Cassiel says. “Who’s that guy talking to Olly? And why do you… look so sad?” she asks, seeing Fenrir’s face. Fenrir points sadly at Rock and Shogun, who are happily wrestling. Despite being much larger and looking more powerful, Shogun lets Rock win their little fights and is always the first to submit to her victory. “That’s… really cute,” Cassiel says. Fenrir grabs Cassiel’s shoulders and shakes her. “This is basically watching some random guy I don't know come up to my daughter and try to seduce her! And it's working! Rock is totally falling for him!” he explains. “I'm tired. Stop shaking me,” Cassiel says, brushing his hands off her shoulders. “Look, they're happy. Isn't that all that matters? Don't tell me you're one of those guys who are overly protective of their kids and try to stop them from dating. That's what my parents did, and it just made me hook up with a bunch of douchebags instead of the respectable ones—those were the only people who would date me behind my parents' backs. Do you really want Rock going behind your back with animals you don't know? At least that fox isn't trying to get her high or pimp her.” “You... had a rough time dating before, didn't you?” Fenrir asks, hearing her rant and seeing how upset she looks remembering those memories. “Our kids are allowed to date as soon as they're old enough. You're already planning on us having kids?” “Sh-shut up, bastard. Of course I am. Don't point out the obvious.” “All right, all right. I promise I won't be overly protective, but I'll still want to meet all of their boyfriends and girlfriends. Hey, wait—I haven't met your parents yet and have never even talked to them. Do they even know I exist?” Fenrir asks. Cassiel looks away and tries to whistle but fails. “Cass. Are you dating me behind your parents' backs?” he asks again. Her attempt at whistling speeds up and grows louder. "You really suck at whistling," he said. Cassiel turned to face him, ready to shout, but he surprised her with a kiss. A startled moan turned into an angry shout when he pulled away. "D-don't scare me like that, you perverted, overprotective dog!" she snapped. "Looks like you've got a troublemaker there, bro!" a stranger called from near Oleander. Cassiel glared at the man and drew her sword. "Who are you calling trouble, asshole?" "Whoa — you actually called someone something other than 'bastard,'" Fenrir teased. "Th‑that's because I don't want to call him what I call you!" she stammered. "Ooh, she's hot and cold. Never been a fan — too much drama," the other man said, prompting Cassiel to point her sword at him. "Who even are you?" "I am this fair maiden's knight! I shall be whoever she wants me to be, and I will sweep her off her feet and show her the world!" he declared. Cassiel glanced at Oleander. He looked like he was trying hard not to laugh as the man faced away. "You know Olly's a guy, right?" she asked. Oleander's eyes went wide. He pouted, shook his head at Cassiel, and raised a finger to his lips, then returned to a neutral expression when the stranger turned back.	The serpent drops Fenrir off on the beach near the others before collapsing. Fenrir, still confused after being carried by her — and it seems the others are too — gets straight to the point. "Can you heal her, Cass?" "Sure," Cassiel says, cautiously approaching the serpent. It's the first time she's been this close to a monster that size that doesn't want to kill her, and the first time she's tried healing a non-player who isn't Rock. The serpent shifts as Cassiel stands beside her, fixing Cassiel with one eye larger than Cassiel's head. Cassiel looks back at Fenrir. "She's safe," he says. "O-okay," Cassiel replies, taking a deep breath and raising her hands. Mystical, golden light flows from her palms into the serpent, which relaxes and closes her eyes at the touch. "Hey there, princess," a familiar voice calls from behind Fenrir. He turns and sees the same odd man as before — only now he's trying to flirt with Oleander. The man hops off his mount and kneels before her. "Would you like to fly with me? I could show you the world on my trusty steed," he proposes, taking Oleander's hand. Fenrir cringes, but when he watches Oleander's reaction he can see the sadistic gears turning in her head. “You’re so sweet!” Oleander says, raising his hands over his heart and speaking in an extra-feminine tone. “And you’re such a cutie, too.” While those two flirt, Fenrir notices something else making the day even crazier: Shogun and Rock are staring at each other—really staring, and looking far too interested. Rock’s tail is wagging, and even Shogun’s feathery tail twitches. “They grow up so fast, Onii-chan,” Saya says. “I am not letting my dog hook up with some oversized fox,” Fenrir thinks. “She’s like a daughter to me!” “So you’re telling me that you not only tongued your own dog, but you did that with a dog you see as a daughter?” Cassiel asks. “I’m going to uninstall you,” Fenrir replies. “You’d have to uninstall your brain to do that, silly Onii-chan,” Cassiel says. Shogun walks up to Rock, rolls onto his back in front of her, and Rock closes the distance, sniffs him all over, then licks his face. “I’ve healed her as much as I can,” Cassiel says. “Who’s that guy talking to Olly? And why do you… look so sad?” she asks, seeing Fenrir’s face. Fenrir points sadly at Rock and Shogun, who are happily wrestling. Despite being much larger and looking more powerful, Shogun lets Rock win their little fights and is always the first to submit to her victory. “That’s… really cute,” Cassiel says. Fenrir grabs Cassiel’s shoulders and shakes her. “This is basically watching some random guy I don't know come up to my daughter and try to seduce her! And it's working! Rock is totally falling for him!” he explains. “I'm tired. Stop shaking me,” Cassiel says, brushing his hands off her shoulders. “Look, they're happy. Isn't that all that matters? Don't tell me you're one of those guys who are overly protective of their kids and try to stop them from dating. That's what my parents did, and it just made me hook up with a bunch of douchebags instead of the respectable ones—those were the only people who would date me behind my parents' backs. Do you really want Rock going behind your back with animals you don't know? At least that fox isn't trying to get her high or pimp her.” “You... had a rough time dating before, didn't you?” Fenrir asks, hearing her rant and seeing how upset she looks remembering those memories. “Our kids are allowed to date as soon as they're old enough. You're already planning on us having kids?” “Sh-shut up, bastard. Of course I am. Don't point out the obvious.” “All right, all right. I promise I won't be overly protective, but I'll still want to meet all of their boyfriends and girlfriends. Hey, wait—I haven't met your parents yet and have never even talked to them. Do they even know I exist?” Fenrir asks. Cassiel looks away and tries to whistle but fails. “Cass. Are you dating me behind your parents' backs?” he asks again. Her attempt at whistling speeds up and grows louder. "You really suck at whistling," he said. Cassiel turned to face him, ready to shout, but he surprised her with a kiss. A startled moan turned into an angry shout when he pulled away. "D-don't scare me like that, you perverted, overprotective dog!" she snapped. "Looks like you've got a troublemaker there, bro!" a stranger called from near Oleander. Cassiel glared at the man and drew her sword. "Who are you calling trouble, asshole?" "Whoa — you actually called someone something other than 'bastard,'" Fenrir teased. "Th‑that's because I don't want to call him what I call you!" she stammered. "Ooh, she's hot and cold. Never been a fan — too much drama," the other man said, prompting Cassiel to point her sword at him. "Who even are you?" "I am this fair maiden's knight! I shall be whoever she wants me to be, and I will sweep her off her feet and show her the world!" he declared. Cassiel glanced at Oleander. He looked like he was trying hard not to laugh as the man faced away. "You know Olly's a guy, right?" she asked. Oleander's eyes went wide. He pouted, shook his head at Cassiel, and raised a finger to his lips, then returned to a neutral expression when the stranger turned back.
long_en_278	wiki_en	829	en	Organonickel chemistry is a branch of organometallic chemistry that deals with organic compounds featuring nickel-carbon bonds. Organonickel compounds are used as catalysts, as building blocks in organic chemistry, and in chemical vapor deposition. They are also short-lived intermediates in organic reactions. The first organonickel compound, nickel tetracarbonyl Ni(CO)4, was reported in 1890 and was quickly applied in the Mond process for nickel purification. Organonickel complexes are prominent in numerous industrial processes, including carbonylations, hydrocyanation, and the Shell Higher Olefin Process. Classes of compounds Alkyl and aryl complexes A popular reagent is Ni(CH3)2(tetramethylethylenediamine) (NiMe2(TMEDA)). Many alkyl and aryl complexes are known with the formula NiR(X)L2. Examples include (dppf)Ni(cinnamyl)Cl, trans-(PCy2Ph)2Ni(o-tolyl)Cl, (dppf)Ni(o-tolyl)Cl, (TMEDA)Ni(o-tolyl)Cl, and (TMEDA)NiMe2. Nickel compounds of the type NiR2 also exist with just 12 valence electrons. In solution, however, solvents always interact with the metal atom, increasing the electron count. One 12-electron compound is di(mesityl)nickel, prepared from (allyl)2Ni2Br2 and the corresponding Grignard reagent: (allyl)2Ni2Br2 + 4 C6H2Me3MgBr → 2 (allyl)MgBr + 2 MgBr2 + 2 (C6H2Me3)2Ni Alkene complexes Many complexes exist of nickel coordinated to an alkene. Practical applications of this theme include polymerization or oligomerization of alkenes, as in the Shell Higher Olefin Process. In these compounds nickel is formally zerovalent, Ni(0), and the bonding is described by the Dewar–Chatt–Duncanson model. A common representative is bis(cyclooctadiene)nickel(0) (Ni(COD)2), which contains two cyclooctadiene ligands. It is an 18-electron (18e-) compound, with 10 electrons provided by nickel and 4×2 electrons from the double bonds. This solid, which melts at 60 °C, is used as a catalyst and as a precursor for many other nickel compounds. Allyl complexes Nickel forms several simple allyl complexes. Allyl halides react with Ni(CO)4 to form π-allyl complexes such as (allyl)2Ni2Cl2; these compounds are sources of allyl nucleophiles. In (allyl)2Ni2Br2 and (allyl)Ni(C5H5), nickel is assigned oxidation number +2, and the electron counts are 16 and 18, respectively. Bis(allyl)nickel is prepared from allylmagnesium bromide and nickel chloride. Cyclopentadienyl complexes Nickelocene, NiCp2, contains Ni in the +2 oxidation state and is a 20-valence-electron metallocene. It can be oxidized by one electron. The corresponding palladocene and platinocene are unknown. From nickelocene many derivatives are generated, e.g. CpNiLCl, CpNiNO, and Cp2Ni2(CO)3. Carbene complexes Nickel forms carbene complexes that formally feature C=Ni double bonds. Reactions — alkene/alkyne oligomerizations Nickel compounds catalyze the oligomerization of alkenes and alkynes, a property that helped validate the research and development of Ziegler–Natta catalysts in the 1950s. The discovery that nickel impurities from an autoclave suppressed the propagation (Aufbau) step and promoted termination to produce a terminal alkene was striking: the polymerization of ethylene suddenly stopped at 1‑butene. This "nickel effect" prompted the search for other catalysts and ultimately led to systems capable of producing high‑molar‑mass polymers, such as modern Ziegler–Natta catalysts. One practical implementation of alkyne oligomerization is the Reppe synthesis, for example in the formation of cyclooctatetraene, a formal [2+2+2+2] cycloaddition. Oligomerization of butadiene with ethylene to give trans‑1,4‑hexadiene was once an industrial process. Formal [2+2+2] cycloadditions also occur in alkyne trimerization, which can be extended to include benzyne: benzyne is generated in situ from an ortho‑trimethylsilyl aryl triflate and reacts with a diyne such as 1,7‑octadiyne in the presence of a NiBr2·bis(diphenylphosphino)ethane/Zn catalyst system to give the corresponding naphthalene derivative. In the catalytic cycle, elemental zinc reduces Ni(II) to Ni(0), which can then coordinate two alkyne bonds. A cyclometalation step converts the nickel cyclopentadiene intermediate into a species that coordinates benzyne, which undergoes C–H insertion to give a nickel cycloheptatriene compound. Reductive elimination then liberates a tetrahydroanthracene product. The formation of organonickel intermediates in this type of reaction is not always obvious, but in a carefully designed experiment two such intermediates were formed quantitatively. One study noted that the reaction only works with acetylene or simple alkynes because more substituted alkynes give poor regioselectivity. From a terminal alkyne, seven isomers are possible, differing in the positions of substituents or double bonds. One strategy to remedy this problem is to employ certain diynes; the selected reaction conditions also minimize formation of the competing [2+2+2] cycloaddition product, which gives the corresponding substituted arene. Nickel compounds promote coupling reactions between allyl and aryl halides. Other nickel-catalyzed coupling reactions include the Kumada and Negishi couplings. Nickel catalyzes the addition of carbon monoxide to alkenes and alkynes. Industrially, acrylic acid was once produced by reacting acetylene, carbon monoxide, and water at 40–55 atm and 160–200 °C in the presence of nickel(II) bromide and a copper halide. See also: nickel(IV) organometallic complexes; nickel(II) precatalysts; lactate racemase.	Organonickel chemistry is a branch of organometallic chemistry that deals with organic compounds featuring nickel-carbon bonds. Organonickel compounds are used as catalysts, as building blocks in organic chemistry, and in chemical vapor deposition. They are also short-lived intermediates in organic reactions. The first organonickel compound, nickel tetracarbonyl Ni(CO) four, was reported in eighteen ninety and was quickly applied in the Mond process for nickel purification. Organonickel complexes are prominent in numerous industrial processes, including carbonylations, hydrocyanation, and the Shell Higher Olefin Process. Classes of compounds Alkyl and aryl complexes A popular reagent is Ni(CH three) two (tetramethylethylenediamine) (NiMe two (TMEDA)). Many alkyl and aryl complexes are known with the formula NiR(X)L two. Examples include (dppf)Ni(cinnamyl)Cl, trans-(PCy two Ph) two Ni(o-tolyl)Cl, (dppf)Ni(o-tolyl)Cl, (TMEDA)Ni(o-tolyl)Cl, and (TMEDA)NiMe two. Nickel compounds of the type NiR two also exist with just twelve valence electrons. In solution, however, solvents always interact with the metal atom, increasing the electron count. One twelve-electron compound is di(mesityl)nickel, prepared from (allyl) two Ni two Br two plus four C six H two Me three MgBr yields two (allyl) MgBr plus two MgBr two plus two (C six H two Me three) two Ni Alkene complexes Many complexes exist of nickel coordinated to an alkene. Practical applications of this theme include polymerization or oligomerization of alkenes, as in the Shell Higher Olefin Process. In these compounds nickel is formally zerovalent, Ni(zero), and the bonding is described by the Dewar–Chatt–Duncanson model. A common representative is bis(cyclooctadiene)nickel(zero) (Ni(COD two)), which contains two cyclooctadiene ligands. It is an eighteen-electron (eighteen e-) compound, with ten electrons provided by nickel and four times two electrons from the double bonds. This solid, which melts at sixty degrees C, is used as a catalyst and as a precursor for many other nickel compounds. Allyl complexes Nickel forms several simple allyl complexes. Allyl halides react with Ni(CO four) to form pi-allyl complexes such as (allyl) two Ni two Cl two; these compounds are sources of allyl nucleophiles. In (allyl) two Ni two Br two and (allyl)Ni(C five H five), nickel is assigned oxidation number plus two, and the electron counts are sixteen and eighteen, respectively. Bis(allyl)nickel is prepared from allylmagnesium bromide and nickel chloride. Cyclopentadienyl complexes Nickelocene, Ni Cp two, contains Ni in the plus two oxidation state and is a twenty-valence-electron metallocene. It can be oxidized by one electron. The corresponding palladocene and platinocene are unknown. From nickelocene many derivatives are generated, e.g. CpNiLCl, CpNiNO, and Cp two Ni two (CO) three. Carbene complexes Nickel forms carbene complexes that formally feature C equals Ni double bonds. Reactions — alkene/alkyne oligomerizations Nickel compounds catalyze the oligomerization of alkenes and alkynes, a property that helped validate the research and development of Ziegler–Natta catalysts in the nineteen fifties. The discovery that nickel impurities from an autoclave suppressed the propagation (Aufbau) step and promoted termination to produce a terminal alkene was striking: the polymerization of ethylene suddenly stopped at one‑butene. This "nickel effect" prompted the search for other catalysts and ultimately led to systems capable of producing high‑molar‑mass polymers, such as modern Ziegler–Natta catalysts. One practical implementation of alkyne oligomerization is the Reppe synthesis, for example in the formation of cyclooctatetraene, a formal [two plus two plus two plus two] cycloaddition. Oligomerization of butadiene with ethylene to give trans‑one, four‑hexadiene was once an industrial process. Formal [two plus two plus two] cycloadditions also occur in alkyne trimerization, which can be extended to include benzyne: benzyne is generated in situ from an ortho‑trimethylsilyl aryl triflate and reacts with a diyne such as one, seven‑octadiyne in the presence of a NiBr two·bis(diphenylphosphino)ethane/Zn catalyst system to give the corresponding naphthalene derivative. In the catalytic cycle, elemental zinc reduces Ni(two) to Ni(zero), which can then coordinate two alkyne bonds. A cyclometalation step converts the nickel cyclopentadiene intermediate into a species that coordinates benzyne, which undergoes C H insertion to give a nickel cycloheptatriene compound. Reductive elimination then liberates a tetrahydroanthracene product. The formation of organonickel intermediates in this type of reaction is not always obvious, but in a carefully designed experiment two such intermediates were formed quantitatively. One study noted that the reaction only works with acetylene or simple alkynes because more substituted alkynes give poor regioselectivity. From a terminal alkyne, seven isomers are possible, differing in the positions of substituents or double bonds. One strategy to remedy this problem is to employ certain diynes; the selected reaction conditions also minimize formation of the competing two plus two plus two cycloaddition product, which gives the corresponding substituted arene. Nickel compounds promote coupling reactions between allyl and aryl halides. Other nickel-catalyzed coupling reactions include the Kumada and Negishi couplings. Nickel catalyzes the addition of carbon monoxide to alkenes and alkynes. Industrially, acrylic acid was once produced by reacting acetylene, carbon monoxide, and water at forty to fifty five atmospheres and one hundred sixty to two hundred degrees Celsius in the presence of nickel two bromide and a copper halide. See also: nickel four organometallic complexes; nickel two precatalysts; lactate racemase.
long_en_372	poet_en	995	en	Andrew from the hotel wrote down the name in Arabic for us to give to the taxi. All seemed to go well, and we were dropped off next to a gallery — just not the one we were after. Oh well; we were here, so we might as well have a look. Inside, a few rooms were devoted solely to photography. The photos were by an artist born in Jerusalem who had been taking photographs since the 1930s. He also owned a collection of photographs going back to 1890. There was a combination of these on show: mostly photos of people and everyday life back then. There were some amazing pictures of people in traditional dress and cityscapes very different from what we had seen on our trip. It made us wish we had taken more photos of people. After visiting this gallery, we had to find out how to get to the one we had intended to go to. We could see a few landmarks, as we were on one of the many hills of Amman, but the street name was not on our small map. We wandered up a bit, but decided we might as well walk all the way down, find out where we were, and go from there. A couple of hundred meters down the road we found the gallery — no idea why the taxi had not dropped us off there in the first place. We started in a small room called "The Lab." There were some photos made with a very old technique that used water reflections. They were well done but a bit depressing and not particularly interesting, to be honest. Off to the side, in another room, was a video installation. A woman filmed her walk from home to work for a few days in a row, across one of the many checkpoints the Israelis have set up in the West Bank. She had to walk about 2 km each way because cars aren't allowed on this stretch of road. Of course she wasn't allowed to record, so we mostly got shots of her feet. It was filmed through a hole in her bag; the first time she tried, the police confiscated the footage and threw her American passport in the mud. That footage wasn't very interesting either—mostly shots of her walking past an armored personnel carrier (APC) and other people making their way to and from the West Bank. In the blurb she mentions that the Israelis can close off the road whenever they want, firing tear gas and live rounds into the street. If we had stayed to watch the whole thing there might have been something in it, but the two days we sat through weren't that interesting—just people walking. From there we had to find the rest of the exhibitions. A couple of other buildings are further down the hill. This is the main section, and out front there is a Byzantine church. Some of the pillars have been restored, and there is a small alcove where the altar was. Andrew tried to conduct a mass there for his new church, but no one was interested in attending; even Anna walked away when he started preaching. The layout of this museum is very confusing. While walking around, we found what we thought was the main entrance, but it was locked, so we went up the other side. There was a small seating area with a fountain and trees providing shade—it was the perfect place to relax. We got a couple of freshly squeezed juices, which topped off the experience. However, we couldn't sit there all day; we still wanted to see what the museum had to offer. Inside another building was an exhibition by an amateur Palestinian artist criticizing Israel and the USA. The works were 3D reliefs made with glue and sawdust—pretty good, if gruesome. He had witnessed a massacre of more than 100 people, and all his work reflected this. One piece depicted a statue of Saddam Hussein being overthrown by American cockroaches; it was very well done. Another section consisted of coloured lines on canvas with sand glued on to give texture. That was about all we could find in the gallery. We were sure we had missed a lot, but it was confusing to tell which buildings were administrative and which were galleries. Back at the hotel, we decided to blog some more and try to catch up. After a while we were ready for dinner. Although we were both a little sick of hummus and falafel, we had promised Abu Hamsa the previous evening that we would go see him at Hashim's. He was there and glad to see us. He even took a bit of time off to come talk to us, but as it was busy (it always is, 24/7) he had to go back to work. We ended up ordering every dish they had: ful (beans), hummus, falafel, and feta. The feta was brilliant, and we hadn't had any in Jordan yet. It was a filling meal as always, and comes highly recommended to anyone visiting Jordan. It's also cheap — 4 JD for more than we could eat. The 16th was supposed to be a day of shopping. Anna wanted some new clothes. Unfortunately, the clothing sold here either covers a woman from ankle to wrist or makes her walk around half naked; there doesn't seem to be an in-between. Anna gave up pretty quickly. To be fair, we only tried the streets in downtown Amman, not the big malls elsewhere. We spent the afternoon drinking several Arabic coffees and smoking a shisha (nargileh) in an all-male cafe. We had one final falafel sandwich for lunch. We'd had many falafel sandwiches, and by then we were happy to leave them behind. We met up with Sofia and David for breakfast.	Andrew from the hotel wrote down the name in Arabic for us to give to the taxi. All seemed to go well, and we were dropped off next to a gallery — just not the one we were after. Oh well; we were here, so we might as well have a look. Inside, a few rooms were devoted solely to photography. The photos were by an artist born in Jerusalem who had been taking photographs since the nineteen thirties. He also owned a collection of photographs going back to eighteen ninety. There was a combination of these on show: mostly photos of people and everyday life back then. There were some amazing pictures of people in traditional dress and cityscapes very different from what we had seen on our trip. It made us wish we had taken more photos of people. After visiting this gallery, we had to find out how to get to the one we had intended to go to. We could see a few landmarks, as we were on one of the many hills of Amman, but the street name was not on our small map. We wandered up a bit, but decided we might as well walk all the way down, find out where we were, and go from there. A couple of hundred meters down the road we found the gallery — no idea why the taxi had not dropped us off there in the first place. We started in a small room called "The Lab." There were some photos made with a very old technique that used water reflections. They were well done but a bit depressing and not particularly interesting, to be honest. Off to the side, in another room, was a video installation. A woman filmed her walk from home to work for a few days in a row, across one of the many checkpoints the Israelis have set up in the West Bank. She had to walk about two kilometers each way because cars aren't allowed on this stretch of road. Of course she wasn't allowed to record, so we mostly got shots of her feet. It was filmed through a hole in her bag; the first time she tried, the police confiscated the footage and threw her American passport in the mud. That footage wasn't very interesting either—mostly shots of her walking past an armored personnel carrier (APC) and other people making their way to and from the West Bank. In the blurb she mentions that the Israelis can close off the road whenever they want, firing tear gas and live rounds into the street. If we had stayed to watch the whole thing there might have been something in it, but the two days we sat through weren't that interesting—just people walking. From there we had to find the rest of the exhibitions. A couple of other buildings are further down the hill. This is the main section, and out front there is a Byzantine church. Some of the pillars have been restored, and there is a small alcove where the altar was. Andrew tried to conduct a mass there for his new church, but no one was interested in attending; even Anna walked away when he started preaching. The layout of this museum is very confusing. While walking around, we found what we thought was the main entrance, but it was locked, so we went up the other side. There was a small seating area with a fountain and trees providing shade—it was the perfect place to relax. We got a couple of freshly squeezed juices, which topped off the experience. However, we couldn't sit there all day; we still wanted to see what the museum had to offer. Inside another building was an exhibition by an amateur Palestinian artist criticizing Israel and the USA. The works were three D reliefs made with glue and sawdust—pretty good, if gruesome. He had witnessed a massacre of more than one hundred people, and all his work reflected this. One piece depicted a statue of Saddam Hussein being overthrown by American cockroaches; it was very well done. Another section consisted of coloured lines on canvas with sand glued on to give texture. That was about all we could find in the gallery. We were sure we had missed a lot, but it was confusing to tell which buildings were administrative and which were galleries. Back at the hotel, we decided to blog some more and try to catch up. After a while we were ready for dinner. Although we were both a little sick of hummus and falafel, we had promised Abu Hamsa the previous evening that we would go see him at Hashim's. He was there and glad to see us. He even took a bit of time off to come talk to us, but as it was busy (it always is, twenty four slash seven) he had to go back to work. We ended up ordering every dish they had: ful (beans), hummus, falafel, and feta. The feta was brilliant, and we hadn't had any in Jordan yet. It was a filling meal as always, and comes highly recommended to anyone visiting Jordan. It's also cheap — four JD for more than we could eat. The sixteenth was supposed to be a day of shopping. Anna wanted some new clothes. Unfortunately, the clothing sold here either covers a woman from ankle to wrist or makes her walk around half naked; there doesn't seem to be an in-between. Anna gave up pretty quickly. To be fair, we only tried the streets in downtown Amman, not the big malls elsewhere. We spent the afternoon drinking several Arabic coffees and smoking a shisha (nargileh) in an all-male cafe. We had one final falafel sandwich for lunch. We'd had many falafel sandwiches, and by then we were happy to leave them behind. We met up with Sofia and David for breakfast.
long_en_353	poet_en	969	en	"Isn't—isn't this a bit overkill? I mean... is it that important?" Ryouta asks. Cassandra, standing behind him with her arms locked underneath his own and holding him back, asks, "You don't want to hide anything from us, right?" "This is good research material. Great idea, Serra!" Aiko says, patting Serra on the top of her head. Even Alice is in the room now, sitting in her wheelchair next to Serra. Ryouta is going to have to start deleting his history. "Alright, I won't try to stop you. It's embarrassing, but there isn't anything too bad in there. I think," Ryouta says. "Promise?" Cassandra asks. "I promise." With that, Cassandra lets go of him and they both walk over to join the other three girls. Ryouta has regrets. He forgot that the very first doujin in his list—the one with the highest rating that he can give—is full-on tentacles. Everybody gets to experience the wonderful flavor of tentacle grape in it, and now his girlfriends and Aiko get to look at the pages full of erotic tentacles and cute girls as well. Serra turns around, looks at him, and gives him a thumbs-up before returning her attention to the doujin. "I think I would rather you have tried to stop me from seeing this, you... perverted dog bastard," Cassandra says, upgrading his nickname to a whole new level. "I see, I see. So, these are the kinds of situations you like, Ryouta? Ooh! That expression she's making... I'll have to remember that." "And making peace signs! But I think I'd probably die if anything that big tried to attack me," Aiko says. "I would prefer monsters or even wild beasts over strange-looking octopus creatures, but I love this scenario! To have your entire being broken down to just accepting whatever your new master forces upon you—it's wonderful!" Alice comments. "How do you even get into stuff like this?" Cassandra asks. "Well, uhh... I guess this kind of thing just sort of happens when you're on the internet for so long," Ryouta tries explaining, but she's not impressed by his answer. None of the girls are surprised by the next doujins in the list, however. Girls wearing sweaters are on the covers of the next seven doujins, then there are more tentacles, then more sweaters. Serra and Cassandra realize that most of the girls in his top-rated favorites look like them. Half the girls are short, petite girls with long silvery hair; the other half are busty blondes. Serra opens one with a girl who resembles herself and notices a controversial tag next to the cover, highlighting it for all to see. Next, she opens a new tab and goes to the city's government website. Her final destination is the police section, where she highlights their phone number and takes out her cell phone. "H-hey now, it's not my fault that people like to label anything with short, petite girls that way even if they aren't," she protests. "I mean, you're not a loli, Serra," Ryouta explains, but since she's looking at her phone and pretending to dial the police, she doesn't hear him. The other girls find it amusing—well, maybe not Cassandra. Cassandra is definitely judging him. "Cass Cass, don't forget that you've done things to her body too. If I'm going to jail over Serra, you're coming with me," Ryouta tells her. "You don't need to say that out loud!" Cassandra shouts, her face red. Serra loses interest in her own joke and moves on to the next doujin on his list. The girls are no longer like her and Cassandra but instead look like monsters: girls with a single giant tail instead of legs, girls with fur-covered arms and legs, girls with the ears and tails of common household pets, and even one with the lower body of a spider. Both Serra and Cassandra judge him for the spider. "Hey, it's different if the spider is a cute girl too," Ryouta explains. One cover seems to have drawn Aiko's attention more than the others. When Ryouta looks, he sees her staring at probably the bustiest girl among the covers—her breasts are utterly massive, she's wearing skimpy clothes, and she has cat ears and a cat tail. "I knew it—you like them huge!" Aiko says accusingly. “I never said I didn’t like them big, just that I like all sizes! Hey, where’d you get that notepad from? Isn’t paper super expensive now?” Ryouta asks. “Oh, this? This is what I’ve been using to keep my notes on you ever since we bumped into each other. I’ve got your sleeping pattern, how often you go to the bathroom, how much time you spend on the computer, and all that!” Aiko explains. “And now I have your fetishes and preferred body types!” “How—how’d you find out all that other stuff?” Ryouta asks. Aiko points at the window across the street. Across the street is a large, tall building he had considered moving into, but it had less privacy, so he chose his current apartment instead. “That window is in a hallway, so I can stand there and look into your room!” Aiko says. Ryouta closes his blinds. “Thanks for telling me I need to keep my blinds closed. I never considered anyone could look in from over there,” he says. “It’s okay! I don’t need to collect information on you from t "Because I want to play! It's not fair that everybody else gets to play with you while I don't!"	"Isn't—isn't this a bit overkill? I mean... is it that important?" Ryouta asks. Cassandra, standing behind him with her arms locked underneath his own and holding him back, asks, "You don't want to hide anything from us, right?" "This is good research material. Great idea, Serra!" Aiko says, patting Serra on the top of her head. Even Alice is in the room now, sitting in her wheelchair next to Serra. Ryouta is going to have to start deleting his history. "Alright, I won't try to stop you. It's embarrassing, but there isn't anything too bad in there. I think," Ryouta says. "Promise?" Cassandra asks. "I promise." With that, Cassandra lets go of him and they both walk over to join the other three girls. Ryouta has regrets. He forgot that the very first doujin in his list—the one with the highest rating that he can give—is full-on tentacles. Everybody gets to experience the wonderful flavor of tentacle grape in it, and now his girlfriends and Aiko get to look at the pages full of erotic tentacles and cute girls as well. Serra turns around, looks at him, and gives him a thumbs-up before returning her attention to the doujin. "I think I would rather you have tried to stop me from seeing this, you... perverted dog bastard," Cassandra says, upgrading his nickname to a whole new level. "I see, I see. So, these are the kinds of situations you like, Ryouta? Ooh! That expression she's making..." I'll have to remember that." "And making peace signs! But I think I'd probably die if anything that big tried to attack me," Aiko says. "I would prefer monsters or even wild beasts over strange-looking octopus creatures, but I love this scenario! To have your entire being broken down to just accepting whatever your new master forces upon you—it's wonderful!" Alice comments. "How do you even get into stuff like this?" Cassandra asks. "Well, uhh... I guess this kind of thing just sort of happens when you're on the internet for so long," Ryouta tries explaining, but she's not impressed by his answer. None of the girls are surprised by the next doujins in the list, however. Girls wearing sweaters are on the covers of the next seven doujins, then there are more tentacles, then more sweaters. Serra and Cassandra realize that most of the girls in his top-rated favorites look like them. Half the girls are short, petite girls with long silvery hair; the other half are busty blondes. Serra opens one with a girl who resembles herself and notices a controversial tag next to the cover, highlighting it for all to see. Next, she opens a new tab and goes to the city's government website. Her final destination is the police section, where she highlights their phone number and takes out her cell phone. "H-hey now, it's not my fault that people like to label anything with short, petite girls that way even if they aren't," she protests. "I mean, you're not a loli, Serra," Ryouta explains, but since she's looking at her phone and pretending to dial the police, she doesn't hear him. The other girls find it amusing—well, maybe not Cassandra. Cassandra is definitely judging him. "Cass Cass, don't forget that you've done things to her body too. If I'm going to jail over Serra, you're coming with me," Ryouta tells her. "You don't need to say that out loud!" Cassandra shouts, her face red. Serra loses interest in her own joke and moves on to the next doujin on his list. The girls are no longer like her and Cassandra but instead look like monsters: girls with a single giant tail instead of legs, girls with fur-covered arms and legs, girls with the ears and tails of common household pets, and even one with the lower body of a spider. Both Serra and Cassandra judge him for the spider. "Hey, it's different if the spider is a cute girl too," Ryouta explains. One cover seems to have drawn Aiko's attention more than the others. When Ryouta looks, he sees her staring at probably the bustiest girl among the covers—her breasts are utterly massive, she's wearing skimpy clothes, and she has cat ears and a cat tail. "I knew it—you like them huge!" Aiko says accusingly. “I never said I didn’t like them big, just that I like all sizes! Hey, where’d you get that notepad from? Isn’t paper super expensive now?” Ryouta asks. “Oh, this? This is what I’ve been using to keep my notes on you ever since we bumped into each other. I’ve got your sleeping pattern, how often you go to the bathroom, how much time you spend on the computer, and all that!” Aiko explains. “And now I have your fetishes and preferred body types!” “How—how’d you find out all that other stuff?” Ryouta asks. Aiko points at the window across the street. Across the street is a large, tall building he had considered moving into, but it had less privacy, so he chose his current apartment instead. “That window is in a hallway, so I can stand there and look into your room!” Aiko says. Ryouta closes his blinds. “Thanks for telling me I need to keep my blinds closed. I never considered anyone could look in from over there,” he says. “It’s okay! I don’t need to collect information on you from t "Because I want to play! It's not fair that everybody else gets to play with you while I don't!".
long_en_374	poet_en	1,047	en	I picked this up at Target the other day — it was so cute, and it was 50% off! I also got a small jug to fit on the bottom shelf of the fridge and filled it with milk. So now Chloe can make her own breakfast while I'm still in bed — hence the title. We also got the popcorn maker, which was 50% off, for movie nights with the girls. Fun times! Okay, the title isn't the best, but I couldn't think of anything else. It's supposed to be a pirate reference — you'll see. Yesterday Chloe and I were talking about her Uncle Brian and Aunt Kristina; they're coming in two days! They've been living in China since July — my sister-in-law is teaching at an American school there, and my brother is an assistant — and they're coming home for Christmas. We're very excited, and Chloe has been counting down the days. So anyway, back to my story. We were talking about when they would be here, and she said, "Yeah, they're probably in the car right now driving to get here!" My parents drive up from Georgia all the time, so that's why she said that. I explained to her that China is very far away; it's across the ocean, so they have to fly here. They can't drive. When she heard me say "ocean," she got very excited and said, "I love the ocean! I wish I were a pirate so I could sail the seas and be near the ocean all the time!" And just to clarify, she meant the feather-in-your-hat, parrot-on-your-shoulder, "argh!"-saying kind of pirate. Not the pirates who are wreaking havoc off the coast of Africa, but I'm sure you knew that. As I was walking Chloe home from the bus stop today, she went out of her way to step in all the snow, which was mixed with mud and leaves. I asked her not to step in the mud because she'll track it all over my clean floors. She stopped, looked quizzical, and said, "What? We have clean floors?!" I guess not. This year for Thanksgiving we went to the Outer Banks with Mike's family. Each of us had a dish to bring, so everyone contributed. We had such a fun weekend, especially the kids! During our time there, Mike and Will (with some help from others) built a kite out of drinking straws and mylar. It took them a long time to make it, but it was fun watching it fly. The weekend didn't start out smoothly and we had some bumps while we were there, but we still had a lot of fun. The day we left, I found out Emily had pink eye. The day after we returned, Chloe had croup. Then Zachary got a small fever and a cough. The day before we left, Mike pulled something in his foot and could barely walk. The day we left, I started to get a cold, which is now at its worst. Even though there was a lot going on, there was still some quiet time to relax and read. I'll leave you with some photos... My Drama Queen Chloe and Aunt Andi. Emily seeing the ocean for the first time. Little Man enjoyed himself too. This is the kite that they made; doesn't it look festive? The finished gingerbread house that the kids made with Aunt Andi. The kids in their PJs. You wouldn't know it by looking at them, but this was taken right after we watched the sunrise. Yes, they get up that early! I started to see this on other blogs and thought that I would join in on the fun (even though I should be starting dinner). The things that I have done are in bold print. 01. Bought everyone in the bar a drink 02. Swam with dolphins 03. Climbed a mountain 04. Taken a Ferrari for a test drive 05. Been inside the Great Pyramid 06. Held a tarantula 07. Taken a candlelit bath with someone 08. Said "I love you" and meant it 09. Hugged a tree 10. Bungee jumped 11. Visited Paris (almost did this; had plane tickets and hotel reservations but had to cancel for a family emergency) 12. Watched a lightning storm at sea 13. Stayed up all night long and saw the sunrise 14. Seen the Northern Lights 15. Gone to a huge sports game 16. Walked the stairs to the top of the Leaning Tower of Pisa 17. Grown and eaten your own vegetables 18. Touched an iceberg 19. Slept under the stars 20. Changed a baby's diaper (still am!) 21. Taken a trip in a hot air balloon 22. Watched a meteor shower 23. Gotten drunk on champagne 24. Given more than you can afford to charity 25. Looked up at the night sky through a telescope 26. Had an uncontrollable giggling fit at the worst possible moment. Had a food fight. Bet on a winning horse. Asked out a stranger. Had a snowball fight. Screamed as loudly as you possibly can. Held a lamb. Seen a total eclipse. Ridden a roller coaster. Hit a home run. Danced like a fool and didn't care who was looking. Adopted an accent for an entire day. Actually felt happy about your life, even for just a moment. Had two hard drives for your computer. Visited all 50 states. Taken care of someone who was drunk. Had amazing friends (still do). Danced with a stranger in a foreign country. Watched whales. Stolen a sign. Backpacked in Europe. Taken a road trip. Gone rock climbing. Taken a midnight walk on the beach. Gone skydiving. Visited Ireland. Been heartbroken longer than you were actually in love. In a restaurant, sat at a stranger's table and had a meal with them. Visited Japan. Milked a cow. Posted by * Update * - Everything has already been claimed. I have some baby things that Zachary has already outgrown (or doesn't really like) that I'd like to get rid of. If you're interested, please let me know as soon as possible.	I picked this up at Target the other day — it was so cute, and it was fifty percent off! I also got a small jug to fit on the bottom shelf of the fridge and filled it with milk. So now Chloe can make her own breakfast while I'm still in bed — hence the title. We also got the popcorn maker, which was fifty percent off, for movie nights with the girls. Fun times! Okay, the title isn't the best, but I couldn't think of anything else. It's supposed to be a pirate reference — you'll see. Yesterday Chloe and I were talking about her Uncle Brian and Aunt Kristina; they're coming in two days! They've been living in China since July — my sister-in-law is teaching at an American school there, and my brother is an assistant — and they're coming home for Christmas. We're very excited, and Chloe has been counting down the days. So anyway, back to my story. We were talking about when they would be here, and she said, "Yeah, they're probably in the car right now driving to get here!" My parents drive up from Georgia all the time, so that's why she said that. I explained to her that China is very far away; it's across the ocean, so they have to fly here. They can't drive. When she heard me say "ocean," she got very excited and said, "I love the ocean! I wish I were a pirate so I could sail the seas and be near the ocean all the time!" And just to clarify, she meant the feather-in-your-hat, parrot-on-your-shoulder, "argh!"-saying kind of pirate. Not the pirates who are wreaking havoc off the coast of Africa, but I'm sure you knew that. As I was walking Chloe home from the bus stop today, she went out of her way to step in all the snow, which was mixed with mud and leaves. I asked her not to step in the mud because she'll track it all over my clean floors. She stopped, looked quizzical, and said, "What? We have clean floors?!" I guess not. This year for Thanksgiving we went to the Outer Banks with Mike's family. Each of us had a dish to bring, so everyone contributed. We had such a fun weekend, especially the kids! During our time there, Mike and Will (with some help from others) built a kite out of drinking straws and mylar. It took them a long time to make it, but it was fun watching it fly. The weekend didn't start out smoothly and we had some bumps while we were there, but we still had a lot of fun. The day we left, I found out Emily had pink eye. The day after we returned, Chloe had croup. Then Zachary got a small fever and a cough. The day before we left, Mike pulled something in his foot and could barely walk. The day we left, I started to get a cold, which is now at its worst. Even though there was a lot going on, there was still some quiet time to relax and read. I'll leave you with some photos... My Drama Queen Chloe and Aunt Andi. Emily seeing the ocean for the first time. Little Man enjoyed himself too. This is the kite that they made; doesn't it look festive? The finished gingerbread house that the kids made with Aunt Andi. The kids in their PJs. You wouldn't know it by looking at them, but this was taken right after we watched the sunrise. Yes, they get up that early! I started to see this on other blogs and thought that I would join in on the fun (even though I should be starting dinner). The things that I have done are in bold print. one. Bought everyone in the bar a drink two. Swam with dolphins three. Climbed a mountain four. Taken a Ferrari for a test drive five. Been inside the Great Pyramid six. Held a tarantula seven. Taken a candlelit bath with someone eight. Said "I love you" and meant it nine. Hugged a tree ten. Bungee jumped eleven. Visited Paris (almost did this; had plane tickets and hotel reservations but had to cancel for a family emergency) twelve. Watched a lightning storm at sea thirteen. Stayed up all night long and saw the sunrise fourteen. Seen the Northern Lights fifteen. Gone to a huge sports game sixteen. Walked the stairs to the top of the Leaning Tower of Pisa seventeen. Grown and eaten your own vegetables eighteen. Touched an iceberg nineteen. Slept under the stars twenty. Changed a baby's diaper (still am!) twenty one. Taken a trip in a hot air balloon twenty two. Watched a meteor shower twenty three. Gotten drunk on champagne twenty four. Given more than you can afford to charity twenty five. Looked up at the night sky through a telescope twenty six. Had an uncontrollable giggling fit at the worst possible moment. Had a food fight. Bet on a winning horse. Asked out a stranger. Had a snowball fight. Screamed as loudly as you possibly can. Held a lamb. Seen a total eclipse. Ridden a roller coaster. Hit a home run. Danced like a fool and didn't care who was looking. Adopted an accent for an entire day. Actually felt happy about your life, even for just a moment. Had two hard drives for your computer. Visited all fifty states. Taken care of someone who was drunk. Had amazing friends (still do). Danced with a stranger in a foreign country. Watched whales. Stolen a sign. Backpacked in Europe. Taken a road trip. Gone rock climbing. Taken a midnight walk on the beach. Gone skydiving. Visited Ireland. Been heartbroken longer than you were actually in love. In a restaurant, sat at a stranger's table and had a meal with them. Visited Japan. Milked a cow. Posted by * Update * - Everything has already been claimed. I have some baby things that Zachary has already outgrown (or doesn't really like) that I'd like to get rid of. If you're interested, please let me know as soon as possible.
long_en_298	wiki_en	1,035	en	The 6555th Aerospace Test Group is an inactive United States Air Force unit. It was last assigned to the Eastern Space and Missile Center and stationed at Patrick Air Force Base, Florida; it was inactivated on 1 October 1990. Prior to the activation of Air Force Space Command, the unit was responsible for the development of USAF missiles, including tactical surface-to-surface missiles, the CIM-10 Bomarc interceptor, the SM-62 Snark intercontinental cruise missile, intercontinental ballistic missiles, and heavy launch rockets for military and satellite deployment. The unit played a key role in the civilian NASA programs Project Mercury, Project Gemini, and Project Apollo, as well as in military Space Shuttle flights. The unit's mission is now performed by the 45th Space Launch Delta, although there is no direct lineage. Activated in December 1950 to replace the 550th Guided Missiles Wing, the 6555th had a distinguished career launching and managing ballistic missiles, space launch vehicles, and payloads for the Ballistic Systems Division, the Space Systems Division, and the Space and Missile Systems Organization. As a wing or a group, the 6555th earned ten Air Force Outstanding Unit Awards between 21 December 1959 and October 1990. In the 1950s the unit underwent several designation changes and organizational realignments; as launches of winged missiles continued, the wing gained two new units: the 1st Pilotless Bomber Squadron in October 1951 and the 69th Pilotless Bomber Squadron in January 1952. Thereafter, the 6555th focused on assembling, testing, and launching B-61 Matador missiles to prepare the 1st and 69th Pilotless Bomber Squadrons for operations in Europe. The 6555th Guided Missile Wing became the 6555th Guided Missile Group on 1 March 1953, and the 1st and 69th Pilotless Bomber Squadrons were reassigned to Tactical Air Command (TAC) on 15 January 1954. Because TAC agreed to train all other B-61 Matador squadrons at its school in Orlando AFB, Florida, the 6555th Guided Missile Group was reduced to little more than a squadron when the 69th completed field training in the summer of 1954. The 6555th Guided Missile Group was discontinued on 7 September 1954. The 6555th Guided Missile Squadron continued as a B-61 Matador research and development testing unit and was reassigned to AFMTC Headquarters on 7 September 1954. The 6555th Guided Missile Squadron became the 6555th Guided Missile Group (Test and Evaluation) on 15 August 1959 and was reassigned to the Air Force Ballistic Missile Division on 21 December 1959 without a change of station; concurrent with the reassignment, the group absorbed the resources of the Division's Assistant Commander for Missile Tests. At the beginning of 1971, the 6555th Aerospace Test Group consisted of a commander's office and three divisions: Support, Atlas Systems, and Titan III Systems. Although the Test Group's launch operations in the early 1970s centered on the Atlas and Titan III systems, the Group established a Space Transportation System (STS) Division on 1 July 1974 to ensure the Department of Defense's Shuttle requirements were incorporated into future Shuttle operations at Kennedy Space Center (KSC). On 1 November 1975 the Group reorganized its Atlas and Titan III launch agencies into a Space Launch Vehicle Systems Division and simultaneously consolidated the Atlas Satellite Launch Systems Branch and the Titan III Space Satellite Systems Launch Operations Branch into a newly created Satellite Systems Division. The 6595th Aerospace Test Wing commander directed these changes to place booster operations under one division chief and payload operations under another. The IUS Operations Branch was placed under the Space Launch Vehicle Systems Division when it was formed on 1 July 1977. After the final Atlas-Agena launch on 6 April 1978, the Space Launch Vehicle Systems and Satellite Systems divisions shifted focus from Atlas-Agena operations on Complex 13 to Atlas-Centaur boosters and Department of Defense payloads on Complex 36. On 1 October 1979 the Group was transferred to the Eastern Space and Missile Center (ESMC), the immediate predecessor of the 45th Space Wing. The unit was inactivated on 1 October 1990 when Air Force Space Command inactivated the provisional unit and merged its organization with ESMC. Most of the 6555th's resources were reorganized as the 1st Space Launch Squadron under ESMC and two Combined Task Forces (CTFs) serving AFSPC and Air Force Systems Command. Ultimately, the last vestiges of the 6555th were inactivated on 1 July 1992 as Air Force Systems Command and Air Force Logistics Command merged to form Air Force Materiel Command. Today, the mission is performed by components of the 45th Space Wing: the 45th Operations Group and the 45th Launch Group. Weapons and missile development — Postwar era Between 1946 and 1950, the group's predecessor units, the 1st Experimental Guided Missiles Group and the 550th Guided Missiles Wing, tested a variety of glided bombs and tactical missiles. They also developed QB-17 drone aircraft for use in atomic bomb testing and later as targets for anti-aircraft missiles. The 550th GMW also launched the first rockets from the Joint Long Range Proving Ground at Cape Canaveral, Florida, in 1950. Republic-Ford JB-2 (1947–1949) The JB-2 was a U.S.-made copy of the German V-1 surface-to-surface, pilotless flying bomb first used against England in June 1944. Planned for use in the invasion of Japan, the missile was never used in combat during World War II. The 1st Experimental Guided Missiles Group began testing the JB-2 at White Sands, New Mexico, in March 1947 and spent several months preparing a detachment for cold‑weather testing in Alaska in November 1948. Testing was also conducted at the Army Air Forces Proving Ground on Santa Rosa Island, Florida. The JB-2 was never used operationally, but it led to the development of the USAF’s first operational cruise missile, the Martin B-61A Matador. QB-17L/BQ-17 Flying Fortresses (1946–1950) were unmanned aircraft used to fly near or even through mushroom clouds during postwar atomic tests. B-17s were withdrawn from storage for conversion into drones with the addition of radio, radar, television, and other equipment. Most of the conversions were performed by the San Antonio Air Depot at Kelly Field, Texas.	The six thousand five hundred fifty fifth Aerospace Test Group is an inactive United States Air Force unit. It was last assigned to the Eastern Space and Missile Center and stationed at Patrick Air Force Base, Florida; it was inactivated on one October nineteen ninety. Prior to the activation of Air Force Space Command, the unit was responsible for the development of USAF missiles, including tactical surface-to-surface missiles, the CIM-ten Bomarc interceptor, the SM-sixty two Snark intercontinental cruise missile, intercontinental ballistic missiles, and heavy launch rockets for military and satellite deployment. The unit played a key role in the civilian NASA programs Project Mercury, Project Gemini, and Project Apollo, as well as in military Space Shuttle flights. The unit's mission is now performed by the forty fifth Space Launch Delta, although there is no direct lineage. Activated in December nineteen fifty to replace the five hundred fiftieth Guided Missiles Wing, the six thousand five hundred fifty fifth had a distinguished career launching and managing ballistic missiles, space launch vehicles, and payloads for the Ballistic Systems Division, the Space Systems Division, and the Space and Missile Systems Organization. As a wing or a group, the six thousand five hundred fifty fifth earned ten Air Force Outstanding Unit Awards between twenty one December nineteen fifty nine and October nineteen ninety. In the nineteen fifties the unit underwent several designation changes and organizational realignments; as launches of winged missiles continued, the wing gained two new units: the first Pilotless Bomber Squadron in October nineteen fifty one and the sixty ninth Pilotless Bomber Squadron in January nineteen fifty two. Thereafter, the six thousand five hundred fifty fifth focused on assembling, testing, and launching B sixty one Matador missiles to prepare the first and sixty ninth Pilotless Bomber Squadrons for operations in Europe. The six thousand five hundred fifty fifth Guided Missile Wing became the six thousand five hundred fifty fifth Guided Missile Group on first March nineteen fifty three, and the first and sixty ninth Pilotless Bomber Squadrons were reassigned to Tactical Air Command (TAC) on fifteenth January nineteen fifty four. Because TAC agreed to train all other B sixty one Matador squadrons at its school in Orlando AFB, Florida, the six thousand five hundred fifty fifth Guided Missile Group was reduced to little more than a squadron when the sixty ninth completed field training in the summer of nineteen fifty four. The six thousand five hundred fifty fifth Guided Missile Group was discontinued on seventh September nineteen fifty four. The six thousand five hundred fifty fifth Guided Missile Squadron continued as a B sixty one Matador research and development testing unit and was reassigned to AFMTC Headquarters on seventh September nineteen fifty four. The six thousand five hundred fifty fifth Guided Missile Squadron became the six thousand five hundred fifty fifth Guided Missile Group (Test and Evaluation) on fifteenth August nineteen fifty nine and was reassigned to the Air Force Ballistic Missile Division on twenty first December nineteen fifty nine without a change of station; concurrent with the reassignment, the group absorbed the resources of the Division's Assistant Commander for Missile Tests. At the beginning of nineteen seventy-one, the six thousand five hundred fifty-fifth Aerospace Test Group consisted of a commander's office and three divisions: Support, Atlas Systems, and Titan three Systems. Although the Test Group's launch operations in the early nineteen seventies centered on the Atlas and Titan three systems, the Group established a Space Transportation System (STS) Division on the first of July nineteen seventy-four to ensure the Department of Defense's Shuttle requirements were incorporated into future Shuttle operations at Kennedy Space Center (KSC). On the first of November nineteen seventy-five the Group reorganized its Atlas and Titan three launch agencies into a Space Launch Vehicle Systems Division and simultaneously consolidated the Atlas Satellite Launch Systems Branch and the Titan three Space Satellite Systems Launch Operations Branch into a newly created Satellite Systems Division. The six thousand five hundred ninety-fifth Aerospace Test Wing commander directed these changes to place booster operations under one division chief and payload operations under another. The IUS Operations Branch was placed under the Space Launch Vehicle Systems Division when it was formed on the first of July nineteen seventy-seven. After the final Atlas-Agena launch on the sixth of April nineteen seventy-eight, the Space Launch Vehicle Systems and Satellite Systems divisions shifted focus from Atlas-Agena operations on Complex thirteen to Atlas-Centaur boosters and Department of Defense payloads on Complex thirty-six. On the first of October nineteen seventy-nine the Group was transferred to the Eastern Space and Missile Center (ESMC), the immediate predecessor of the forty-fifth Space Wing. The unit was inactivated on one October nineteen ninety when Air Force Space Command inactivated the provisional unit and merged its organization with ESMC. Most of the six thousand five hundred fifty-fifth's resources were reorganized as the first Space Launch Squadron under ESMC and two Combined Task Forces (CTFs) serving AFSPC and Air Force Systems Command. Ultimately, the last vestiges of the six thousand five hundred fifty-fifth were inactivated on one July nineteen ninety-two as Air Force Systems Command and Air Force Logistics Command merged to form Air Force Materiel Command. Today, the mission is performed by components of the forty-fifth Space Wing: the forty-fifth Operations Group and the forty-fifth Launch Group. Weapons and missile development — Postwar era Between nineteen forty-six and nineteen fifty, the group's predecessor units, the first Experimental Guided Missiles Group and the five hundred fiftieth Guided Missiles Wing, tested a variety of glided bombs and tactical missiles. They also developed QB seventeen drone aircraft for use in atomic bomb testing and later as targets for anti-aircraft missiles. The five hundred fiftieth GMW also launched the first rockets from the Joint Long Range Proving Ground at Cape Canaveral, Florida, in nineteen fifty. Republic-Ford JB two (nineteen forty-seven–nineteen forty-nine) The JB two was a U.S.-made copy of the German V one surface-to-surface, pilotless flying bomb first used against England in June nineteen forty-four. Planned for use in the invasion of Japan, the missile was never used in combat during World War II. The first Experimental Guided Missiles Group began testing the JB two at White Sands, New Mexico, in March nineteen forty-seven and spent several months preparing a detachment for cold‑weather testing in Alaska in November nineteen forty-eight. Testing was also conducted at the Army Air Forces Proving Ground on Santa Rosa Island, Florida. The JB two was never used operationally, but it led to the development of the USAF’s first operational cruise missile, the Martin B sixty-one A Matador. QB seventeen L slash BQ seventeen Flying Fortresses (nineteen forty-six to nineteen fifty) were unmanned aircraft used to fly near or even through mushroom clouds during postwar atomic tests. B seventeen s were withdrawn from storage for conversion into drones with the addition of radio, radar, television, and other equipment. Most of the conversions were performed by the San Antonio Air Depot at Kelly Field, Texas.
long_en_352	poet_en	966	en	I headed out last night to see Daniel's last baseball game of the season. It started at 7, so it was too late to bring the boys with me. As I was leaving the house, I heard some screaming. Colin and Ciaran were both standing in the giant bay window, hysterically crying and yelling, "No, Mommy, no—leave!" You would have thought I was leaving them to be tortured. Good God, they were going nuts. Luckily I called Frank, and they settled down pretty quickly after I left. The last thing I saw was Frank scoop them both up and walk away with two screaming kids. I was curious to see how the night would turn out. If I'm not home, will they still say, "Mommy, do it"? Or will it be forgotten because I'm out of the equation? I'm hoping it will be forgotten and it's only something they will do when I am home. Well, they all went to sleep with Frank. No "Mommy, do it"—apparently Daddy can do it! All was quiet on the homefront until 6:45 a.m. That's a good thing in this house. Here's a funny one for you. I was getting dressed to take the kids to the mall play area before dinner. I was talking to the boys and said, "Okay, Mommy's just going to throw on her jeans and we'll go." To that, Ciaran turned around and said, "No, jeans, Mommy, no jeans." I said, "Well, I have to wear something." He grabbed a red-and-blue sundress I had worn on Sunday and said, "Mommy, you wear a dress, not jeans!" Lord only knows what was going on in his head. But he didn't want me in jeans, that's for sure! We continue to make progress in our speech. Each day the boys are talking more and more. Sometimes I have no clue what the heck they are saying, but it gets clearer and clearer as each day passes. Cormac continues to progress beautifully. He's so funny now. He is so much more of a loner than his brothers. He loves to play with them, but sometimes just needs to do his own thing. He is also starting to catch on that he can blame another brother for his not-so-kind acts. This all started with Colin. No matter what happened, Colin would say, "ree ree did it." He was so quick to blame. Okay, so this is a popular statement in our house lately, especially over the last few days. Ciaran is really the one who says it the most, but it is spilling over to the other two. The problem? Well, they won't let "Daddy do it" if they want "Mommy do it." That means put their blanket on them at night, get them their milk, feed them their food, change their d I guess I should just let Frank do whatever it is they want me to do, and we do that a lot, especially when it comes to changing diapers. They usually settle down after a minute or so. But I guess this comes with the territory of being a stay-at-home mom. I do everything. We don't really ever leave them with a sitter other than my mom or Maggie, and that's rare. Let me say, it's mostly Ciaran. Cormac rarely does this, and Colin only occasionally. Our weekend was exciting. I took the boys to Maggie's on Friday. We played in the baby pool and hung out. We went to the boardwalk in Point Pleasant and met up with my cousins and their kids. It was the most beautiful night. The boys got pinwheels and loved them. I went to bed about 12:30 a.m., and my little Cormac woke up at 1:00 a.m. and decided to stay up until 4:00 a.m. This was so unusual for him, or for any of them for that matter. I guess he got some quality alone time with Mommy. Finally I went to sleep about 4:30 a.m., and Ciaran was awake by 6:30 a.m. I think that means I slept two hours. We headed home and went to Richie's surprise 60th birthday party. Lots of fun. The boys ran wild in the yard from 5:00 until 8:00 p.m. I was thinking I would get a really good night's sleep and catch up from the triplets getting a hold of glitter and then spilling bubbles all over. While I wasn't too thrilled with the glitter, it was partly my fault — I left them alone in their room for a second to get something. I didn't know they could open the container with the craft things in it. I do now! When I left the room to get the vacuum to clean up the glitter, Cormac managed to open a bottle of bubbles. He unscrewed the cap; it was new, so there was a seal on it — he opened the seal too, all in a few seconds. Anyway, once they spilled the bubbles it was just hilarious because they couldn't stand on the hardwood floors. They were laughing hysterically. And what did I do — freak out and start yelling? No, I grabbed the camera and took a video of the chaos. I did tell them they weren't allowed to do it again, but look how much fun they were having. I'll be finding glitter, I'm sure, for the next two years! That stuff never goes away. It's stuck all over their bodies. No matter how much you vacuum, it will never go away.	I headed out last night to see Daniel's last baseball game of the season. It started at seven, so it was too late to bring the boys with me. As I was leaving the house, I heard some screaming. Colin and Ciaran were both standing in the giant bay window, hysterically crying and yelling, "No, Mommy, no—leave!" You would have thought I was leaving them to be tortured. Good God, they were going nuts. Luckily I called Frank, and they settled down pretty quickly after I left. The last thing I saw was Frank scoop them both up and walk away with two screaming kids. I was curious to see how the night would turn out. If I'm not home, will they still say, "Mommy, do it"? Or will it be forgotten because I'm out of the equation? I'm hoping it will be forgotten and it's only something they will do when I am home. Well, they all went to sleep with Frank. No "Mommy, do it"—apparently Daddy can do it! All was quiet on the homefront until six:forty-five a.m. That's a good thing in this house. Here's a funny one for you. I was getting dressed to take the kids to the mall play area before dinner. I was talking to the boys and said, "Okay, Mommy's just going to throw on her jeans and we'll go." To that, Ciaran turned around and said, "No, jeans, Mommy, no jeans." I said, "Well, I have to wear something." He grabbed a red-and-blue sundress I had worn on Sunday and said, "Mommy, you wear a dress, not jeans!" Lord only knows what was going on in his head. But he didn't want me in jeans, that's for sure! We continue to make progress in our speech. Each day the boys are talking more and more. Sometimes I have no clue what the heck they are saying, but it gets clearer and clearer as each day passes. Cormac continues to progress beautifully. He's so funny now. He is so much more of a loner than his brothers. He loves to play with them, but sometimes just needs to do his own thing. He is also starting to catch on that he can blame another brother for his not-so-kind acts. This all started with Colin. No matter what happened, Colin would say, "ree ree did it." He was so quick to blame. Okay, so this is a popular statement in our house lately, especially over the last few days. Ciaran is really the one who says it the most, but it is spilling over to the other two. The problem? Well, they won't let "Daddy do it" if they want "Mommy do it." That means put their blanket on them at night, get them their milk, feed them their food, change their d I guess I should just let Frank do whatever it is they want me to do, and we do that a lot, especially when it comes to changing diapers. They usually settle down after a minute or so. But I guess this comes with the territory of being a stay-at-home mom. I do everything. We don't really ever leave them with a sitter other than my mom or Maggie, and that's rare. Let me say, it's mostly Ciaran. Cormac rarely does this, and Colin only occasionally. Our weekend was exciting. I took the boys to Maggie's on Friday. We played in the baby pool and hung out. We went to the boardwalk in Point Pleasant and met up with my cousins and their kids. It was the most beautiful night. The boys got pinwheels and loved them. I went to bed about twelve colon thirty a dot m dot, and my little Cormac woke up at one colon zero zero a dot m dot and decided to stay up until four colon zero zero a dot m dot. This was so unusual for him, or for any of them for that matter. I guess he got some quality alone time with Mommy. Finally I went to sleep about four colon thirty a dot m dot, and Ciaran was awake by six colon thirty a dot m dot. I think that means I slept two hours. We headed home and went to Richie's surprise sixtieth birthday party. Lots of fun. The boys ran wild in the yard from five colon zero zero until eight colon zero zero p dot m dot. I was thinking I would get a really good night's sleep and catch up from the triplets getting a hold of glitter and then spilling bubbles all over. While I wasn't too thrilled with the glitter, it was partly my fault — I left them alone in their room for a second to get something. I didn't know they could open the container with the craft things in it. I do now! When I left the room to get the vacuum to clean up the glitter, Cormac managed to open a bottle of bubbles. He unscrewed the cap; it was new, so there was a seal on it — he opened the seal too, all in a few seconds. Anyway, once they spilled the bubbles it was just hilarious because they couldn't stand on the hardwood floors. They were laughing hysterically. And what did I do — freak out and start yelling? No, I grabbed the camera and took a video of the chaos. I did tell them they weren't allowed to do it again, but look how much fun they were having. I'll be finding glitter, I'm sure, for the next two years! That stuff never goes away. It's stuck all over their bodies. No matter how much you vacuum, it will never go away.
long_en_379	poet_en	999	en	We were married on August 27, 2011. We are having a little girl, and her name is Louisa Jean Riley. This blog has now grown into our journey. We will walk through Louisa's diagnosis of an atrioventricular canal defect and Down syndrome. Mike and I will both write to serve as a ministry and to keep people up to date on Louisa's progress and delivery. Saturday, June 30, 2007. One of my older brothers, Troy, had a bonfire at his house. It was really cool, and we planned it together so I could bring my parents to meet him for the first time. Both of my brothers were there, Travis and Troy. Both are married and both have beautiful kids. I'll try to explain the people in the picture. In the close right-hand corner is Chris, of course. Behind Chris is my dad. In the lower-left corner on the couch are Travis and his wife, Amber. On the far back wall, all the way to the left, is Grandma; to the right of her is my older brother, Troy. To the right of Troy are my sister Brandy and my brother-in-law-to-be, John. At this point we were inside talking, looking at pictures, and just enjoying each other. My mom saw a picture of my biological father. She was immediately struck by how much we look alike. It's indeed a little surprising to see someone much older than me, someone you've never seen before, who looks so much like me. It's almost a little creepy. Next we went outside and had some hot dogs and s'mores, along with other good food. A little later we started playing with the fireworks John had bought. This only led to what could be called reckless endangerment — you can form your own opinion, though. I don't know if you're familiar with men and shooting off fireworks, but it usually goes like this: "Oooooo, ahhhhh, wow — neat!" Then someone inevitably shouts, "Hey, I got a good idea!" That's exactly what happened. Soon we had a Barney doll filled to the brim with gunpowder, and my brother‑in‑law‑to‑be stuck a bottle rocket up Barney's butt and launched it. Yes, that's right, and it was amazing. Watch the video below. With Troy (my big brother), we haven't had any DNA testing done or anything of that sort, but the fact that he enjoys doing something like this as much as I do is proof enough for me. I had a really good time today. I sighed with relief after finishing work. This sounds kind of odd and funny, but I didn't miss this country too much; I definitely missed my job and working. I like staying busy and earning some cash — it makes the day go by faster. I like the people at Vision Scapes; they're fun to work with and easy to get along with. Very chill work atmosphere. After work I went over to the house Tony and Amber share with Nick and Brandon. I work with Tony, Nick, and Brandon, and our boss owns and rents the house to them. It is really cool and pretty big. They have a huge backyard, and it's right by work, so you can easily walk to the office. After that, Chris took me home and I got cleaned up. My buddy Brad called me—here is a picture of Brad. Brad is a good old boy from way back; we have known each other since elementary school. I like these relationships because I can put my trust in these people and feel confident. Brad is a good guy, plain and simple. So Brad called, and after Chris and I chilled at the house for a bit, it was suddenly 9:30 p.m. We were both surprised, so we drove to Brad's, hung out with some people, and played with Kaido. Kaido is Brad's pit bull; he is an awesome dog. I have had the pleasure of watching Kaido grow up from a pup. He is definitely not a pup anymore, and if you make him angry, you will regret it. MS update: Today, at work and during other activities, I noticed a definite difference in my tremors. They are hardly noticeable unless I am under a great deal of stress. This is great—I think the real test will be in church on Sunday during worship, which is where I had noticed them before. Wouldn't it be awesome if that's where I noticed they were gone? I kept myself busy today. This morning I had to report to the alcohol assessment class. It was supposed to be two hours, but I was out of there in about 45 minutes. Thankfully I got the woman with a level head. She realized that I do not have an alcohol problem and that the minimal alcohol classes would do just fine. So on a Saturday and Sunday from 8 AM to 4:30 PM I have class — fun. At a cost of $200, I better be learning everything there is to know about alcohol: why people drink it and why it is a problem in America, even though I already know all of that. Let me clear something up: I am taking this alcohol class because of when I got in trouble back in September 2006. There is no new charge, and drinking is not a part of my lifestyle. Around 3 PM the photographer for the News Sentinel showed up. We told her some stories from China, and then Jennifer Boen from the newspaper arrived. She did the interview, asked many good questions, and did a good job as she always does. She will be writing a good article. Chris's tire on his Lincoln was making the whole thing shake really badly for the past couple of days.	We were married on August twenty seven, two thousand eleven. We are having a little girl, and her name is Louisa Jean Riley. This blog has now grown into our journey. We will walk through Louisa's diagnosis of an atrioventricular canal defect and Down syndrome. Mike and I will both write to serve as a ministry and to keep people up to date on Louisa's progress and delivery. Saturday, June thirty, two thousand seven. One of my older brothers, Troy, had a bonfire at his house. It was really cool, and we planned it together so I could bring my parents to meet him for the first time. Both of my brothers were there, Travis and Troy. Both are married and both have beautiful kids. I'll try to explain the people in the picture. In the close right-hand corner is Chris, of course. Behind Chris is my dad. In the lower-left corner on the couch are Travis and his wife, Amber. On the far back wall, all the way to the left, is Grandma; to the right of her is my older brother, Troy. To the right of Troy are my sister Brandy and my brother-in-law-to-be, John. At this point we were inside talking, looking at pictures, and just enjoying each other. My mom saw a picture of my biological father. She was immediately struck by how much we look alike. It's indeed a little surprising to see someone much older than me, someone you've never seen before, who looks so much like me. It's almost a little creepy. Next we went outside and had some hot dogs and s'mores, along with other good food. A little later we started playing with the fireworks John had bought. This only led to what could be called reckless endangerment — you can form your own opinion, though. I don't know if you're familiar with men and shooting off fireworks, but it usually goes like this: "Oooooo, ahhhhh, wow — neat!" Then someone inevitably shouts, "Hey, I got a good idea!" That's exactly what happened. Soon we had a Barney doll filled to the brim with gunpowder, and my brother‑in‑law‑to‑be stuck a bottle rocket up Barney's butt and launched it. Yes, that's right, and it was amazing. Watch the video below. With Troy (my big brother), we haven't had any DNA testing done or anything of that sort, but the fact that he enjoys doing something like this as much as I do is proof enough for me. I had a really good time today. I sighed with relief after finishing work. This sounds kind of odd and funny, but I didn't miss this country too much; I definitely missed my job and working. I like staying busy and earning some cash — it makes the day go by faster. I like the people at Vision Scapes; they're fun to work with and easy to get along with. Very chill work atmosphere. After work I went over to the house Tony and Amber share with Nick and Brandon. I work with Tony, Nick, and Brandon, and our boss owns and rents the house to them. It is really cool and pretty big. They have a huge backyard, and it's right by work, so you can easily walk to the office. After that, Chris took me home and I got cleaned up. My buddy Brad called me—here is a picture of Brad. Brad is a good old boy from way back; we have known each other since elementary school. I like these relationships because I can put my trust in these people and feel confident. Brad is a good guy, plain and simple. So Brad called, and after Chris and I chilled at the house for a bit, it was suddenly nine: thirty p.m. We were both surprised, so we drove to Brad's, hung out with some people, and played with Kaido. Kaido is Brad's pit bull; he is an awesome dog. I have had the pleasure of watching Kaido grow up from a pup. He is definitely not a pup anymore, and if you make him angry, you will regret it. MS update: Today, at work and during other activities, I noticed a definite difference in my tremors. They are hardly noticeable unless I am under a great deal of stress. This is great—I think the real test will be in church on Sunday during worship, which is where I had noticed them before. Wouldn't it be awesome if that's where I noticed they were gone? I kept myself busy today. This morning I had to report to the alcohol assessment class. It was supposed to be two hours, but I was out of there in about forty five minutes. Thankfully I got the woman with a level head. She realized that I do not have an alcohol problem and that the minimal alcohol classes would do just fine. So on a Saturday and Sunday from eight AM to four: thirty PM I have class — fun. At a cost of dollar two hundred, I better be learning everything there is to know about alcohol: why people drink it and why it is a problem in America, even though I already know all of that. Let me clear something up: I am taking this alcohol class because of when I got in trouble back in September two thousand six. There is no new charge, and drinking is not a part of my lifestyle. Around three P M the photographer for the News Sentinel showed up. We told her some stories from China, and then Jennifer Boen from the newspaper arrived. She did the interview, asked many good questions, and did a good job as she always does. She will be writing a good article. Chris's tire on his Lincoln was making the whole thing shake really badly for the past couple of days.
long_en_356	poet_en	906	en	“You’re different now,” Ull says, bringing Nell into the room serving as her temporary bedroom. “Am I?” Nell asks, refusing to look at him. “You seem disappointed to be back here. You knew I would find you, surely.” Nell gives no response. “Don’t tell me you actually believed he could help you.” Her hands ball into fists, then stop. “You belong to me, my precious Nehalennia. He would have used you or sold you off like everybody else wanted. Are you forgetting that I am the only one who truly wants to keep you?” “That doesn’t stop you from wanting to use me, just like they all did,” Nell answers. “Of course I want to use you. You are mine. It’s only natural to use what belongs to you, isn’t it? Even so, I care for you and wish to take care of you.” “I would rather be uncared for than have the perverse care of someone like you.” Ull grabs her by the hair again, lifts her off the ground, and pushes her against the wall. “I do not like defiant pets. You would be wise to destroy any lingering hope inside that childish mind of yours. You are mine, and unless you start over with a new body, that will never change. Do I make myself clear?” he asks, tossing her down onto the floor. Nell does not answer. Ull wastes no more time; he places heavy chains around her ankles and wrists. “I don’t normally mind you escaping, but until we deal with this new attitude of yours, you will not be permitted to waste my time.” With that, Ull leaves the room, posts two guards inside with her, and locks the door behind him. “Think he’ll be on soon?” Cassiel asks Serra. The two girls sit at the end of the pier. Cassiel dangles her legs off it, gently kicking back and forth, and Serra kneels behind the blonde, her chin resting on Cassiel’s head and her arms draped over her. Serra occasionally slips a hand to Cassiel’s chest; to her surprise, she only gets a blushing pout in response. “Yeah. He won’t miss this,” Serra answers. “He’s cutting it pretty close, that bastard. I swear if he’s late—” “He won’t be.” “He better not be. Oh, and I wanted to ask, what’s it like having so many parents?” “Lots of presents,” Serra says with a thumbs-up. “Between your parents and now me and Ryouta, don’t you think you’re a bit spoiled?” “I’m super spoiled,” Cassiel replies with a smile. “Hmph. I—I don’t get spoiled…” “Cass, Cass.” “What?” Cassiel asks, turning her head toward Serra now that her head is off hers. Serra plants a kiss on Cassiel’s lips. “I’ll spoil you whenever you want me to,” Serra says. “I—o-okay, maybe I get spoiled, too,” Cassiel says, cheeks flushed. “Oi, big guy, hand me that wrench,” Tabitha orders Bonekraka. Bonekraka tosses the wrench so it lands next to Tabitha, scratching the floorboard of the Shoebill. "Hey! I just cleaned this baby up! You gonna fix that yourself?!" "Nyet," Bonekraka answers, and a moment later a wrench hits the back of his head as a pissed-off Tabitha glares at him from the engine. "Now come bring it to me the proper way!" Bonekraka turns to look at Tabitha with his hands balled into fists, but when he sees the look in her eyes he's reminded of his wife. The grumpy orc submissively takes the wrench over to her, handing it to her properly. "Thank ya. See, now was that so hard to do? Next time don't try sassin' me and scratchin' up my baby." "Da." "O-Olly, you mustn't—we have such little time," Corwin says, looking both left and right to make sure nobody is coming down the alleyway where Oleander has him pressed against the wall. "Come on, Cor, we both know I can make you fire your cannon quickly enough," Oleander coos, running one hand up Corwin's chest while the other travels dangerously lower down his abdomen. "Shouldn't you be happy that you have a lover so eager to please you?" "I—I assure you that I am beyond happy, but what if somebody sees?" "Are you telling me you wouldn't like getting caught in the act?" Oleander's teasing fingers trace Corwin's body. Corwin gulps. "I... do not entirely dislike the thought, but—" "Then," Oleander pauses to sink his hands into Corwin's pants, "let's have some fun before the tournament starts." "We don't know when we're going to be able to do this again." "I'm sure we'll have time to in only a few hours! It's not as if we are leaving one another." "If you want me to stop, all you have to say is 'stop,'" Oleander says, looking up at Corwin with a smug grin. Corwin doesn't tell him to stop. Rock is in the middle of a stare-down with one of the stray cats that has been visiting The Shoebill to taunt her on a daily basis. Rock knows she's not allowed off the ship right now, and the cat picks up on this because Rock doesn't even try to chase him away.	“You’re different now,” Ull says, bringing Nell into the room serving as her temporary bedroom. “Am I?” Nell asks, refusing to look at him. “You seem disappointed to be back here. You knew I would find you, surely.” Nell gives no response. “Don’t tell me you actually believed he could help you.” Her hands ball into fists, then stop. “You belong to me, my precious Nehalennia. He would have used you or sold you off like everybody else wanted. Are you forgetting that I am the only one who truly wants to keep you?” “That doesn’t stop you from wanting to use me, just like they all did,” Nell answers. “Of course I want to use you. You are mine. It’s only natural to use what belongs to you, isn’t it? Even so, I care for you and wish to take care of you.” “I would rather be uncared for than have the perverse care of someone like you.” Ull grabs her by the hair again, lifts her off the ground, and pushes her against the wall. “I do not like defiant pets. You would be wise to destroy any lingering hope inside that childish mind of yours. You are mine, and unless you start over with a new body, that will never change. Do I make myself clear?” he asks, tossing her down onto the floor. Nell does not answer. Ull wastes no more time; he places heavy chains around her ankles and wrists. “I don’t normally mind you escaping, but until we deal with this new attitude of yours, you will not be permitted to waste my time.” With that, Ull leaves the room, posts two guards inside with her, and locks the door behind him. “Think he’ll be on soon?” Cassiel asks Serra. The two girls sit at the end of the pier. Cassiel dangles her legs off it, gently kicking back and forth, and Serra kneels behind the blonde, her chin resting on Cassiel’s head and her arms draped over her. Serra occasionally slips a hand to Cassiel’s chest; to her surprise, she only gets a blushing pout in response. “Yeah. He won’t miss this,” Serra answers. “He’s cutting it pretty close, that bastard. I swear if he’s late—” “He won’t be.” “He better not be. Oh, and I wanted to ask, what’s it like having so many parents?” “Lots of presents,” Serra says with a thumbs-up. “Between your parents and now me and Ryouta, don’t you think you’re a bit spoiled?” “I’m super spoiled,” Cassiel replies with a smile. “Hmph. I—I don’t get spoiled…” “Cass, Cass.” “What?” Cassiel asks, turning her head toward Serra now that her head is off hers. Serra plants a kiss on Cassiel’s lips. “I’ll spoil you whenever you want me to,” Serra says. “I—o-okay, maybe I get spoiled, too,” Cassiel says, cheeks flushed. “Oi, big guy, hand me that wrench,” Tabitha orders Bonekraka. Bonekraka tosses the wrench so it lands next to Tabitha, scratching the floorboard of the Shoebill. "Hey! I just cleaned this baby up! You gonna fix that yourself?!" "Nyet," Bonekraka answers, and a moment later a wrench hits the back of his head as a pissed off Tabitha glares at him from the engine. "Now come bring it to me the proper way!" Bonekraka turns to look at Tabitha with his hands balled into fists, but when he sees the look in her eyes he's reminded of his wife. The grumpy orc submissively takes the wrench over to her, handing it to her properly. "Thank ya. See, now was that so hard to do? Next time don't try sassin' me and scratchin' up my baby." "Da." "O Olly, you mustn't—we have such little time," Corwin says, looking both left and right to make sure nobody is coming down the alleyway where Oleander has him pressed against the wall. "Come on, Cor, we both know I can make you fire your cannon quickly enough," Oleander coos, running one hand up Corwin's chest while the other travels dangerously lower down his abdomen. "Shouldn't you be happy that you have a lover so eager to please you?" "I—I assure you that I am beyond happy, but what if somebody sees?" "Are you telling me you wouldn't like getting caught in the act?" Oleander's teasing fingers trace Corwin's body. Corwin gulps. "I... do not entirely dislike the thought, but—" "Then," Oleander pauses to sink his hands into Corwin's pants, "let's have some fun before the tournament starts." "We don't know when we're going to be able to do this again." "I'm sure we'll have time to in only a few hours! It's not as if we are leaving one another." "If you want me to stop, all you have to say is 'stop,'" Oleander says, looking up at Corwin with a smug grin. Corwin doesn't tell him to stop. Rock is in the middle of a stare-down with one of the stray cats that has been visiting The Shoebill to taunt her on a daily basis. Rock knows she's not allowed off the ship right now, and the cat picks up on this because Rock doesn't even try to chase him away.
long_en_250	news_en	1,080	en	March 10 (Reuters) - For other diaries, please see: Top Economic Events; Emerging Markets Economic Events; Government Debt Auctions; Political and General News; U.S. Federal Reserve. This diary is filed daily. MONDAY, MARCH 12 - Brussels: Participation by ECB President Mario Draghi and ECB Executive Board member Benoit Coeure in the Eurogroup meeting. - Brussels: Eurogroup meeting. TUESDAY, MARCH 13 - Brussels: Participation by ECB Vice‑President Vitor Constancio in the ECOFIN meeting. - Kingston, Ontario: Bank of Canada Governor Stephen Poloz speaks at Queen's University at 14:30 GMT. - London: UK Finance Minister Philip Hammond delivers spring statement on updated fiscal forecasts. - Brussels: Economic and Financial Affairs Council. - Tokyo: Bank of Japan releases minutes of the monetary policy meeting held on Jan 22-23 at 23:50 GMT. WEDNESDAY, MARCH 14 - Frankfurt: Address by ECB President Mario Draghi at "The ECB and Its Watchers XIX" conference organised by the Institute for Monetary and Financial Stability at 08:00 GMT. - Rome: Speech by ECB Supervisor Ignazio Angeloni at the annual conference "Tapering e final QE: gli effetti sugli asset in gestione" organised by Itinerari Previdenziali at 08:45 GMT. FRANKFURT - Participation by ECB chief economist Peter Praet in Debate 1 "Assessment of Quantitative Easing and Challenges of Policy Normalization" at The ECB and Its Watchers XIX conference organised by the Institute for Monetary and Financial Stability in Frankfurt am Main, Germany - 08:45 GMT. FRANKFURT - Participation by ECB Vice-President Vitor Constancio in Debate 2 "The Pursuit of Financial Stability and Tasks for Monetary, Regulatory and Macro-Prudential Policies" at The ECB and Its Watchers XIX conference organised by the Institute for Monetary and Financial Stability in Frankfurt am Main, Germany - 10:45 GMT. BERLIN - ECB executive board member Benoit Coeure in panel on "Nullzinspolitik" at CDU-Wirtschaftsrat: Finanzmarktklausur - 16:15 GMT. BERLIN - ECB director Benoit Coeure participates in an event with fintech companies organised by FinLeap - 07:30 GMT. FRANKFURT - ECB President Mario Draghi, chief economist Peter Praet, Vice President Vitor Constancio and Banque de France Governor Francois Villeroy de Galhau speak at The ECB and Its Watchers conference in Frankfurt. THURSDAY, MARCH 15 - FLORENCE, Italy - ECB board member Sabine Lautenschlaeger speaks at Florence School of Banking and Finance - 15:45 GMT. THURSDAY, MARCH 15 - BERN - Swiss National Bank (SNB) monetary policy assessment - 08:30 GMT. OSLO - Norges Bank to announce the Executive Board's interest rate decision and publish its monetary policy report, followed by a press conference at 0900 GMT. FRIDAY, MARCH 16 LONDON - Bank of England Financial Policy Committee issues a statement from its meeting at 0930 GMT. KRISTIANSUND, Norway - Norges Bank Deputy Governors Jon Nicolaisen and Egil Matsen give speeches to Norges Bank's Regional Network, Region North‑West. SUNDAY, MARCH 18 MIAMI - Federal Reserve Banks of Atlanta, Chicago and San Francisco co‑sponsor the conference "National Interagency Community Reinvestment Conference: Aligning to Build Resilient and Inclusive Communities." Speakers include Federal Reserve Bank of Atlanta President Raphael Bostic (through Mar. 21). TOKYO - Bank of Japan to release the summary of opinions from board members at its March 8–9 policy meeting at 2350 GMT. TUESDAY, MARCH 20 OSLO - Norges Bank Deputy Governor Egil Matsen gives a speech at a conference hosted by Finance Norway at 0955 GMT. WASHINGTON, D.C. - U.S. Federal Reserve's Federal Open Market Committee (FOMC) starts its two‑day meeting on interest rates. WEDNESDAY, MARCH 21 WASHINGTON, D.C. - U.S. Federal Reserve's Federal Open Market Committee (FOMC) announces its decision on interest rates, followed by a statement at 1800 GMT. WASHINGTON - U.S. Federal Reserve chairperson holds a news conference on the interest rate at 1830 GMT. FRANKFURT - ECB Governing Council meeting; no interest rate announcements scheduled. Thursday, March 22 — Toronto, Ontario: Bank of Canada Senior Deputy Governor Carolyn Wilkins speaks at the Rotman School of Management, 1900 GMT. Thursday, March 22 — Washington, D.C.: Governor of Norges Bank Oystein Olsen gives a speech in Washington. Thursday, March 22 — Frankfurt, Germany: ECB General Council meeting. Thursday, March 22 — Wellington: Reserve Bank of New Zealand issues Official Cash Rate (OCR) announcement. Thursday, March 22 — London: Bank of England announces its interest rate decision and publishes the minutes of the meeting, 1200 GMT. Friday, March 23 — New York: Federal Reserve Bank of Minneapolis President Neel Kashkari participates in a Q&A moderated by Kathleen Hays of Bloomberg, 1430 GMT. Friday, March 23 — Stockholm: Riksbank General Council meeting, 1200 GMT. Monday, March 26 — Vienna: German central bank chief Jens Weidmann delivers a speech, "New Impetus for Europe", 0930 GMT. Tuesday, March 27 — London: Bank of England's Record of the Financial Policy Committee meeting held on March 12 will be published, 0930 GMT. Tuesday, March 27 — Stockholm: Riksbank Executive Board meeting, 0700 GMT. Thursday, March 29 — New York: Federal Reserve Bank of Philadelphia President Patrick Harker speaks on the economic outlook before a New York Association of Business Economics luncheon, 1700 GMT. Thursday, April 5 — Zurich, Switzerland: Alternate member of the Governing Board of the Swiss National Bank Dewet Moser gives a speech, "Gestern und heute: Wandel am Geld- und Devisenmarkt", at the Money Market Event. Zurich, Switzerland - Speech by Member of the Governing Board of the Swiss National Bank Andrea Maechler, "Heute und morgen: Ein Blick in die digitale Zukunft", Money Market Event, Zurich. Friday, April 6 - London: Governor of Norges Bank Oystein Olsen will give a speech in London. Tuesday, April 10 - Oslo: Governor of Norges Bank Oystein Olsen will give a speech to foreign embassy representatives. Wednesday, April 11 - Frankfurt, Germany: ECB Governing Council meeting. No interest rate announcements scheduled. Wednesday, April 11 - Washington, D.C.: U.S. Federal Reserve Federal Open Market Committee (FOMC) will release minutes from its March 20-21 policy meeting, 1800 GMT. Thursday, April 12 - Stavanger, Norway: Norges Bank Governor Oystein Olsen and Deputy Governor Egil Matsen give speeches to the regional network (Region South-West) and lectures at the University of Stavanger. Thursday, April 12 - Oslo: Deputy Governor of Norges Bank Jon Nicolaisen speaks at the Norwegian Academy of Science and Letters. Friday, April 13 - Tromso, Norway: Norges Bank Governor Oystein Olsen and Deputy Governor Egil Matsen give speeches to the regional network (Region North) and in Tromso. Monday, April 16 - Stockholm: Riksbank executive board meeting, 0700 GMT. Wednesday, April 18 - Washington, D.C.: U.S. Federal Reserve issues its Beige Book on economic conditions, 1800 GMT. Wednesday, April 18 - Ottawa: Bank of Canada key policy interest rate announcement and monetary policy report, 1400 GMT.	March ten (Reuters) - For other diaries, please see: Top Economic Events; Emerging Markets Economic Events; Government Debt Auctions; Political and General News; U.S. Federal Reserve. This diary is filed daily. MONDAY, MARCH twelve - Brussels: Participation by ECB President Mario Draghi and ECB Executive Board member Benoit Coeure in the Eurogroup meeting. - Brussels: Eurogroup meeting. TUESDAY, MARCH thirteen - Brussels: Participation by ECB Vice‑President Vitor Constancio in the ECOFIN meeting. - Kingston, Ontario: Bank of Canada Governor Stephen Poloz speaks at Queen's University at fourteen thirty GMT. - London: UK Finance Minister Philip Hammond delivers spring statement on updated fiscal forecasts. - Brussels: Economic and Financial Affairs Council. - Tokyo: Bank of Japan releases minutes of the monetary policy meeting held on Jan twenty two to twenty three at twenty three fifty GMT. WEDNESDAY, MARCH fourteen - Frankfurt: Address by ECB President Mario Draghi at "The ECB and Its Watchers XIX" conference organised by the Institute for Monetary and Financial Stability at zero eight zero zero GMT. - Rome: Speech by ECB Supervisor Ignazio Angeloni at the annual conference "Tapering e final QE: gli effetti sugli asset in gestione" organised by Itinerari Previdenziali at zero eight forty five GMT. FRANKFURT - Participation by ECB chief economist Peter Praet in Debate one "Assessment of Quantitative Easing and Challenges of Policy Normalization" at The ECB and Its Watchers XIX conference organised by the Institute for Monetary and Financial Stability in Frankfurt am Main, Germany - zero eight:forty five GMT. FRANKFURT - Participation by ECB Vice-President Vitor Constancio in Debate two "The Pursuit of Financial Stability and Tasks for Monetary, Regulatory and Macro-Prudential Policies" at The ECB and Its Watchers XIX conference organised by the Institute for Monetary and Financial Stability in Frankfurt am Main, Germany - ten:forty five GMT. BERLIN - ECB executive board member Benoit Coeure in panel on "Nullzinspolitik" at CDU-Wirtschaftsrat: Finanzmarktklausur - sixteen:fifteen GMT. BERLIN - ECB director Benoit Coeure participates in an event with fintech companies organised by FinLeap - zero seven:thirty GMT. FRANKFURT - ECB President Mario Draghi, chief economist Peter Praet, Vice President Vitor Constancio and Banque de France Governor Francois Villeroy de Galhau speak at The ECB and Its Watchers conference in Frankfurt. THURSDAY, MARCH fifteen - FLORENCE, Italy - ECB board member Sabine Lautenschlaeger speaks at Florence School of Banking and Finance - fifteen:forty five GMT. THURSDAY, MARCH fifteen - BERN - Swiss National Bank (SNB) monetary policy assessment - zero eight:thirty GMT. OSLO - Norges Bank to announce the Executive Board's interest rate decision and publish its monetary policy report, followed by a press conference at zero nine zero zero GMT. FRIDAY, MARCH sixteen LONDON - Bank of England Financial Policy Committee issues a statement from its meeting at zero nine three zero GMT. KRISTIANSUND, Norway - Norges Bank Deputy Governors Jon Nicolaisen and Egil Matsen give speeches to Norges Bank's Regional Network, Region North‑West. SUNDAY, MARCH eighteen MIAMI - Federal Reserve Banks of Atlanta, Chicago and San Francisco co‑sponsor the conference "National Interagency Community Reinvestment Conference: Aligning to Build Resilient and Inclusive Communities." Speakers include Federal Reserve Bank of Atlanta President Raphael Bostic (through Mar. twenty one). TOKYO - Bank of Japan to release the summary of opinions from board members at its March eight to nine policy meeting at two three five zero GMT. TUESDAY, MARCH twenty OSLO - Norges Bank Deputy Governor Egil Matsen gives a speech at a conference hosted by Finance Norway at zero nine five five GMT. WASHINGTON, D.C. - U.S. Federal Reserve's Federal Open Market Committee (FOMC) starts its two‑day meeting on interest rates. WEDNESDAY, MARCH twenty one WASHINGTON, D.C. - U.S. Federal Reserve's Federal Open Market Committee (FOMC) announces its decision on interest rates, followed by a statement at one eight zero zero GMT. WASHINGTON - U.S. Federal Reserve chairperson holds a news conference on the interest rate at one eight three zero GMT. FRANKFURT - ECB Governing Council meeting; no interest rate announcements scheduled. Thursday, March twenty two — Toronto, Ontario: Bank of Canada Senior Deputy Governor Carolyn Wilkins speaks at the Rotman School of Management, one thousand nine hundred GMT. Thursday, March twenty two — Washington, D.C.: Governor of Norges Bank Oystein Olsen gives a speech in Washington. Thursday, March twenty two — Frankfurt, Germany: ECB General Council meeting. Thursday, March twenty two — Wellington: Reserve Bank of New Zealand issues Official Cash Rate (OCR) announcement. Thursday, March twenty two — London: Bank of England announces its interest rate decision and publishes the minutes of the meeting, one thousand two hundred GMT. Friday, March twenty three — New York: Federal Reserve Bank of Minneapolis President Neel Kashkari participates in a Q&A moderated by Kathleen Hays of Bloomberg, one thousand four hundred thirty GMT. Friday, March twenty three — Stockholm: Riksbank General Council meeting, one thousand two hundred GMT. Monday, March twenty six — Vienna: German central bank chief Jens Weidmann delivers a speech, "New Impetus for Europe", nine hundred thirty GMT. Tuesday, March twenty seven — London: Bank of England's Record of the Financial Policy Committee meeting held on March twelve will be published, nine hundred thirty GMT. Tuesday, March twenty seven — Stockholm: Riksbank Executive Board meeting, seven hundred GMT. Thursday, March twenty nine — New York: Federal Reserve Bank of Philadelphia President Patrick Harker speaks on the economic outlook before a New York Association of Business Economics luncheon, one thousand seven hundred GMT. Thursday, April five — Zurich, Switzerland: Alternate member of the Governing Board of the Swiss National Bank Dewet Moser gives a speech, "Gestern und heute: Wandel am Geld- und Devisenmarkt", at the Money Market Event. Zurich, Switzerland - Speech by Member of the Governing Board of the Swiss National Bank Andrea Maechler, "Heute und morgen: Ein Blick in die digitale Zukunft", Money Market Event, Zurich. Friday, April six - London: Governor of Norges Bank Oystein Olsen will give a speech in London. Tuesday, April ten - Oslo: Governor of Norges Bank Oystein Olsen will give a speech to foreign embassy representatives. Wednesday, April eleven - Frankfurt, Germany: ECB Governing Council meeting. No interest rate announcements scheduled. Wednesday, April eleven - Washington, D.C.: U.S. Federal Reserve Federal Open Market Committee (FOMC) will release minutes from its March twenty- twenty one policy meeting, one eight zero zero GMT. Thursday, April twelve - Stavanger, Norway: Norges Bank Governor Oystein Olsen and Deputy Governor Egil Matsen give speeches to the regional network (Region South-West) and lectures at the University of Stavanger. Thursday, April twelve - Oslo: Deputy Governor of Norges Bank Jon Nicolaisen speaks at the Norwegian Academy of Science and Letters. Friday, April thirteen - Tromso, Norway: Norges Bank Governor Oystein Olsen and Deputy Governor Egil Matsen give speeches to the regional network (Region North) and in Tromso. Monday, April sixteen - Stockholm: Riksbank executive board meeting, zero seven zero zero GMT. Wednesday, April eighteen - Washington, D.C.: U.S. Federal Reserve issues its Beige Book on economic conditions, eighteen hundred GMT. Wednesday, April eighteen - Ottawa: Bank of Canada key policy interest rate announcement and monetary policy report, fourteen hundred GMT.
long_en_176	paper_en	1,680	en	We build upon the DeepSeek-V3 pipeline and adopt a similar distribution of preference pairs and training prompts. For helpfulness, we focus exclusively on the final summary, ensuring that the assessment emphasizes the utility and relevance of the response to the user while minimizing interference with the underlying reasoning process. For harmlessness, we evaluate the entire response of the model, including both the reasoning process and the summary, to identify and mitigate any potential risks, biases, or harmful content that may arise during the generation process. Ultimately, the integration of reward signals and diverse data distributions enables us to train a model that excels in reasoning while prioritizing helpfulness and harmlessness. Distillation: Empower Small Models with Reasoning Capability To equip more efficient smaller models with reasoning capabilities like DeepSeek-R1, we directly fine-tuned open-source models like Qwen and Llama using the 800k samples curated with DeepSeek-R1, as detailed in Section 2.3.3. Our findings indicate that this straightforward distillation method significantly enhances the reasoning abilities of smaller models. The base models we use here are Qwen2.5-Math-1.5B, Qwen2.5-Math-7B, Qwen2.5-14B, Qwen2.5-32B, Llama-3.1-8B, and Llama-3.3-70B-Instruct. We select Llama-3.3 because its reasoning capability is slightly better than that of Llama-3.1. For distilled models, we apply only SFT and do not include an RL stage, even though incorporating RL could substantially boost model performance. Our primary goal here is to demonstrate the effectiveness of the distillation technique, leaving the exploration of the RL stage to the broader research community. Section 3: Experiment Benchmarks We evaluate models on MMLU, MMLU-Redux, MMLU-Pro, C-Eval, and CMMLU, IFEval, FRAMES, GPQA Diamond, SimpleQA, C-SimpleQA, SWE-Bench Verified, Aider, LiveCodeBench (2024-08 -- 2025-01), Codeforces, Chinese National High School Mathematics Olympiad (CNMO 2024), and American Invitational Mathematics Examination 2024 (AIME 2024). In addition to standard benchmarks, we also evaluate our models on open-ended generation tasks using LLMs as judges. Specifically, we adhere to the original configurations of AlpacaEval 2.0 and Arena-Hard, which leverage GPT-4-Turbo-1106 as judges for pairwise comparisons. Here, we only feed the final summary to evaluation to avoid the length bias. For distilled models, we report representative results on AIME 2024, MATH-500, GPQA Diamond, Codeforces, and LiveCodeBench. Evaluation Prompts Following the setup in DeepSeek-V3, standard benchmarks such as MMLU, DROP, GPQA Diamond, and SimpleQA are evaluated using prompts from the simple-evals framework. For MMLU-Redux, we adopt the Zero-Eval prompt format in a zero-shot setting. In terms of MMLU-Pro, C-Eval and CLUE-WSC, since the original prompts are few-shot, we slightly modify the prompt to the zero-shot setting. The CoT in few-shot may hurt the performance of DeepSeek-R1. Other datasets follow their original evaluation protocols with default prompts provided by their creators. For code and math benchmarks, the HumanEval-Mul dataset covers eight mainstream programming languages (Python, Java, C++, C#, JavaScript, TypeScript, PHP, and Bash). Model performance on LiveCodeBench is evaluated using CoT format, with data collected between August 2024 and January 2025. The Codeforces dataset is evaluated using problems from 10 Div.2 contests along with expert-crafted test cases, after which the expected ratings and percentages of competitors are calculated. SWE-Bench verified results are obtained via the agentless framework. AIDER-related benchmarks are measured using a "diff" format. DeepSeek-R1 outputs are capped at a maximum of 32,768 tokens for each benchmark. Baselines We conduct comprehensive evaluations against several strong baselines, including DeepSeek-V3, Claude-Sonnet-3.5-1022, GPT-4o-0513, OpenAI-o1-mini, and OpenAI-o1-1217. Since accessing the OpenAI-o1-1217 API is challenging in mainland China, we report its performance based on official reports. For distilled models, we also compare the open-source model QwQ-32B-Preview. Evaluation Setup We set the maximum generation length to 32,768 tokens for the models. We found that using greedy decoding to evaluate long-output reasoning models results in higher repetition rates and significant variability across different checkpoints. Therefore, we default to pass at k evaluation and report pass at 1 using a non-zero temperature. Specifically, we use a sampling temperature of 0.6 and a top-p value of 0.95 to generate k responses (typically between 4 and 64, depending on the test set size) for each question. Pass at 1 is then calculated. This method provides more reliable performance estimates. For AIME 2024, we also report consensus (majority vote) results using 64 samples, denoted as consensus at 64. DeepSeek-R1 Evaluation For education-oriented knowledge benchmarks such as MMLU, MMLU-Pro, and GPQA Diamond, DeepSeek-R1 demonstrates superior performance compared to DeepSeek-V3. This improvement is primarily attributed to enhanced accuracy in STEM-related questions, where significant gains are achieved through large-scale reinforcement learning. Additionally, DeepSeek-R1 excels on FRAMES, a long-context-dependent QA task, showcasing its strong document analysis capabilities. This highlights the potential of reasoning models in AI-driven search and data analysis tasks. On the factual benchmark SimpleQA, DeepSeek-R1 outperforms DeepSeek-V3, demonstrating its capability in handling fact-based queries. A similar trend is observed where OpenAI-o1 surpasses GPT-4o on this benchmark. However, DeepSeek-R1 performs worse than DeepSeek-V3 on the Chinese SimpleQA benchmark, primarily due to its tendency to refuse answering certain queries after safety RL. Without safety RL, DeepSeek-R1 could achieve an accuracy of over 70%. DeepSeek-R1 also delivers impressive results on IF-Eval, a benchmark designed to assess a model's ability to follow format instructions. These improvements can be linked to the inclusion of instruction-following data during the final stages of supervised fine-tuning (SFT) and RL training. Furthermore, remarkable performance is observed on AlpacaEval2.0 and ArenaHard, indicating DeepSeek-R1’s strengths in writing tasks and open-domain question answering. Its significant outperformance of DeepSeek-V3 underscores the generalization benefits of large-scale RL, which not only boosts reasoning capabilities but also improves performance across diverse domains. Moreover, the summary lengths generated by DeepSeek-R1 are concise, with an average of 689 tokens on ArenaHard and 2,218 characters on AlpacaEval 2.0. This indicates that DeepSeek-R1 avoids introducing length bias during GPT-based evaluations, further solidifying its robustness across multiple tasks. On math tasks, DeepSeek-R1 demonstrates performance on par with OpenAI-o1-1217, surpassing other models by a large margin. A similar trend is observed on coding algorithm tasks, such as LiveCodeBench and Codeforces, where reasoning-focused models dominate these benchmarks. On engineering-oriented coding tasks, OpenAI-o1-1217 outperforms DeepSeek-R1 on Aider but achieves comparable performance on SWE Verified. We believe the engineering performance of DeepSeek-R1 will improve in the next version, as the amount of related RL training data currently remains very limited. Distilled Model Evaluation Simply distilling DeepSeek-R1's outputs enables the efficient DeepSeek-R1-7B (i.e., DeepSeek-R1-Distill-Qwen-7B, abbreviated similarly below) to outperform non-reasoning models like GPT-4o-0513 across the board. DeepSeek-R1-14B surpasses QwQ-32B-Preview on all evaluation metrics, while DeepSeek-R1-32B and DeepSeek-R1-70B significantly exceed o1-mini on most benchmarks. These results demonstrate the strong potential of distillation. Additionally, we found that applying RL to these distilled models yields significant further gains. We believe this warrants further exploration and therefore present only the results of the simple SFT-distilled models here. Section 4: Discussion Distillation v.s. Reinforcement Learning In Section 3.2, we can see that by distilling DeepSeek-R1, the small model can achieve impressive results. However, there is still one question left: can the model achieve comparable performance through the large-scale RL training discussed in the paper without distillation? To answer this question, we conduct large-scale RL training on Qwen-32B-Base using math, code, and STEM data, training for over 10K steps, resulting in DeepSeek-R1-Zero-Qwen-32B. The experimental results demonstrate that the 32B base model, after large-scale RL training, achieves performance on par with QwQ-32B-Preview. However, DeepSeek-R1-Distill-Qwen-32B, which is distilled from DeepSeek-R1, performs significantly better than DeepSeek-R1-Zero-Qwen-32B across all benchmarks. Therefore, we can draw two conclusions: First, distilling more powerful models into smaller ones yields excellent results, whereas smaller models relying on the large-scale RL mentioned in this paper require enormous computational power and may not even achieve the performance of distillation. Second, while distillation strategies are both economical and effective, advancing beyond the boundaries of intelligence may still require more powerful base models and larger-scale reinforcement learning. Unsuccessful Attempts In the early stages of developing DeepSeek-R1, we also encountered failures and setbacks along the way. We share our failure experiences here to provide insights, but this does not imply that these approaches are incapable of developing effective reasoning models. Process Reward Model (PRM) PRM is a reasonable method to guide the model toward better approaches for solving reasoning tasks. However, in practice, PRM has three main limitations that may hinder its ultimate success. First, it is challenging to explicitly define a fine-grain step in general reasoning. Second, determining whether the current intermediate step is correct is a challenging task. Automated annotation using models may not yield satisfactory results, while manual annotation is not conducive to scaling up. Third, once a model-based PRM is introduced, it inevitably leads to reward hacking, and retraining the reward model needs additional training resources and it complicates the whole training pipeline. In conclusion, while PRM demonstrates a good ability to rerank the top-N responses generated by the model or assist in guided search, its advantages are limited compared to the additional computational overhead it introduces during the large-scale reinforcement learning process in our experiments. Monte Carlo Tree Search (MCTS) Inspired by AlphaGo and AlphaZero, we explored using Monte Carlo Tree Search (MCTS) to enhance test-time compute scalability.	We build upon the DeepSeek V three pipeline and adopt a similar distribution of preference pairs and training prompts. For helpfulness, we focus exclusively on the final summary, ensuring that the assessment emphasizes the utility and relevance of the response to the user while minimizing interference with the underlying reasoning process. For harmlessness, we evaluate the entire response of the model, including both the reasoning process and the summary, to identify and mitigate any potential risks, biases, or harmful content that may arise during the generation process. Ultimately, the integration of reward signals and diverse data distributions enables us to train a model that excels in reasoning while prioritizing helpfulness and harmlessness. Distillation: Empower Small Models with Reasoning Capability To equip more efficient smaller models with reasoning capabilities like DeepSeek R one, we directly fine-tuned open-source models like Qwen and Llama using the eight hundred thousand samples curated with DeepSeek R one, as detailed in Section two point three point three. Our findings indicate that this straightforward distillation method significantly enhances the reasoning abilities of smaller models. The base models we use here are Qwen two point five Math one point five B, Qwen two point five Math seven B, Qwen two point five fourteen B, Qwen two point five thirty two B, Llama three point one eight B, and Llama three point three seventy B Instruct. We select Llama three point three because its reasoning capability is slightly better than that of Llama three point one. For distilled models, we apply only SFT and do not include an RL stage, even though incorporating RL could substantially boost model performance. Our primary goal here is to demonstrate the effectiveness of the distillation technique, leaving the exploration of the RL stage to the broader research community. Section three: Experiment Benchmarks We evaluate models on MMLU, MMLU-Redux, MMLU-Pro, C-Eval, and CMMLU, IFEval, FRAMES, GPQA Diamond, SimpleQA, C-SimpleQA, SWE-Bench Verified, Aider, LiveCodeBench (two thousand twenty four zero eight -- two thousand twenty five zero one), Codeforces, Chinese National High School Mathematics Olympiad (CNMO two thousand twenty four), and American Invitational Mathematics Examination two thousand twenty four (AIME two thousand twenty four). In addition to standard benchmarks, we also evaluate our models on open-ended generation tasks using LLMs as judges. Specifically, we adhere to the original configurations of AlpacaEval two point zero and Arena-Hard, which leverage GPT four Turbo one thousand one hundred six as judges for pairwise comparisons. Here, we only feed the final summary to evaluation to avoid the length bias. For distilled models, we report representative results on AIME two thousand twenty four, MATH five hundred, GPQA Diamond, Codeforces, and LiveCodeBench. Evaluation Prompts Following the setup in DeepSeek V three, standard benchmarks such as MMLU, DROP, GPQA Diamond, and SimpleQA are evaluated using prompts from the simple-evals framework. For MMLU-Redux, we adopt the Zero-Eval prompt format in a zero-shot setting. In terms of MMLU-Pro, C-Eval and CLUE-WSC, since the original prompts are few-shot, we slightly modify the prompt to the zero-shot setting. The CoT in few-shot may hurt the performance of DeepSeek R one. Other datasets follow their original evaluation protocols with default prompts provided by their creators. For code and math benchmarks, the HumanEval-Mul dataset covers eight mainstream programming languages (Python, Java, C plus plus, C sharp, JavaScript, TypeScript, PHP, and Bash). Model performance on LiveCodeBench is evaluated using CoT format, with data collected between August two thousand twenty four and January two thousand twenty five. The Codeforces dataset is evaluated using problems from ten Div. two contests along with expert-crafted test cases, after which the expected ratings and percentages of competitors are calculated. SWE-Bench verified results are obtained via the agentless framework. AIDER-related benchmarks are measured using a "diff" format. DeepSeek R one outputs are capped at a maximum of thirty two thousand seven hundred sixty eight tokens for each benchmark. Baselines We conduct comprehensive evaluations against several strong baselines, including DeepSeek V three, Claude-Sonnet three point five one thousand twenty two, GPT four o five hundred thirteen, OpenAI o one mini, and OpenAI o one one thousand two hundred seventeen. Since accessing the OpenAI o one one thousand two hundred seventeen API is challenging in mainland China, we report its performance based on official reports. For distilled models, we also compare the open-source model QwQ thirty two B Preview. Evaluation Setup We set the maximum generation length to thirty two thousand seven hundred sixty eight tokens for the models. We found that using greedy decoding to evaluate long-output reasoning models results in higher repetition rates and significant variability across different checkpoints. Therefore, we default to pass at k evaluation and report pass at one using a non-zero temperature. Specifically, we use a sampling temperature of zero point six and a top-p value of zero point nine five to generate k responses (typically between four and sixty four, depending on the test set size) for each question. Pass at one is then calculated. This method provides more reliable performance estimates. For AIME two thousand twenty four, we also report consensus (majority vote) results using sixty four samples, denoted as consensus at sixty four. DeepSeek-R1 Evaluation For education-oriented knowledge benchmarks such as MMLU, MMLU-Pro, and GPQA Diamond, DeepSeek-R1 demonstrates superior performance compared to DeepSeek-V3. This improvement is primarily attributed to enhanced accuracy in STEM-related questions, where significant gains are achieved through large-scale reinforcement learning. Additionally, DeepSeek-R1 excels on FRAMES, a long-context-dependent QA task, showcasing its strong document analysis capabilities. This highlights the potential of reasoning models in AI-driven search and data analysis tasks. On the factual benchmark SimpleQA, DeepSeek-R1 outperforms DeepSeek-V3, demonstrating its capability in handling fact-based queries. A similar trend is observed where OpenAI-o1 surpasses GPT-4o on this benchmark. However, DeepSeek R one performs worse than DeepSeek V three on the Chinese SimpleQA benchmark, primarily due to its tendency to refuse answering certain queries after safety RL. Without safety RL, DeepSeek R one could achieve an accuracy of over seventy percent. DeepSeek R one also delivers impressive results on IF Eval, a benchmark designed to assess a model's ability to follow format instructions. These improvements can be linked to the inclusion of instruction following data during the final stages of supervised fine tuning (SFT) and RL training. Furthermore, remarkable performance is observed on AlpacaEval two point zero and ArenaHard, indicating DeepSeek R one’s strengths in writing tasks and open-domain question answering. Its significant outperformance of DeepSeek V three underscores the generalization benefits of large scale RL, which not only boosts reasoning capabilities but also improves performance across diverse domains. Moreover, the summary lengths generated by DeepSeek R one are concise, with an average of six hundred eighty nine tokens on ArenaHard and two thousand two hundred eighteen characters on AlpacaEval two point zero. This indicates that DeepSeek R one avoids introducing length bias during GPT based evaluations, further solidifying its robustness across multiple tasks. On math tasks, DeepSeek R one demonstrates performance on par with OpenAI o one one thousand two hundred seventeen, surpassing other models by a large margin. A similar trend is observed on coding algorithm tasks, such as LiveCodeBench and Codeforces, where reasoning focused models dominate these benchmarks. On engineering-oriented coding tasks, OpenAI-o one- one thousand two hundred seventeen outperforms DeepSeek-R one on Aider but achieves comparable performance on SWE Verified. We believe the engineering performance of DeepSeek-R one will improve in the next version, as the amount of related RL training data currently remains very limited. Distilled Model Evaluation Simply distilling DeepSeek-R one's outputs enables the efficient DeepSeek-R one- seven B (i.e., DeepSeek-R one- Distill- Qwen- seven B, abbreviated similarly below) to outperform non-reasoning models like GPT-four o-zero five one three across the board. DeepSeek-R one- fourteen B surpasses QwQ- thirty two B-Preview on all evaluation metrics, while DeepSeek-R one- thirty two B and DeepSeek-R one- seventy B significantly exceed o one-mini on most benchmarks. These results demonstrate the strong potential of distillation. Additionally, we found that applying RL to these distilled models yields significant further gains. We believe this warrants further exploration and therefore present only the results of the simple SFT-distilled models here. Section four: Discussion Distillation v.s. Reinforcement Learning In Section three point two, we can see that by distilling DeepSeek-R one, the small model can achieve impressive results. However, there is still one question left: can the model achieve comparable performance through the large scale RL training discussed in the paper without distillation? To answer this question, we conduct large scale RL training on Qwen thirty two B Base using math, code, and STEM data, training for over ten thousand steps, resulting in DeepSeek R one Zero Qwen thirty two B. The experimental results demonstrate that the thirty two B base model, after large scale RL training, achieves performance on par with QwQ thirty two B Preview. However, DeepSeek R one Distill Qwen thirty two B, which is distilled from DeepSeek R one, performs significantly better than DeepSeek R one Zero Qwen thirty two B across all benchmarks. Therefore, we can draw two conclusions: First, distilling more powerful models into smaller ones yields excellent results, whereas smaller models relying on the large scale RL mentioned in this paper require enormous computational power and may not even achieve the performance of distillation. Second, while distillation strategies are both economical and effective, advancing beyond the boundaries of intelligence may still require more powerful base models and larger scale reinforcement learning. Unsuccessful Attempts In the early stages of developing DeepSeek R one, we also encountered failures and setbacks along the way. We share our failure experiences here to provide insights, but this does not imply that these approaches are incapable of developing effective reasoning models. Process Reward Model (PRM) PRM is a reasonable method to guide the model toward better approaches for solving reasoning tasks. However, in practice, PRM has three main limitations that may hinder its ultimate success. First, it is challenging to explicitly define a fine-grain step in general reasoning. Second, determining whether the current intermediate step is correct is a challenging task. Automated annotation using models may not yield satisfactory results, while manual annotation is not conducive to scaling up. Third, once a model-based PRM is introduced, it inevitably leads to reward hacking, and retraining the reward model needs additional training resources and it complicates the whole training pipeline. In conclusion, while PRM demonstrates a good ability to rerank the top-N responses generated by the model or assist in guided search, its advantages are limited compared to the additional computational overhead it introduces during the large-scale reinforcement learning process in our experiments. Monte Carlo Tree Search (MCTS) Inspired by AlphaGo and AlphaZero, we explored using Monte Carlo Tree Search (MCTS) to enhance test-time compute scalability.
long_en_167	paper_en	2,405	en	In the MMT-Bench evaluation, which assesses advanced reasoning and instruction following across 32 core meta-tasks and 162 subtasks in multimodal understanding, Qwen2-VL-72B achieves 71.7, markedly surpassing the previous best (63.4) and demonstrating its prowess in applying expert knowledge and executing deliberate visual recognition, localization, reasoning, and planning. On MMBench, which evaluates fine-grained abilities across 20 dimensions, Qwen2-VL-72B exhibits strong performance, achieving 86.5 on the English test set, matching the state-of-the-art, and 86.6 on the Chinese test set, establishing a new benchmark. For MME, which measures a wide spectrum of perception and cognition abilities across 14 subtasks, Qwen2-VL-72B achieves a cumulative score of 2482.7, significantly outperforming the previous best (2414.7), underscoring its advanced capabilities in both visual perception and high-level cognition tasks. These comprehensive results underscore the Qwen2-VL series' exceptional proficiency in general visual question answering tasks. The models demonstrate advanced capabilities in real-world spatial comprehension, genuine multimodal integration, complex reasoning, instruction following, and a broad range of perception and cognition tasks. The consistent superior performance across diverse benchmarks, particularly the outstanding results of the 72B model, positions the Qwen2-VL series as a leading solution in the field of visual question answering. Our models excel in handling visually indispensable tasks, integrating core vision-language capabilities, and demonstrating expertise across diverse multimodal scenarios, ranging from fundamental perception tasks to complex reasoning and planning. This exhaustive evaluation highlights the Qwen2-VL series' versatility and effectiveness in addressing the multifaceted challenges posed by state-of-the-art multimodal benchmarks, thereby setting a new standard for large vision-language models. Document and Diagrams Reading We tested our model's OCR and document and diagram comprehension on DocVQA, ChartQA, InfoVQA, TextVQA, AI2D datasets. The DocVQA/InfoVQA/ChartQA dataset focuses on the model's ability to comprehend text in documents/high-resolution infographics/charts, while the TextVQA dataset examines the ability to comprehend text in naturalistic images. The OCRBench dataset is a a dataset of mixed tasks, which focuses on mathematical formula parsing and information extraction in addition to the text-based VQA. The AI2D dataset focuses on multiple-choice questions on scientific diagrams containing text. In addition, we also tested the OCR and formula recognition capabilities of our model on OCRBench, as well as the multilingual OCR capabilities of our model on the MTVQA dataset. The experimental results show that our model achieves SoTA level in several metrics, including DocVQA, InfoVQA, TextVQA and OCRBench, demonstrating that our model has good comprehension of textual content in images from multiple domains. Multilingual Text Recognition and Understanding In particular, our model surpasses all existing general-purpose LVLMs in multilingual OCR. Our model not only outperforms existing LVLMs (including proprietary models such as GPT-4o, Claude 3.5 Sonnet, etc.) on the public-available MTVQA dataset, it also outperforms GPT-4o on the in-house internal benchmark across all foreign languages except Arabic. Mathematical Reasoning We've conducted experiments on the MathVista and MathVision datasets to assess mathematical reasoning capabilities. MathVista is a comprehensive benchmark featuring 6,141 diverse examples of mathematical and visual tasks. The MathVision dataset comprises 3,040 math problems embedded in visual contexts from actual math competitions, covering 16 mathematical disciplines and varying in difficulty across five levels. These challenges underscore the necessity for LVLMs to exhibit strong visual comprehension, a deep understanding of mathematics, and sound logical reasoning skills. The Qwen2-VL series has demonstrated superior performance on MathVista, achieving a 70.5 outperforming other LVLMs. Additionally, it has set a new open-source benchmark on MathVision with 25.9. Referring Expression Comprehension Regarding visual localization task, we evaluate Qwen2-VL on RefCOCO, RefCOCO+, and RefCOCOg datasets. The results demonstrate that Qwen2-VL attains top-tier results among generalist models. Benefiting from a more rational structure design, Qwen2-VL is able to perceive details in high-resolution images, leading to significant improvements over Qwen-VL. The superiority of these models in comparison to both generalist and specialized models highlights their potential for advancing the field of visual localization and their capacity for real-world implementation in tasks requiring precise visual understanding. Video Understanding We evaluate our models on various video understanding tasks, with related benchmarks covering short videos of a few seconds to long videos of up to one hour. Qwen2-VL demonstrates strong results across 2B, 7B, and 72B sizes, with Qwen2-VL-72B achieving the best performance on MVBench, PerceptionTest, and EgoSchema. This showcases Qwen2-VL's superior capabilities in video understanding tasks, and scaling up Qwen2-VL yields significant improvements. For the challenging Video-MME benchmark, which includes videos up to one hour, it is noteworthy that we limited the maximum number of frames extracted per video to 768 during evaluation, potentially impacting performance on longer videos. Future work will focus on extending Qwen2-VL to support longer sequences, thereby accommodating longer videos. Visual Agent Qwen2-VL is evaluated first for its ability to interact with the environment via function calls and then for its capacity to complete complex sequential decision tasks through multiple rounds of interaction. The implementation is based on the Qwen-Agent framework. Function Calling Unlike function calling in LLMs, function calling in LVLMs often involves extracting information from visual cues. Due to the absence of public benchmarks for evaluating the capabilities of LVLMs in function calling, we constructed our internal evaluation dataset. To construct the evaluation dataset, we undertook the following procedures: Scene Categorization, Image Collection, Image Content Extraction, and Question/Functions/Arguments Generation. Firstly, we classified scenes into categories based on different visual applications. Subsequently, we downloaded and meticulously selected high-quality, representative images from the internet for each category. Thereafter, utilizing an advanced LVLM, we analyzed each image to extract key visual elements and textual information. Finally, based on the content information from the images, we used an advanced LLM to generate a series of questions that required specific functions to answer, along with specifying the input parameters needed for these function calls. Similar to the function calling evaluation method in LLMs, we designed two metrics to evaluate the accuracy of the function selection and the correctness of the arguments input. Specifically, Type Match(TM), is calculated as the ratio of times the model successfully invoked the correct function to the total number of calls attempted. Exact Match(EM), for each function calling, we checked whether the arguments passed to the function exactly matched those recorded in the image's content information, calculating this correctness ratio. The performance of Qwen2-VL in both Type Match(93.1 vs. 90.2) and Exact Match(53.2 vs. 50.0) over GPT-4o substantiates the efficacy of Qwen2-VL's capability in function calling, thereby underscoring its significant potential for application expansion through external tool integration. The evaluation results demonstrated that GPT-4o underperformed, primarily due to two factors: in scenarios where uncertainty arises, GPT-4o demonstrates a conservative approach by avoiding using external tools. The Optical Character Recognition (OCR) capability of GPT-4o is outperformed by Qwen2-VL, particularly in the context of Chinese characters. UI Operations/Games/Robotics/Navigation To assess Qwen2-VL’s ability to generally handle complex tasks, we conduct evaluations across multiple VL agent tasks, including mobile operations, robotic control, card games, and vision-language navigation. As these tasks need multiple actions to complete tasks, we keep the history (observation, action) through Qwen2-VL supports a 32K context length, then append each new observation image after every action, enabling continuous reasoning about subsequent steps. UI Operations: we evaluate Qwen2-VL using the AITZ task, which constructs a core clean test set derived from AITW. Based on common operation patterns of phone, we define actions such as tap, input and swipe for Qwen2-VL to interact with on-screen icons for task completion. For example, when Qwen2-VL is tasked with finding a pizza restaurant nearby by Google Maps, it should input "pizza" in the search term, swipe to select the appropriate restaurant, and tap the corresponding link. Following the AITZ setting, we report both type match (correctness of tap, input, or swipe) and exact match (correctness of tap location, input text, or swipe direction). With the support of grounding capability on UI, Qwen2-VL surpasses GPT-4 and previous SoTA. Robotic Control: we evaluate Qwen2-VL on the ALFRED task in AI2THOR. The task requires agent to perform complex household tasks, such as toasting bread and slicing an apple to prepare a meal. To work in the virtual environment, we define high-level actions (GotoLocation, Pickup, PutDown, Open, Close, Clean, Heat, Cool, Slice) as the action set. Moreover, agent needs to localize objects for manipulation (e.g., it can only pick up an apple if the apple is recognized). To improve the accuracy of manipulation, we integrate SAM. ALFRED task reports task success rate (SR) (e.g., preparing dinner) and sub-goal completion metrics (GC) (e.g., whether the bread is toasted or the apple is sliced). Qwen2-VL slightly outperforms the previously specialized model ThinkBot on the valid-unseen set. Card Games: we leverage the card game environment from RL4VLM to assess Qwen2-VL's performance in a series of card-based games: Number Line, BlackJack, EZPoint, and Point24. Each game presents distinct challenges: (1) reaching a target number using +1 or -1 operations, (2) drawing or holding cards to compete against the dealer, (3) applying basic arithmetic operations to reach a total of 12, and (4) using arithmetic operations to achieve a total of 24. We report the success rate of the tasks. They not only evaluate agent capabilities but also require strong OCR skills to recognize these cards and understand the progression of the game. Qwen2-VL demonstrates superior performance across all tasks. Vision-Language Navigation: we evaluate Qwen2-VL on the Vision-and-Language Navigation (VLN) task using the R2R and REVERIE. In VLN, the model must autonomously determine the next location based on instruction, current observations. We report the success rate (SR) of VLM in reaching the predetermined destination for this task. The performance of Qwen2-VL is comparable to that of GPT-4o, but both models fall significantly behind current specialized VLN models. We attribute this gap to the incomplete and unstructured map information generated by the model from multiple images. Accurately modeling maps and locations in a 3D environment remains a major challenge for multimodal models. Ablation Study In this section, we present ablation studies on image dynamic resolution, M-RoPE, and model scale. These experiments aim to provide insights into the impact of these key components on our model's performance. Dynamic Resolution We compare the performance between dynamic resolution and fixed resolution. For fixed resolution, we resize the images to ensure a constant number of image tokens being input to the model, rather than resizing to a specific height and width, as this would distort the original aspect ratio. For dynamic resolution, we only set min_pixels=100 by 28 by 28 and max_pixels=16384 by 28 by 28, allowing the number of image tokens depend primarily on the image's native resolution. It can be observed that adjusting image sizes only results in small perturbations in performance, demonstrating the model robustness to varying image sizes. Moreover, dynamic resolution approach is more efficient. We can observe that no single fixed resolution achieves optimal performance across all benchmarks. In contrast, the dynamic resolution approach consistently achieves top-tier performance while consuming fewer tokens on average. Additionally, we observe that merely increasing the image size does not always lead to improved performance. It is more important to choose an appropriate resolution for different images. We upscale small images to surpass a specified min_pixels threshold. Evaluations on upscaled images shows enhanced performance on perceptual tasks like InfoVQA, HallusionBench, and OCRBench. We attribute these gains to increased computational load. However, for OCRBench, a too-high min_pixels value leads to a severe performance decline. This is likely because OCRBench contains numerous extremely small images, and excessive enlargement causes these images to deviate from the training data distribution, turning them into out-of-distribution samples. In contrast, the effect of increasing min_pixels on the MMMU benchmark is negligible. We hypothesize that the performance bottleneck in MMMU is more related to the model's reasoning capability rather than image resolution. M-RoPE In this subsection, we demonstrate the effectiveness of M-RoPE. First, we validate its capability on various downstream tasks. We employ Qwen2-1.5B and ViT-L as the backbone and report the results of the pre-trained models. Compared to 1D-RoPE, using M-RoPE achieves better performance in downstream tasks, particularly in video benchmarks. Furthermore, we assess the length extrapolation capability of M-RoPE on Video-MME medium-length videos. The performance of Qwen2-VL-72B at different inference lengths is illustrated. Leveraging M-RoPE, the model demonstrates robust results across various inference lengths. Notably, despite limiting the maximum tokens per video to 16K during training, the model still exhibits exceptional performance at a maximum inference length of 80K tokens. Model Scaling We evaluate the performance of models of varying scales across multiple capability dimensions. Specifically, we categorize these dimensions into complex college-level problem-solving, mathematical abilities, document and table comprehension, general scenario question-answering, and video comprehension. The overall capability of a model is assessed by averaging its scores across different benchmarks associated with each dimension. In particular, we use the MMMU benchmark to represent college-level problem-solving ability, while the average scores from MathVista and MathVision serve as indicators of mathematical ability. For general scenario question-answering, we compute the average score across the RealWorldQA, MMBench-V1.1, MMT-Bench, HallBench, MMVet, and MMStar benchmarks. Document and table comprehension capability is reflected through the average score from benchmarks like DocVQA, InfoVQA, ChartQA, TextVQA, OCRBench, and MTVQA. Lastly, video comprehension ability is measured by averaging scores across MVBench, PerceptionTest, EgoSchema, and Video-MME. There is a consistent improvement in performance with increasing model size, particularly with respect to mathematical abilities, which show a positive correlation with the number of model parameters. On the other hand, for optical character recognition (OCR)-related tasks, even smaller-scale models exhibit relatively strong performance. We visualize the relationship between model performance and the number of training tokens during the second stage of pretraining for Qwen2-VL-7B.	In the MMT-Bench evaluation, which assesses advanced reasoning and instruction following across thirty two core meta-tasks and one hundred sixty two subtasks in multimodal understanding, Qwen two VL seventy two B achieves seventy one point seven, markedly surpassing the previous best (sixty three point four) and demonstrating its prowess in applying expert knowledge and executing deliberate visual recognition, localization, reasoning, and planning. On MMBench, which evaluates fine-grained abilities across twenty dimensions, Qwen two VL seventy two B exhibits strong performance, achieving eighty six point five on the English test set, matching the state-of-the-art, and eighty six point six on the Chinese test set, establishing a new benchmark. For MME, which measures a wide spectrum of perception and cognition abilities across fourteen subtasks, Qwen two VL seventy two B achieves a cumulative score of two thousand four hundred eighty two point seven, significantly outperforming the previous best (two thousand four hundred fourteen point seven), underscoring its advanced capabilities in both visual perception and high-level cognition tasks. These comprehensive results underscore the Qwen two VL series' exceptional proficiency in general visual question answering tasks. The models demonstrate advanced capabilities in real-world spatial comprehension, genuine multimodal integration, complex reasoning, instruction following, and a broad range of perception and cognition tasks. The consistent superior performance across diverse benchmarks, particularly the outstanding results of the seventy two B model, positions the Qwen two VL series as a leading solution in the field of visual question answering. Our models excel in handling visually indispensable tasks, integrating core vision-language capabilities, and demonstrating expertise across diverse multimodal scenarios, ranging from fundamental perception tasks to complex reasoning and planning. This exhaustive evaluation highlights the Qwen two VL series' versatility and effectiveness in addressing the multifaceted challenges posed by state-of-the-art multimodal benchmarks, thereby setting a new standard for large vision-language models. Document and Diagrams Reading We tested our model's OCR and document and diagram comprehension on DocVQA, ChartQA, InfoVQA, TextVQA, AI two D datasets. The DocVQA/InfoVQA/ChartQA dataset focuses on the model's ability to comprehend text in documents/high-resolution infographics/charts, while the TextVQA dataset examines the ability to comprehend text in naturalistic images. The OCRBench dataset is a a dataset of mixed tasks, which focuses on mathematical formula parsing and information extraction in addition to the text-based VQA. The AI two D dataset focuses on multiple-choice questions on scientific diagrams containing text. In addition, we also tested the OCR and formula recognition capabilities of our model on OCRBench, as well as the multilingual OCR capabilities of our model on the MTVQA dataset. The experimental results show that our model achieves SoTA level in several metrics, including DocVQA, InfoVQA, TextVQA and OCRBench, demonstrating that our model has good comprehension of textual content in images from multiple domains. Multilingual Text Recognition and Understanding In particular, our model surpasses all existing general-purpose LVLMs in multilingual OCR. Our model not only outperforms existing LVLMs (including proprietary models such as GPT four o, Claude three point five Sonnet, etc.) on the public-available MTVQA dataset, it also outperforms GPT four o on the in-house internal benchmark across all foreign languages except Arabic. Mathematical Reasoning We've conducted experiments on the MathVista and MathVision datasets to assess mathematical reasoning capabilities. MathVista is a comprehensive benchmark featuring six thousand one hundred forty-one diverse examples of mathematical and visual tasks. The MathVision dataset comprises three thousand forty math problems embedded in visual contexts from actual math competitions, covering sixteen mathematical disciplines and varying in difficulty across five levels. These challenges underscore the necessity for LVLMs to exhibit strong visual comprehension, a deep understanding of mathematics, and sound logical reasoning skills. The Qwen two VL series has demonstrated superior performance on MathVista, achieving a seventy point five outperforming other LVLMs. Additionally, it has set a new open-source benchmark on MathVision with twenty-five point nine. Referring Expression Comprehension Regarding visual localization task, we evaluate Qwen two-VL on RefCOCO, RefCOCO+, and RefCOCOg datasets. The results demonstrate that Qwen two-VL attains top-tier results among generalist models. Benefiting from a more rational structure design, Qwen two-VL is able to perceive details in high-resolution images, leading to significant improvements over Qwen-VL. The superiority of these models in comparison to both generalist and specialized models highlights their potential for advancing the field of visual localization and their capacity for real-world implementation in tasks requiring precise visual understanding. Video Understanding We evaluate our models on various video understanding tasks, with related benchmarks covering short videos of a few seconds to long videos of up to one hour. Qwen two-VL demonstrates strong results across two B, seven B, and seventy two B sizes, with Qwen two-VL-seventy two B achieving the best performance on MVBench, PerceptionTest, and EgoSchema. This showcases Qwen two-VL's superior capabilities in video understanding tasks, and scaling up Qwen two-VL yields significant improvements. For the challenging Video-MME benchmark, which includes videos up to one hour, it is noteworthy that we limited the maximum number of frames extracted per video to seven hundred sixty eight during evaluation, potentially impacting performance on longer videos. Future work will focus on extending Qwen two-VL to support longer sequences, thereby accommodating longer videos. Visual Agent Qwen two VL is evaluated first for its ability to interact with the environment via function calls and then for its capacity to complete complex sequential decision tasks through multiple rounds of interaction. The implementation is based on the Qwen-Agent framework. Function Calling Unlike function calling in LLMs, function calling in LVLMs often involves extracting information from visual cues. Due to the absence of public benchmarks for evaluating the capabilities of LVLMs in function calling, we constructed our internal evaluation dataset. To construct the evaluation dataset, we undertook the following procedures: Scene Categorization, Image Collection, Image Content Extraction, and Question/Functions/Arguments Generation. Firstly, we classified scenes into categories based on different visual applications. Subsequently, we downloaded and meticulously selected high-quality, representative images from the internet for each category. Thereafter, utilizing an advanced LVLM, we analyzed each image to extract key visual elements and textual information. Finally, based on the content information from the images, we used an advanced LLM to generate a series of questions that required specific functions to answer, along with specifying the input parameters needed for these function calls. Similar to the function calling evaluation method in LLMs, we designed two metrics to evaluate the accuracy of the function selection and the correctness of the arguments input. Specifically, Type Match(TM), is calculated as the ratio of times the model successfully invoked the correct function to the total number of calls attempted. Exact Match(EM), for each function calling, we checked whether the arguments passed to the function exactly matched those recorded in the image's content information, calculating this correctness ratio. The performance of Qwen two VL in both Type Match(ninety three point one vs. ninety point two) and Exact Match(fifty three point two vs. fifty point zero) over GPT four o substantiates the efficacy of Qwen two VL's capability in function calling, thereby underscoring its significant potential for application expansion through external tool integration. The evaluation results demonstrated that GPT four o underperformed, primarily due to two factors: in scenarios where uncertainty arises, GPT four o demonstrates a conservative approach by avoiding using external tools. The Optical Character Recognition (OCR) capability of GPT four o is outperformed by Qwen two VL, particularly in the context of Chinese characters. UI Operations/Games/Robotics/Navigation To assess Qwen two VL’s ability to generally handle complex tasks, we conduct evaluations across multiple VL agent tasks, including mobile operations, robotic control, card games, and vision-language navigation. As these tasks need multiple actions to complete tasks, we keep the history (observation, action) through Qwen two VL supports a thirty two K context length, then append each new observation image after every action, enabling continuous reasoning about subsequent steps. UI Operations: we evaluate Qwen two VL using the AITZ task, which constructs a core clean test set derived from AITW. Based on common operation patterns of phone, we define actions such as tap, input and swipe for Qwen two VL to interact with on-screen icons for task completion. For example, when Qwen two VL is tasked with finding a pizza restaurant nearby by Google Maps, it should input "pizza" in the search term, swipe to select the appropriate restaurant, and tap the corresponding link. Following the AITZ setting, we report both type match (correctness of tap, input, or swipe) and exact match (correctness of tap location, input text, or swipe direction). With the support of grounding capability on UI, Qwen two VL surpasses GPT four and previous SoTA. Robotic Control: we evaluate Qwen two VL on the ALFRED task in AI two THOR. The task requires agent to perform complex household tasks, such as toasting bread and slicing an apple to prepare a meal. To work in the virtual environment, we define high-level actions (GotoLocation, Pickup, PutDown, Open, Close, Clean, Heat, Cool, Slice) as the action set. Moreover, agent needs to localize objects for manipulation (e.g., it can only pick up an apple if the apple is recognized). To improve the accuracy of manipulation, we integrate SAM. ALFRED task reports task success rate (SR) (e.g., preparing dinner) and sub-goal completion metrics (GC) (e.g., whether the bread is toasted or the apple is sliced). Qwen two VL slightly outperforms the previously specialized model ThinkBot on the valid-unseen set. Card Games: we leverage the card game environment from RL four VLM to assess Qwen two VL's performance in a series of card-based games: Number Line, BlackJack, EZPoint, and Point twenty four. Each game presents distinct challenges: (one) reaching a target number using plus one or minus one operations, (two) drawing or holding cards to compete against the dealer, (three) applying basic arithmetic operations to reach a total of twelve, and (four) using arithmetic operations to achieve a total of twenty four. We report the success rate of the tasks. They not only evaluate agent capabilities but also require strong OCR skills to recognize these cards and understand the progression of the game. Qwen two VL demonstrates superior performance across all tasks. Vision-Language Navigation: we evaluate Qwen two VL on the Vision-and-Language Navigation (VLN) task using the R two R and REVERIE. In VLN, the model must autonomously determine the next location based on instruction, current observations. We report the success rate (SR) of VLM in reaching the predetermined destination for this task. The performance of Qwen two VL is comparable to that of GPT four o, but both models fall significantly behind current specialized VLN models. We attribute this gap to the incomplete and unstructured map information generated by the model from multiple images. Accurately modeling maps and locations in a three D environment remains a major challenge for multimodal models. Ablation Study In this section, we present ablation studies on image dynamic resolution, M RoPE, and model scale. These experiments aim to provide insights into the impact of these key components on our model's performance. Dynamic Resolution We compare the performance between dynamic resolution and fixed resolution. For fixed resolution, we resize the images to ensure a constant number of image tokens being input to the model, rather than resizing to a specific height and width, as this would distort the original aspect ratio. For dynamic resolution, we only set min pixels equals one hundred by twenty eight by twenty eight and max pixels equals sixteen thousand three hundred eighty four by twenty eight by twenty eight, allowing the number of image tokens depend primarily on the image's native resolution. It can be observed that adjusting image sizes only results in small perturbations in performance, demonstrating the model robustness to varying image sizes. Moreover, dynamic resolution approach is more efficient. We can observe that no single fixed resolution achieves optimal performance across all benchmarks. In contrast, the dynamic resolution approach consistently achieves top-tier performance while consuming fewer tokens on average. Additionally, we observe that merely increasing the image size does not always lead to improved performance. It is more important to choose an appropriate resolution for different images. We upscale small images to surpass a specified min pixels threshold. Evaluations on upscaled images shows enhanced performance on perceptual tasks like InfoVQA, HallusionBench, and OCRBench. We attribute these gains to increased computational load. However, for OCRBench, a too-high min pixels value leads to a severe performance decline. This is likely because OCRBench contains numerous extremely small images, and excessive enlargement causes these images to deviate from the training data distribution, turning them into out-of-distribution samples. In contrast, the effect of increasing min pixels on the MMMU benchmark is negligible. We hypothesize that the performance bottleneck in MMMU is more related to the model's reasoning capability rather than image resolution. M RoPE In this subsection, we demonstrate the effectiveness of M RoPE. First, we validate its capability on various downstream tasks. We employ Qwen two one point five B and ViT L as the backbone and report the results of the pre-trained models. Compared to one D RoPE, using M RoPE achieves better performance in downstream tasks, particularly in video benchmarks. Furthermore, we assess the length extrapolation capability of M RoPE on Video MME medium-length videos. The performance of Qwen two VL seventy two B at different inference lengths is illustrated. Leveraging M RoPE, the model demonstrates robust results across various inference lengths. Notably, despite limiting the maximum tokens per video to sixteen K during training, the model still exhibits exceptional performance at a maximum inference length of eighty K tokens. Model Scaling We evaluate the performance of models of varying scales across multiple capability dimensions. Specifically, we categorize these dimensions into complex college-level problem-solving, mathematical abilities, document and table comprehension, general scenario question-answering, and video comprehension. The overall capability of a model is assessed by averaging its scores across different benchmarks associated with each dimension. In particular, we use the MMMU benchmark to represent college-level problem-solving ability, while the average scores from MathVista and MathVision serve as indicators of mathematical ability. For general scenario question-answering, we compute the average score across the RealWorldQA, MMBench-V1.1, MMT-Bench, HallBench, MMVet, and MMStar benchmarks. Document and table comprehension capability is reflected through the average score from benchmarks like DocVQA, InfoVQA, ChartQA, TextVQA, OCRBench, and MTVQA. Lastly, video comprehension ability is measured by averaging scores across MVBench, PerceptionTest, EgoSchema, and Video-MME. There is a consistent improvement in performance with increasing model size, particularly with respect to mathematical abilities, which show a positive correlation with the number of model parameters. On the other hand, for optical character recognition (OCR)-related tasks, even smaller-scale models exhibit relatively strong performance. We visualize the relationship between model performance and the number of training tokens during the second stage of pretraining for Qwen two-VL-seven B.
long_en_312	wiki_en	1,181	en	Salvation (2017–2018) Salvage 1 (1979) Sanctuary (2008–2012, Canada) Sapphire & Steel (1979–1982, UK) Science Fiction Theatre (1955–1957, anthology) seaQuest DSV (1993–1996; renamed seaQuest 2032 for third season) The Second Hundred Years (1967–1968) The Secret Adventures of Jules Verne (2000) Secret Agent Man (2000) The Secret Service (1969, UK, puppetry) The Secret World of Alex Mack (1994–1998) The Secrets of Isis (1975–1977) Sense8 (2015–2018) The Sentinel (1997–1999) Seven Days (1998–2001) Shadows (1995–present) Sharad of Atlantis (1966, film) a.k.a. Undersea Kingdom (1936, film serial) Shazam! (1974–1977) Silversun (2004, Australia) Sir Arthur Conan Doyle's The Lost World (1999–2002, Canada/Australia/New Zealand) The Six Million Dollar Man (1974–1978) Sleepwalkers (1997–1998) Sliders (1995–2000) Small Wonder (1985–1989) So Weird (1999–2001) Something Is Out There (1988, miniseries) Space: 1999 (1975–1977, UK) Space: Above and Beyond (1995–1996) Space Academy (1977–1979) Space Cases (1996–1997) Space Command (1953–1954, Canada) Space Island One (1998, UK/Germany) Space Odyssey: Voyage To The Planets (2004, UK, docufiction) Space Patrol (1950–1955) Space Patrol (1963–1964, UK) a.k.a. Planet Patrol (US) Space Police (1986, Space Precinct pilot) Space Precinct (1994–1995) Space Rangers (1993) Spadla z oblakov She Came Out of the Blue Sky (1978–1979, Czechoslovakia) The Sparticle Mystery (2011–2013) Special Unit 2 (2001–2002) Spectre (1977, UK, film) Speed Racer (US adaptation) a.k.a. Mach Go Go Go (Japan) (franchise) Spellbinder (1995, Australia/Poland) Spider-Man (franchise): - Spider-Man (1978–1979, Japan) - The Amazing Spider-Man (1977–1979) - Spidey Super Stories (1974–1977) The Stand (1994, miniseries) Star Command (1996, film) Star Cops (1987, UK) Star-Crossed (2014) Star Maidens (1976, UK) Star Trek (franchise): - Star Trek: The Original Series a.k.a. ST:TOS (1966–1969) - Star Trek: The Next Generation a.k.a. ST:TNG (1987–1994) - Star Trek: Deep Space Nine a.k.a. ST:DS9 (1993–1999) - Star Trek: Voyager a.k.a. ST:VOY (1995–2001) - Star Trek: Enterprise a.k.a. ST:ENT (2001–2005) - Star Trek: Discovery a.k.a. ST:DSC (2017–present) - Star Trek: Short Treks a.k.a. ST:ST (2018–present) - Star Trek: Picard a.k.a. ST:PIC (2020–2023) - Star Trek: Strange New Worlds (2022–present) Star Wars (franchise): - The Star Wars Holiday Special (1978, special, film) - Caravan of Courage: An Ewok Adventure (1984, film) - Ewoks: The Battle for Endor (1985, film) - R2-D2: Beneath the Dome (2001, mockumentary) - Star Wars: The Legacy Revealed (2007, special, documentary) Star Wolf (1978, Japan) Starflight: The Plane That Couldn't Land a.k.a. Starflight One (1983, film) Stargate (franchise): - Stargate Atlantis a.k.a. SGA (2004–2009, Canada/US) - Stargate SG-1 SG-1 (1997–2007, Canada/US, ship-based seasons 6–10) Stargate Universe (aka SGU) (2009–2011, Canada/US) Starhunter (2000–2004, Canada) Starhyke (2011, UK) The Starlost (1973–1974, Canada) Starman (1986–1987) The Stepford Children (1987) Stingray (1964–1965, UK, puppetry) Stormworld (2009, Australia/Canada) Stowaway to the Moon (1975, film) Strange Days at Blake Holsey High (aka Black Hole High) (2002, Canada) Strange Frequency (2001) Strange Luck (1995–1996) Strange World (1999–2002) The Stranger (1964–1965, Australia) The Stranger (aka Stranded in Space) (1973, film, pilot) Stranger from Space (1951, UK) Stranger Things (2016–present) The Strangerers (2000, UK) Street Hawk (1985) Super Force (1990–1992) Himitsu Sentai Gorenger (aka Gorenger, Goranger) (1975–1977, Japan) J.A.K.Q. Dengekitai (aka The Jackers) (1977, Japan) Battle Fever J (1979–1980, Japan) Denshi Sentai Denziman (aka Denjiman, Electric Fighters, Denziman) (1980–1981, Japan) Taiyo Sentai Sun Vulcan (aka Sun Vulcan) (1981–1982, Japan) Dai Sentai Goggle-V (aka Dai Sentai Goggle Five, Goggle V) (1982–1983, Japan) Kagaku Sentai Dynaman (aka Dynaman) (1983–1984, Japan) Choudenshi Bioman (aka Bioman) (1984–1985, Japan) Dengeki Sentai Changeman (aka Changeman) (1985–1986, Japan) Choushinsei Flashman (aka Flashman) (1986–1987, Japan) Hikari Sentai Maskman (aka Maskman) (1987–1988, Japan) Bioman 2 (France) Choujuu Sentai Liveman (1988–1989, Japan) a.k.a. Liveman; Bioman 3 (France) Kousoku Sentai Turboranger (1989–1990, Japan) a.k.a. Turbo Rangers Chikyu Sentai Fiveman (1990–1991, Japan) a.k.a. Fiveman; Sky Rangers Chōjin Sentai Jetman (1991–1992, Japan) a.k.a. Jetman Kyōryū Sentai Zyuranger (1992–1993, Japan) a.k.a. Zyuranger; Galaxy Rangers Gosei Sentai Dairanger (1993–1994, Japan) a.k.a. Dairanger; Star Rangers Ninja Sentai Kakuranger (1994–1995, Japan) a.k.a. Kakuranger; Ninja Rangers Chōriki Sentai Ohranger (1995–1996, Japan) a.k.a. Ohranger Gekisou Sentai Carranger (1996–1997, Japan) a.k.a. Carranger Denji Sentai Megaranger (1997–1998, Japan) a.k.a. Megaranger Seijuu Sentai Gingaman (1998–1999, Japan) a.k.a. Gingaman Kyuukyuu Sentai GoGoFive (1999–2000, Japan) a.k.a. GoGoFive Mirai Sentai Timeranger (2000–2001, Japan) a.k.a. Timeranger Hyakujuu Sentai Gaoranger (2001–2002, Japan) a.k.a. Gaoranger Ninpuu Sentai Hurricaneger (2002–2003, Japan) a.k.a. Hurricaneger Bakuryū Sentai Abaranger (2003–2004, Japan) a.k.a. Abaranger Tokusou Sentai Dekaranger (2004–2005, Japan) a.k.a. Dekaranger Mahou Sentai Magiranger (2005–2006, Japan) a.k.a. Magiranger GoGo Sentai Boukenger (2006–2007, Japan) a.k.a. Boukenger Juken Sentai Gekiranger (2007–2008, Japan) a.k.a. Gekiranger Engine Sentai Go-onger (2008–2009, Japan) a.k.a. Go-onger Samurai Sentai Shinkenger (2009–2010, Japan) a.k.a. Shinkenger Tensou Sentai Goseiger Goseiger (2010–2011, Japan) Kaizoku Sentai Gokaiger (aka Gokaiger) (2011–2012, Japan) Tokumei Sentai Go-Busters (aka Go-Busters) (2012–2013, Japan) Unofficial Sentai Akibaranger (aka Akibaranger) (2012–2013, Japan) Zyuden Sentai Kyoryuger (aka Kyoryuger) (2013–2014, Japan) Ressha Sentai ToQger (aka ToQger) (2014–2015, Japan) Shuriken Sentai Ninninger (aka Ninninger) (2015–2016, Japan) Doubutsu Sentai Zyuohger (aka Zyuohger) (2016–2017, Japan) Uchu Sentai Kyuranger (aka Kyuranger) (2017–2018, Japan) Kaitou Sentai Lupinranger VS Keisatsu Sentai Patranger (aka Lupinranger VS Patranger) (2018–2019, Japan) Super Sentai Strongest Battle (2019, Japan) Kishiryu Sentai Ryusoulger (aka Ryusoulger) (2019–2020, Japan) Mashin Sentai Kiramager (aka Kiramager) (2020–2021, Japan) Supercar (1961–1962, puppetry) Supergirl (2015–2021) Superhuman Samurai Syber-Squad (1994–1995, US; adaptation of Gridman the Hyper Agent) Superman (franchise) The Adventures of Superman (1952–1958) The Adventures of Superboy (1961, pilot) Superboy Adventures of Superboy, The (1988–1992) Lois & Clark: The New Adventures of Superman (1993–1997) Smallville (2001–2011) Superman & Lois (2021–present) Surface (2005–2006) Survivors (franchise): Survivors (2008–2010, UK) Survivors (1975–1977, UK) Swamp Thing (franchise): Swamp Thing (1990–1993) Swamp Thing (2019) Animation: Saber Marionette (franchise): Saber Marionette J (1996–1997, Japan, animated) Saber Marionette J to X (1998–1999, Japan, animated) Saber Rider and the Star Sheriffs (1987–1988, Japan, animated) Samurai 7 a.k.a. Samurai Sebun (2004, Japan, animated) Samurai Jack (2001–2004, 2017, animated) Savage Dragon (1995–1996, animated) Sealab 2020 (1972, animated) Sealab 2021 (2000–2005, animated) Secret of Cerulean Sand (2002, Japan, animated) Secret Saturdays, The (2008–2010, animated) Sectaurs (1985, animated) Serial Experiments Lain (1998, Japan, animated)	Salvation (two thousand seventeen–two thousand eighteen) Salvage one (nineteen seventy nine) Sanctuary (two thousand eight–two thousand twelve, Canada) Sapphire and Steel (nineteen seventy nine–nineteen eighty two, UK) Science Fiction Theatre (nineteen fifty five–nineteen fifty seven, anthology) seaQuest DSV (nineteen ninety three–nineteen ninety six; renamed seaQuest two thousand thirty two for third season) The Second Hundred Years (nineteen sixty seven–nineteen sixty eight) The Secret Adventures of Jules Verne (two thousand) Secret Agent Man (two thousand) The Secret Service (nineteen sixty nine, UK, puppetry) The Secret World of Alex Mack (nineteen ninety four–nineteen ninety eight) The Secrets of Isis (nineteen seventy five–nineteen seventy seven) Sense eight (two thousand fifteen–two thousand eighteen) The Sentinel (nineteen ninety seven–nineteen ninety nine) Seven Days (nineteen ninety eight–two thousand one) Shadows (nineteen ninety five–present) Sharad of Atlantis (nineteen sixty six, film) a.k.a. Undersea Kingdom (nineteen thirty six, film serial) Shazam! (nineteen seventy four–nineteen seventy seven) Silversun (two thousand four, Australia) Sir Arthur Conan Doyle's The Lost World (nineteen ninety nine–two thousand two, Canada/Australia/New Zealand) The Six Million Dollar Man (nineteen seventy four–nineteen seventy eight) Sleepwalkers (nineteen ninety seven–nineteen ninety eight) Sliders (nineteen ninety five–two thousand) Small Wonder (nineteen eighty five–nineteen eighty nine) So Weird (nineteen ninety nine–two thousand one) Something Is Out There (nineteen eighty eight, miniseries) Space: nineteen ninety nine (nineteen seventy five–nineteen seventy seven, UK) Space: Above and Beyond (nineteen ninety five–nineteen ninety six) Space Academy (nineteen seventy seven–nineteen seventy nine) Space Cases (nineteen ninety six–nineteen ninety seven) Space Command (nineteen fifty three–nineteen fifty four, Canada) Space Island One (nineteen ninety eight, UK/Germany) Space Odyssey: Voyage To The Planets (two thousand four, UK, docufiction) Space Patrol (nineteen fifty–nineteen fifty five) Space Patrol (nineteen sixty three–nineteen sixty four, UK) a.k.a. Planet Patrol (US) Space Police (nineteen eighty six, Space Precinct pilot) Space Precinct (nineteen ninety four–nineteen ninety five) Space Rangers (nineteen ninety three) Spadla z oblakov She Came Out of the Blue Sky (nineteen seventy eight–nineteen seventy nine, Czechoslovakia) The Sparticle Mystery (two thousand eleven–two thousand thirteen) Special Unit two (two thousand one–two thousand two) Spectre (nineteen seventy seven, UK, film) Speed Racer (US adaptation) a.k.a. Mach Go Go Go (Japan) (franchise) Spellbinder (nineteen ninety five, Australia/Poland) Spider-Man (franchise): - Spider-Man (nineteen seventy eight–nineteen seventy nine, Japan) - The Amazing Spider-Man (nineteen seventy seven–nineteen seventy nine) - Spidey Super Stories (nineteen seventy four–nineteen seventy seven) The Stand (nineteen ninety four, miniseries) Star Command (nineteen ninety six, film) Star Cops (nineteen eighty seven, UK) Star-Crossed (two thousand fourteen) Star Maidens (nineteen seventy six, UK) Star Trek (franchise): - Star Trek: The Original Series a.k.a. ST:TOS (nineteen sixty six–nineteen sixty nine) - Star Trek: The Next Generation a.k.a. ST:TNG (nineteen eighty seven–nineteen ninety four) - Star Trek: Deep Space Nine a.k.a. ST:DS9 (nineteen ninety three–nineteen ninety nine) - Star Trek: Voyager a.k.a. ST:VOY (nineteen ninety five–two thousand one) - Star Trek: Enterprise a.k.a. ST:ENT (two thousand one–two thousand five) - Star Trek: Discovery a.k.a. ST:DSC (two thousand seventeen–present) - Star Trek: Short Treks a.k.a. ST:ST (two thousand eighteen–present) - Star Trek: Picard a.k.a. ST:PIC (two thousand twenty–two thousand twenty three) - Star Trek: Strange New Worlds (two thousand twenty two–present) Star Wars (franchise): - The Star Wars Holiday Special (nineteen seventy eight, special, film) - Caravan of Courage: An Ewok Adventure (nineteen eighty four, film) - Ewoks: The Battle for Endor (nineteen eighty five, film) - R two D two: Beneath the Dome (two thousand one, mockumentary) - Star Wars: The Legacy Revealed (two thousand seven, special, documentary) Star Wolf (nineteen seventy eight, Japan) Starflight: The Plane That Couldn't Land a.k.a. Starflight One (one thousand nine hundred eighty three, film) Stargate (franchise): - Stargate Atlantis a.k.a. SGA (two thousand four–two thousand nine, Canada/US) - Stargate SG one SG one (one thousand nine hundred ninety seven–two thousand seven, Canada/US, ship-based seasons six–ten) Stargate Universe (aka SGU) (two thousand nine–two thousand eleven, Canada/US) Starhunter (two thousand–two thousand four, Canada) Starhyke (two thousand eleven, UK) The Starlost (one thousand nine hundred seventy three–one thousand nine hundred seventy four, Canada) Starman (one thousand nine hundred eighty six–one thousand nine hundred eighty seven) The Stepford Children (one thousand nine hundred eighty seven) Stingray (one thousand nine hundred sixty four–one thousand nine hundred sixty five, UK, puppetry) Stormworld (two thousand nine, Australia/Canada) Stowaway to the Moon (one thousand nine hundred seventy five, film) Strange Days at Blake Holsey High (aka Black Hole High) (two thousand two, Canada) Strange Frequency (two thousand one) Strange Luck (one thousand nine hundred ninety five–one thousand nine hundred ninety six) Strange World (one thousand nine hundred ninety nine–two thousand two) The Stranger (one thousand nine hundred sixty four–one thousand nine hundred sixty five, Australia) The Stranger (aka Stranded in Space) (one thousand nine hundred seventy three, film, pilot) Stranger from Space (one thousand nine hundred fifty one, UK) Stranger Things (two thousand sixteen–present) The Strangerers (two thousand, UK) Street Hawk (one thousand nine hundred eighty five) Super Force (one thousand nine hundred ninety–one thousand nine hundred ninety two) Himitsu Sentai Gorenger (aka Gorenger, Goranger) (one thousand nine hundred seventy five–one thousand nine hundred seventy seven, Japan) J.A.K.Q. Dengekitai (aka The Jackers) (nineteen seventy-seven, Japan) Battle Fever J (nineteen seventy-nine–nineteen eighty, Japan) Denshi Sentai Denziman (aka Denjiman, Electric Fighters, Denziman) (nineteen eighty–nineteen eighty-one, Japan) Taiyo Sentai Sun Vulcan (aka Sun Vulcan) (nineteen eighty-one–nineteen eighty-two, Japan) Dai Sentai Goggle-V (aka Dai Sentai Goggle Five, Goggle V) (nineteen eighty-two–nineteen eighty-three, Japan) Kagaku Sentai Dynaman (aka Dynaman) (nineteen eighty-three–nineteen eighty-four, Japan) Choudenshi Bioman (aka Bioman) (nineteen eighty-four–nineteen eighty-five, Japan) Dengeki Sentai Changeman (aka Changeman) (nineteen eighty-five–nineteen eighty-six, Japan) Choushinsei Flashman (aka Flashman) (nineteen eighty-six–nineteen eighty-seven, Japan) Hikari Sentai Maskman (aka Maskman) (nineteen eighty-seven–nineteen eighty-eight, Japan) Bioman two (France) Choujuu Sentai Liveman (nineteen eighty-eight–nineteen eighty-nine, Japan) a.k.a. Liveman; Bioman three (France) Kousoku Sentai Turboranger (nineteen eighty-nine–nineteen ninety, Japan) a.k.a. Turbo Rangers Chikyu Sentai Fiveman (nineteen ninety–nineteen ninety-one, Japan) a.k.a. Fiveman; Sky Rangers Chōjin Sentai Jetman (nineteen ninety-one–nineteen ninety-two, Japan) a.k.a. Jetman Kyōryū Sentai Zyuranger (nineteen ninety-two–nineteen ninety-three, Japan) a.k.a. Zyuranger; Galaxy Rangers Gosei Sentai Dairanger (nineteen ninety-three–nineteen ninety-four, Japan) a.k.a. Dairanger; Star Rangers Ninja Sentai Kakuranger (nineteen ninety-four–nineteen ninety-five, Japan) a.k.a. Kakuranger; Ninja Rangers Chōriki Sentai Ohranger (nineteen ninety-five–nineteen ninety-six, Japan) a.k.a. Ohranger Gekisou Sentai Carranger (nineteen ninety-six–nineteen ninety-seven, Japan) a.k.a. Carranger Denji Sentai Megaranger (nineteen ninety-seven–nineteen ninety-eight, Japan) a.k.a. Megaranger Seijuu Sentai Gingaman (nineteen ninety-eight–nineteen ninety-nine, Japan) a.k.a. Gingaman Kyuukyuu Sentai GoGoFive (nineteen ninety-nine–two thousand, Japan) a.k.a. GoGoFive Mirai Sentai Timeranger (two thousand–two thousand one, Japan) a.k.a. Timeranger Hyakujuu Sentai Gaoranger (two thousand one–two thousand two, Japan) a.k.a. Gaoranger Ninpuu Sentai Hurricaneger (two thousand two to two thousand three, Japan) a.k.a. Hurricaneger Bakuryū Sentai Abaranger (two thousand three to two thousand four, Japan) a.k.a. Abaranger Tokusou Sentai Dekaranger (two thousand four to two thousand five, Japan) a.k.a. Dekaranger Mahou Sentai Magiranger (two thousand five to two thousand six, Japan) a.k.a. Magiranger GoGo Sentai Boukenger (two thousand six to two thousand seven, Japan) a.k.a. Boukenger Juken Sentai Gekiranger (two thousand seven to two thousand eight, Japan) a.k.a. Gekiranger Engine Sentai Go-onger (two thousand eight to two thousand nine, Japan) a.k.a. Go-onger Samurai Sentai Shinkenger (two thousand nine to two thousand ten, Japan) a.k.a. Shinkenger Tensou Sentai Goseiger Goseiger (two thousand ten–two thousand eleven, Japan) Kaizoku Sentai Gokaiger (aka Gokaiger) (two thousand eleven–two thousand twelve, Japan) Tokumei Sentai Go-Busters (aka Go-Busters) (two thousand twelve–two thousand thirteen, Japan) Unofficial Sentai Akibaranger (aka Akibaranger) (two thousand twelve–two thousand thirteen, Japan) Zyuden Sentai Kyoryuger (aka Kyoryuger) (two thousand thirteen–two thousand fourteen, Japan) Ressha Sentai ToQger (aka ToQger) (two thousand fourteen–two thousand fifteen, Japan) Shuriken Sentai Ninninger (aka Ninninger) (two thousand fifteen–two thousand sixteen, Japan) Doubutsu Sentai Zyuohger (aka Zyuohger) (two thousand sixteen–two thousand seventeen, Japan) Uchu Sentai Kyuranger (aka Kyuranger) (two thousand seventeen–two thousand eighteen, Japan) Kaitou Sentai Lupinranger VS Keisatsu Sentai Patranger (aka Lupinranger VS Patranger) (two thousand eighteen–two thousand nineteen, Japan) Super Sentai Strongest Battle (two thousand nineteen, Japan) Kishiryu Sentai Ryusoulger (aka Ryusoulger) (two thousand nineteen–two thousand twenty, Japan) Mashin Sentai Kiramager (aka Kiramager) (two thousand twenty–two thousand twenty-one, Japan) Supercar (nineteen sixty-one–nineteen sixty-two, puppetry) Supergirl (two thousand fifteen–two thousand twenty-one) Superhuman Samurai Syber-Squad (nineteen ninety-four–nineteen ninety-five, US; adaptation of Gridman the Hyper Agent) Superman (franchise) The Adventures of Superman (nineteen fifty-two–nineteen fifty-eight) The Adventures of Superboy (nineteen sixty-one, pilot) Superboy Adventures of Superboy, The (nineteen eighty-eight–nineteen ninety-two) Lois and Clark: The New Adventures of Superman (nineteen ninety-three–nineteen ninety-seven) Smallville (two thousand one–two thousand eleven) Superman and Lois (two thousand twenty-one–present) Surface (two thousand five–two thousand six) Survivors (franchise): Survivors (two thousand eight–two thousand ten, UK) Survivors (nineteen seventy-five–nineteen seventy-seven, UK) Swamp Thing (franchise): Swamp Thing (nineteen ninety–nineteen ninety-three) Swamp Thing (two thousand nineteen) Animation: Saber Marionette (franchise): Saber Marionette J (nineteen ninety-six–nineteen ninety-seven, Japan, animated) Saber Marionette J to X (nineteen ninety-eight–nineteen ninety-nine, Japan, animated) Saber Rider and the Star Sheriffs (nineteen eighty-seven–nineteen eighty-eight, Japan, animated) Samurai seven a.k.a. Samurai Sebun (two thousand four, Japan, animated) Samurai Jack (two thousand one–two thousand four, two thousand seventeen, animated) Savage Dragon (one thousand nine hundred ninety five–one thousand nine hundred ninety six, animated) Sealab two thousand twenty (one thousand nine hundred seventy two, animated) Sealab two thousand twenty one (two thousand–two thousand five, animated) Secret of Cerulean Sand (two thousand two, Japan, animated) Secret Saturdays, The (two thousand eight–two thousand ten, animated) Sectaurs (one thousand nine hundred eighty five, animated) Serial Experiments Lain (one thousand nine hundred ninety eight, Japan, animated).

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.