ResearchArcade
Collection
23 items
•
Updated
Dataset Viewer
id
int64 1
42.8k
| arxiv_id
stringlengths 10
13
| base_arxiv_id
float64 1.11k
2.51k
| version
float64 1
14
⌀ | title
stringlengths 8
224
| abstract
stringlengths 51
4.03k
| submit_date
stringdate 2011-09-01 00:00:00
2025-08-13 00:00:00
| metadata
stringlengths 283
56.2k
|
|---|---|---|---|---|---|---|---|
1
|
2008.03299
| 2,008.03299
| null |
Topology in cyber research
|
We give an idiosyncratic overview of applications of topology to cyber
research, spanning the analysis of variables/assignments and control flow in
computer programs, a brief sketch of topological data analysis in one
dimension, and the use of sheaves to analyze wireless networks.
The text is from a chapter in the forthcoming book Mathematics in Cyber
Research, to be published by Taylor and Francis.
|
2020-08-10
|
{'id': '2008.03299', 'title': 'Topology in cyber research', 'abstract': 'We give an idiosyncratic overview of applications of topology to cyber\nresearch, spanning the analysis of variables/assignments and control flow in\ncomputer programs, a brief sketch of topological data analysis in one\ndimension, and the use of sheaves to analyze wireless networks.\n The text is from a chapter in the forthcoming book Mathematics in Cyber\nResearch, to be published by Taylor and Francis.', 'authors': ['Steve Huntsman', 'Jimmy Palladino', 'Michael Robinson'], 'published': '2020-08-10 02:38:47+00:00', 'categories': ['math.AT', 'cs.NI', 'math.CO'], 'url': 'http://arxiv.org/abs/2008.03299v1'}
|
2
|
2008.03300
| 2,008.033
| null |
Electrodynamics in flat spacetime of six dimensions
|
We consider the dynamics of a classical charge in flat spacetime of six
dimensions. The mass shell relation of a free charge admits nonlinear
oscillations. Having analyzed the problem of on eigenvalues and eigenvectors of
Faraday tensor, we establish the algebraic structure of electromagnetic field
in 6D. We elaborate the classification scheme based on three field's
invariants. Using the basic algebraic properties of the electromagnetic field
tensor we analyze the motion of a charge in constant electromagnetic field. Its
world line is a combination of hyperbolic and circular orbits which lie in
three mutually orthogonal sheets of two dimensions. Within the braneworld
scenario, we project the theory on the de Sitter space of four dimensions.
Actually, as it turns out, spins of elementary particles themselves are
manifestations of extra dimensions.
|
2020-08-10
|
{'id': '2008.03300', 'title': 'Electrodynamics in flat spacetime of six dimensions', 'abstract': "We consider the dynamics of a classical charge in flat spacetime of six\ndimensions. The mass shell relation of a free charge admits nonlinear\noscillations. Having analyzed the problem of on eigenvalues and eigenvectors of\nFaraday tensor, we establish the algebraic structure of electromagnetic field\nin 6D. We elaborate the classification scheme based on three field's\ninvariants. Using the basic algebraic properties of the electromagnetic field\ntensor we analyze the motion of a charge in constant electromagnetic field. Its\nworld line is a combination of hyperbolic and circular orbits which lie in\nthree mutually orthogonal sheets of two dimensions. Within the braneworld\nscenario, we project the theory on the de Sitter space of four dimensions.\nActually, as it turns out, spins of elementary particles themselves are\nmanifestations of extra dimensions.", 'authors': ['Yurij Yaremko'], 'published': '2020-08-10 12:33:45+00:00', 'categories': ['physics.class-ph', 'hep-th'], 'url': 'http://arxiv.org/abs/2008.03300v1'}
|
3
|
2008.03302
| 2,008.03302
| null |
The operational reality of quantum nonlocality
|
Does the remote measurement-disturbance of the quantum state of a system $B$
by measurement on system $A$ entangled with $B$, constitute a real disturbance
-- i.e., an objective alteration -- of $B$ in an operational sense? Employing
information theoretic criteria motivated by operational considerations alone,
we argue that the disturbance in question is real for a subset of steerable
correlations. This result highlights the distinction between quantum
no-signaling and relativistic signal-locality. It furthermore suggests a
natural reason why a convex operational theory should be non-signaling: namely,
to ensure the consistency between the properties of reduced systems and those
of single systems.
|
2020-08-10
|
{'id': '2008.03302', 'title': 'The operational reality of quantum nonlocality', 'abstract': 'Does the remote measurement-disturbance of the quantum state of a system $B$\nby measurement on system $A$ entangled with $B$, constitute a real disturbance\n-- i.e., an objective alteration -- of $B$ in an operational sense? Employing\ninformation theoretic criteria motivated by operational considerations alone,\nwe argue that the disturbance in question is real for a subset of steerable\ncorrelations. This result highlights the distinction between quantum\nno-signaling and relativistic signal-locality. It furthermore suggests a\nnatural reason why a convex operational theory should be non-signaling: namely,\nto ensure the consistency between the properties of reduced systems and those\nof single systems.', 'authors': ['R. Srikanth'], 'published': '2020-08-10 16:07:45+00:00', 'categories': ['quant-ph'], 'url': 'http://arxiv.org/abs/2008.03302v3'}
|
4
|
2008.03842
| 2,008.03842
| null |
Modeling the Galactic Compact Binary Neutron Star Population and Studying the Double Pulsar System
|
In this dissertation, we estimate the population of different classes of BNS
systems that are visible to gravitational-wave observatories. Given that no
ultra-compact BNS systems have been discovered in pulsar radio surveys, we
place a 95\% confidence upper limit of $$850 and $$1100 ultra-compact
neutron star--white dwarf and double neutron star (DNS) systems that are
beaming towards the Earth, respectively. We show that among all of the current
radio pulsar surveys, the ones at the Arecibo radio telescope have the best
chance of detecting an ultra-compact BNS system. We also show that adopting a
survey integration time of $t_int 1$~min will maximize the
signal-to-noise ratio, and thus, the probability of detecting an ultra-compact
BNS system. Similarly, we use the sample of nine observed DNS systems to derive
a Galactic DNS merger rate of $R_MW =
37^+24_-11$~Myr$^-1$, where the errors represent 90\% confidence
intervals. Extrapolating this rate to the observable volume for LIGO, we derive
a merger detection rate of $R = 1.9^+1.2_-0.6 (D_r/100 \ Mpc )^3 yr^-1$, where $D_r$ is the range
distance for LIGO. This rate is consistent with that derived using the DNS
mergers observed by LIGO. Finally, we measure the sense of rotation of the
older millisecond pulsar, pulsar A, in the DNS J0737--3039 system and find that
it rotates prograde with respect to its orbit. This is the first direct
measurement of the sense of rotation of a pulsar and a direct confirmation of
the rotating lighthouse model for pulsars. This result confirms that the spin
angular momentum vector is closely aligned with the orbital angular momentum,
suggesting that kick of the supernova producing the second born pulsar
J0737--3039B was small.
|
2020-08-10
|
{'id': '2008.03842', 'title': 'Modeling the Galactic Compact Binary Neutron Star Population and Studying the Double Pulsar System', 'abstract': 'In this dissertation, we estimate the population of different classes of BNS\nsystems that are visible to gravitational-wave observatories. Given that no\nultra-compact BNS systems have been discovered in pulsar radio surveys, we\nplace a 95\\% confidence upper limit of $\\sim$850 and $\\sim$1100 ultra-compact\nneutron star--white dwarf and double neutron star (DNS) systems that are\nbeaming towards the Earth, respectively. We show that among all of the current\nradio pulsar surveys, the ones at the Arecibo radio telescope have the best\nchance of detecting an ultra-compact BNS system. We also show that adopting a\nsurvey integration time of $t_{\\rm int} \\sim 1$~min will maximize the\nsignal-to-noise ratio, and thus, the probability of detecting an ultra-compact\nBNS system. Similarly, we use the sample of nine observed DNS systems to derive\na Galactic DNS merger rate of $\\mathcal{R}_{\\rm MW} =\n37^{+24}_{-11}$~Myr$^{-1}$, where the errors represent 90\\% confidence\nintervals. Extrapolating this rate to the observable volume for LIGO, we derive\na merger detection rate of $\\mathcal{R} = 1.9^{+1.2}_{-0.6} \\times \\left(D_{\\rm\nr}/100 \\ \\rm Mpc \\right)^3 \\rm yr^{-1}$, where $D_{\\rm r}$ is the range\ndistance for LIGO. This rate is consistent with that derived using the DNS\nmergers observed by LIGO. Finally, we measure the sense of rotation of the\nolder millisecond pulsar, pulsar A, in the DNS J0737--3039 system and find that\nit rotates prograde with respect to its orbit. This is the first direct\nmeasurement of the sense of rotation of a pulsar and a direct confirmation of\nthe rotating lighthouse model for pulsars. This result confirms that the spin\nangular momentum vector is closely aligned with the orbital angular momentum,\nsuggesting that kick of the supernova producing the second born pulsar\nJ0737--3039B was small.', 'authors': ['Nihan Pol'], 'published': '2020-08-10 00:16:14+00:00', 'categories': ['astro-ph.HE'], 'url': 'http://arxiv.org/abs/2008.03842v2'}
|
5
|
2008.03845
| 2,008.03845
| null |
On the Gap between Epidemiological Surveillance and Preparedness
|
Contemporary Epidemiological Surveillance (ES) relies heavily on data
analytics. These analytics are critical input for pandemics preparedness
networks; however, this input is not integrated into a form suitable for
decision makers or experts in preparedness. A decision support system (DSS)
with Computational Intelligence (CI) tools is required to bridge the gap
between epidemiological model of evidence and expert group decision. We argue
that such DSS shall be a cognitive dynamic system enabling the CI and human
expert to work together. The core of such DSS must be based on machine
reasoning techniques such as probabilistic inference, and shall be capable of
estimating risks, reliability and biases in decision making.
|
2020-08-10
|
{'id': '2008.03845', 'title': 'On the Gap between Epidemiological Surveillance and Preparedness', 'abstract': 'Contemporary Epidemiological Surveillance (ES) relies heavily on data\nanalytics. These analytics are critical input for pandemics preparedness\nnetworks; however, this input is not integrated into a form suitable for\ndecision makers or experts in preparedness. A decision support system (DSS)\nwith Computational Intelligence (CI) tools is required to bridge the gap\nbetween epidemiological model of evidence and expert group decision. We argue\nthat such DSS shall be a cognitive dynamic system enabling the CI and human\nexpert to work together. The core of such DSS must be based on machine\nreasoning techniques such as probabilistic inference, and shall be capable of\nestimating risks, reliability and biases in decision making.', 'authors': ['Svetlana Yanushkevich', 'Vlad Shmerko'], 'published': '2020-08-10 00:43:52+00:00', 'categories': ['cs.AI', 'cs.SI'], 'url': 'http://arxiv.org/abs/2008.03845v1'}
|
3,681
|
2501.17762
| 2,501.17762
| null |
Improving Privacy Benefits of Redaction
|
We propose a novel redaction methodology that can be used to sanitize natural
text data. Our new technique provides better privacy benefits than other state
of the art techniques while maintaining lower redaction levels.
|
2025-01-29
|
{'id': '2501.17762', 'title': 'Improving Privacy Benefits of Redaction', 'abstract': 'We propose a novel redaction methodology that can be used to sanitize natural\ntext data. Our new technique provides better privacy benefits than other state\nof the art techniques while maintaining lower redaction levels.', 'authors': ['Vaibhav Gusain', 'Douglas Leith'], 'published': '2025-01-29 16:53:16+00:00', 'categories': ['cs.CR', 'cs.CL', 'cs.LG'], 'url': 'http://arxiv.org/abs/2501.17762v3'}
|
6
|
2008.03846
| 2,008.03846
| null |
Multistability of Small Reaction Networks
|
For three typical sets of small reaction networks (networks with two
reactions, one irreversible and one reversible reaction, or two
reversible-reaction pairs), we completely answer the challenging question: what
is the smallest subset of all multistable networks such that any multistable
network outside of the subset contains either more species or more reactants
than any network in this subset?
|
2020-08-10
|
{'id': '2008.03846', 'title': 'Multistability of Small Reaction Networks', 'abstract': 'For three typical sets of small reaction networks (networks with two\nreactions, one irreversible and one reversible reaction, or two\nreversible-reaction pairs), we completely answer the challenging question: what\nis the smallest subset of all multistable networks such that any multistable\nnetwork outside of the subset contains either more species or more reactants\nthan any network in this subset?', 'authors': ['Xiaoxian Tang', 'Hao Xu'], 'published': '2020-08-10 00:50:22+00:00', 'categories': ['q-bio.MN', 'math.AG', 'math.DS'], 'url': 'http://arxiv.org/abs/2008.03846v2'}
|
7
|
2008.03847
| 2,008.03847
| null |
Color, Flavor, Temperature and Magnetic Field Dependence of QCD Phase Diagram: Magnetic Catalysis and its Inverse
|
We study dynamical chiral symmetry breaking for quarks in the fundamental
representation of $SU(N_c)$ for $N_f$ number of light quark flavors. We also
investigate the phase diagram of quantum chromodynamics at finite temperature
$T$ and/or in the presence of a constant external magnetic field $eB$. The
unified formalism for this analysis is provided by a symmetry-preserving
Schwinger-Dyson equations treatment of a vector$$vector contact
interaction model which encodes several well-established features of quantum
chromodynamics to mimic the latter as closely as possible. Deconfinement and
chiral symmetry restoration are triggered above a critical value of $N_f$ at
$T=0=eB$. On the other hand, increasing temperature itself screens strong
interactions, thus ensuring that a smaller value of $N_f$ is sufficient to
restore chiral symmetry at higher temperatures. We also observe the well-known
phenomenon of magnetic catalysis for a strong enough magnetic field. However,
we note that if the effective coupling strength of the model decreases as a
function of magnetic field, it can trigger inverse magnetic catalysis in a
certain window of this functional dependence. Our model allows for the
simultaneous onset of dynamical chiral symmetry breaking and confinement for
each case. Qualitative as well as quantitative predictions of our simple but
effective model are in reasonably satisfactory agreement with lattice results
and other reliable and refined predictions based upon intricate continuum
studies of quantum chromodynamics.
|
2020-08-10
|
{'id': '2008.03847', 'title': 'Color, Flavor, Temperature and Magnetic Field Dependence of QCD Phase Diagram: Magnetic Catalysis and its Inverse', 'abstract': 'We study dynamical chiral symmetry breaking for quarks in the fundamental\nrepresentation of $SU(N_c)$ for $N_f$ number of light quark flavors. We also\ninvestigate the phase diagram of quantum chromodynamics at finite temperature\n$T$ and/or in the presence of a constant external magnetic field $eB$. The\nunified formalism for this analysis is provided by a symmetry-preserving\nSchwinger-Dyson equations treatment of a vector$\\times$vector contact\ninteraction model which encodes several well-established features of quantum\nchromodynamics to mimic the latter as closely as possible. Deconfinement and\nchiral symmetry restoration are triggered above a critical value of $N_f$ at\n$T=0=eB$. On the other hand, increasing temperature itself screens strong\ninteractions, thus ensuring that a smaller value of $N_f$ is sufficient to\nrestore chiral symmetry at higher temperatures. We also observe the well-known\nphenomenon of magnetic catalysis for a strong enough magnetic field. However,\nwe note that if the effective coupling strength of the model decreases as a\nfunction of magnetic field, it can trigger inverse magnetic catalysis in a\ncertain window of this functional dependence. Our model allows for the\nsimultaneous onset of dynamical chiral symmetry breaking and confinement for\neach case. Qualitative as well as quantitative predictions of our simple but\neffective model are in reasonably satisfactory agreement with lattice results\nand other reliable and refined predictions based upon intricate continuum\nstudies of quantum chromodynamics.', 'authors': ['Aftab Ahmad', 'Adnan Bashir', 'Marco A. Bedolla', 'J. J. Cobos-Martínez'], 'published': '2020-08-10 00:54:09+00:00', 'categories': ['hep-ph', 'nucl-th'], 'url': 'http://arxiv.org/abs/2008.03847v1'}
|
8
|
2508.00602
| 2,508.00602
| null |
LeakSealer: A Semisupervised Defense for LLMs Against Prompt Injection and Leakage Attacks
|
The generalization capabilities of Large Language Models (LLMs) have led to
their widespread deployment across various applications. However, this
increased adoption has introduced several security threats, notably in the
forms of jailbreaking and data leakage attacks. Additionally, Retrieval
Augmented Generation (RAG), while enhancing context-awareness in LLM responses,
has inadvertently introduced vulnerabilities that can result in the leakage of
sensitive information. Our contributions are twofold. First, we introduce a
methodology to analyze historical interaction data from an LLM system, enabling
the generation of usage maps categorized by topics (including adversarial
interactions). This approach further provides forensic insights for tracking
the evolution of jailbreaking attack patterns. Second, we propose LeakSealer, a
model-agnostic framework that combines static analysis for forensic insights
with dynamic defenses in a Human-In-The-Loop (HITL) pipeline. This technique
identifies topic groups and detects anomalous patterns, allowing for proactive
defense mechanisms. We empirically evaluate LeakSealer under two scenarios: (1)
jailbreak attempts, employing a public benchmark dataset, and (2) PII leakage,
supported by a curated dataset of labeled LLM interactions. In the static
setting, LeakSealer achieves the highest precision and recall on the ToxicChat
dataset when identifying prompt injection. In the dynamic setting, PII leakage
detection achieves an AUPRC of $0.97$, significantly outperforming baselines
such as Llama Guard.
|
2025-08-01
|
{'id': '2508.00602', 'title': 'LeakSealer: A Semisupervised Defense for LLMs Against Prompt Injection and Leakage Attacks', 'abstract': 'The generalization capabilities of Large Language Models (LLMs) have led to\ntheir widespread deployment across various applications. However, this\nincreased adoption has introduced several security threats, notably in the\nforms of jailbreaking and data leakage attacks. Additionally, Retrieval\nAugmented Generation (RAG), while enhancing context-awareness in LLM responses,\nhas inadvertently introduced vulnerabilities that can result in the leakage of\nsensitive information. Our contributions are twofold. First, we introduce a\nmethodology to analyze historical interaction data from an LLM system, enabling\nthe generation of usage maps categorized by topics (including adversarial\ninteractions). This approach further provides forensic insights for tracking\nthe evolution of jailbreaking attack patterns. Second, we propose LeakSealer, a\nmodel-agnostic framework that combines static analysis for forensic insights\nwith dynamic defenses in a Human-In-The-Loop (HITL) pipeline. This technique\nidentifies topic groups and detects anomalous patterns, allowing for proactive\ndefense mechanisms. We empirically evaluate LeakSealer under two scenarios: (1)\njailbreak attempts, employing a public benchmark dataset, and (2) PII leakage,\nsupported by a curated dataset of labeled LLM interactions. In the static\nsetting, LeakSealer achieves the highest precision and recall on the ToxicChat\ndataset when identifying prompt injection. In the dynamic setting, PII leakage\ndetection achieves an AUPRC of $0.97$, significantly outperforming baselines\nsuch as Llama Guard.', 'authors': ['Francesco Panebianco', 'Stefano Bonfanti', 'Francesco Trovò', 'Michele Carminati'], 'published': '2025-08-01 13:04:28+00:00', 'categories': ['cs.CR', 'cs.AI', 'cs.LG'], 'url': 'http://arxiv.org/abs/2508.00602v1'}
|
9
|
2508.00228
| 2,508.00228
| null |
Benchmarking XRootD-HTTPS on 400Gbps Links with Variable Latencies
|
In anticipation of the High Luminosity-LHC era, there is a critical need to
oversee software readiness for upcoming growth in network traffic for
production and user data analysis access. This paper looks into software and
hardware required improvements in US-CMS Tier-2 sites to be able to sustain and
meet the projected 400 Gbps bandwidth demands while tackling the challenge
posed by varying latencies between sites. Specifically, our study focuses on
identifying the performance of XRootD HTTP third-party copies across multiple
400 Gbps links and exploring different host and transfer configurations. Our
approach involves systematic testing with variations in the number of origins
per cluster and CPU allocations for each origin. By replicating real network
conditions and creating network "loops" that traverse multiple switches across
the wide area network, we are able to replicate authentic network conditions
|
2025-08-01
|
{'id': '2508.00228', 'title': 'Benchmarking XRootD-HTTPS on 400Gbps Links with Variable Latencies', 'abstract': 'In anticipation of the High Luminosity-LHC era, there is a critical need to\noversee software readiness for upcoming growth in network traffic for\nproduction and user data analysis access. This paper looks into software and\nhardware required improvements in US-CMS Tier-2 sites to be able to sustain and\nmeet the projected 400 Gbps bandwidth demands while tackling the challenge\nposed by varying latencies between sites. Specifically, our study focuses on\nidentifying the performance of XRootD HTTP third-party copies across multiple\n400 Gbps links and exploring different host and transfer configurations. Our\napproach involves systematic testing with variations in the number of origins\nper cluster and CPU allocations for each origin. By replicating real network\nconditions and creating network "loops" that traverse multiple switches across\nthe wide area network, we are able to replicate authentic network conditions', 'authors': ['Aashay Arora', 'Diego Davila', 'Frank Würthwein', 'John Graham', 'Dima Mishin', 'Justas Balcas', 'Tom Lehman', 'Xi Yang', 'Chin Guok', 'Harvey Newman'], 'published': '2025-08-01 00:21:24+00:00', 'categories': ['cs.NI'], 'url': 'http://arxiv.org/abs/2508.00228v1'}
|
10
|
2508.00329
| 2,508.00329
| null |
Radially Locked Sun-Ray Patterns in Autocatalytic Reaction-Diffusion-Advection Systems
|
Traveling fronts ubiquitous in physics, chemistry, and biology are prone to
transverse cellular deformations due to diffusive or convective instabilities.
Here we show both theoretically and experimentally that new patterns can be
obtained if the destabilization is triggered around a front locked radially by
advection. Specifically, angularly shifting sun-ray-like patterns can develop
around radially advected autocatalytic fronts due to a diffusive instability
developing when the autocatalyst X and the reactant Y diffuse at different
rates. The properties of these shining-star structures can be controlled by
tuning the flow rate $Q$ and the ratio of diffusion coefficients $$ as
evidenced by linear stability analysis, nonlinear simulations, and experiments
on the chlorite-tetrathionate reaction.
|
2025-08-01
|
{'id': '2508.00329', 'title': 'Radially Locked Sun-Ray Patterns in Autocatalytic Reaction-Diffusion-Advection Systems', 'abstract': 'Traveling fronts ubiquitous in physics, chemistry, and biology are prone to\ntransverse cellular deformations due to diffusive or convective instabilities.\nHere we show both theoretically and experimentally that new patterns can be\nobtained if the destabilization is triggered around a front locked radially by\nadvection. Specifically, angularly shifting sun-ray-like patterns can develop\naround radially advected autocatalytic fronts due to a diffusive instability\ndeveloping when the autocatalyst X and the reactant Y diffuse at different\nrates. The properties of these shining-star structures can be controlled by\ntuning the flow rate $Q$ and the ratio of diffusion coefficients $\\delta$ as\nevidenced by linear stability analysis, nonlinear simulations, and experiments\non the chlorite-tetrathionate reaction.', 'authors': ['Surya Narayan Maharana', 'Luka Negrojević', 'Alessandro Comolli', 'Anne De Wit'], 'published': '2025-08-01 05:30:43+00:00', 'categories': ['physics.class-ph'], 'url': 'http://arxiv.org/abs/2508.00329v2'}
|
11
|
2508.05628v1
| 2,508.05628
| 1
|
H-Net++: Hierarchical Dynamic Chunking for Tokenizer-Free Language Modelling in Morphologically-Rich Languages
|
Byte-level language models eliminate fragile tokenizers but face
computational challenges in morphologically-rich languages (MRLs), where words
span many bytes. We propose H-NET++, a hierarchical dynamic-chunking model that
learns linguistically-informed segmentation through end-to-end training. Key
innovations include: (1) a lightweight Transformer context-mixer (1.9M
parameters) for cross-chunk attention, (2) a two-level latent hyper-prior for
document-level consistency, (3) specialized handling of orthographic artifacts
(e.g. Persian ZWNJ), and (4) curriculum-based training with staged sequence
lengths. On a 1.4B-token Persian corpus, H-NET++ achieves state-of-the-art
results: 0.159 BPB reduction versus BPE-based GPT-2-fa (12% better
compression), 5.4pp gain on ParsGLUE, 53% improved robustness to ZWNJ
corruption, and 73.8% F1 on gold morphological boundaries. Our learned chunks
align with Persian morphology without explicit supervision, demonstrating that
hierarchical dynamic chunking provides an effective tokenizer-free solution for
MRLs while maintaining computational efficiency.
|
2025-08-07
|
{'id': '2508.05628v1', 'title': 'H-Net++: Hierarchical Dynamic Chunking for Tokenizer-Free Language Modelling in Morphologically-Rich Languages', 'abstract': 'Byte-level language models eliminate fragile tokenizers but face\ncomputational challenges in morphologically-rich languages (MRLs), where words\nspan many bytes. We propose H-NET++, a hierarchical dynamic-chunking model that\nlearns linguistically-informed segmentation through end-to-end training. Key\ninnovations include: (1) a lightweight Transformer context-mixer (1.9M\nparameters) for cross-chunk attention, (2) a two-level latent hyper-prior for\ndocument-level consistency, (3) specialized handling of orthographic artifacts\n(e.g. Persian ZWNJ), and (4) curriculum-based training with staged sequence\nlengths. On a 1.4B-token Persian corpus, H-NET++ achieves state-of-the-art\nresults: 0.159 BPB reduction versus BPE-based GPT-2-fa (12% better\ncompression), 5.4pp gain on ParsGLUE, 53% improved robustness to ZWNJ\ncorruption, and 73.8% F1 on gold morphological boundaries. Our learned chunks\nalign with Persian morphology without explicit supervision, demonstrating that\nhierarchical dynamic chunking provides an effective tokenizer-free solution for\nMRLs while maintaining computational efficiency.', 'authors': ['Mehrdad Zakershahrak', 'Samira Ghodratnama'], 'published': '2025-08-07 17:59:01+00:00', 'categories': ['cs.CL', 'cs.AI'], 'url': 'http://arxiv.org/abs/2508.05628v1'}
|
12
|
2508.00311
| 2,508.00311
| null |
DocTron-Formula: Generalized Formula Recognition in Complex and Structured Scenarios
|
Optical Character Recognition (OCR) for mathematical formula is essential for
the intelligent analysis of scientific literature. However, both task-specific
and general vision-language models often struggle to handle the structural
diversity, complexity, and real-world variability inherent in mathematical
content. In this work, we present DocTron-Formula, a unified framework built
upon general vision-language models, thereby eliminating the need for
specialized architectures. Furthermore, we introduce CSFormula, a large-scale
and challenging dataset that encompasses multidisciplinary and structurally
complex formulas at the line, paragraph, and page levels. Through
straightforward supervised fine-tuning, our approach achieves state-of-the-art
performance across a variety of styles, scientific domains, and complex
layouts. Experimental results demonstrate that our method not only surpasses
specialized models in terms of accuracy and robustness, but also establishes a
new paradigm for the automated understanding of complex scientific documents.
|
2025-08-01
|
{'id': '2508.00311', 'title': 'DocTron-Formula: Generalized Formula Recognition in Complex and Structured Scenarios', 'abstract': 'Optical Character Recognition (OCR) for mathematical formula is essential for\nthe intelligent analysis of scientific literature. However, both task-specific\nand general vision-language models often struggle to handle the structural\ndiversity, complexity, and real-world variability inherent in mathematical\ncontent. In this work, we present DocTron-Formula, a unified framework built\nupon general vision-language models, thereby eliminating the need for\nspecialized architectures. Furthermore, we introduce CSFormula, a large-scale\nand challenging dataset that encompasses multidisciplinary and structurally\ncomplex formulas at the line, paragraph, and page levels. Through\nstraightforward supervised fine-tuning, our approach achieves state-of-the-art\nperformance across a variety of styles, scientific domains, and complex\nlayouts. Experimental results demonstrate that our method not only surpasses\nspecialized models in terms of accuracy and robustness, but also establishes a\nnew paradigm for the automated understanding of complex scientific documents.', 'authors': ['Yufeng Zhong', 'Zhixiong Zeng', 'Lei Chen', 'Longrong Yang', 'Liming Zheng', 'Jing Huang', 'Siqi Yang', 'Lin Ma'], 'published': '2025-08-01 04:34:17+00:00', 'categories': ['cs.CV'], 'url': 'http://arxiv.org/abs/2508.00311v1'}
|
13
|
2508.00539
| 2,508.00539
| null |
Phase-Locked SNR Band Selection for Weak Mineral Signal Detection in Hyperspectral Imagery
|
Hyperspectral imaging offers detailed spectral information for mineral
mapping; however, weak mineral signatures are often masked by noisy and
redundant bands, limiting detection performance. To address this, we propose a
two-stage integrated framework for enhanced mineral detection in the Cuprite
mining district. In the first stage, we compute the signal-to-noise ratio (SNR)
for each spectral band and apply a phase-locked thresholding technique to
discard low-SNR bands, effectively removing redundancy and suppressing
background noise. Savitzky-Golay filtering is then employed for spectral
smoothing, serving a dual role first to stabilize trends during band selection,
and second to preserve fine-grained spectral features during preprocessing. In
the second stage, the refined HSI data is reintroduced into the model, where
KMeans clustering is used to extract 12 endmember spectra (W1 custom), followed
by non negative least squares (NNLS) for abundance unmixing. The resulting
endmembers are quantitatively compared with laboratory spectra (W1 raw) using
cosine similarity and RMSE metrics. Experimental results confirm that our
proposed pipeline improves unmixing accuracy and enhances the detection of weak
mineral zones. This two-pass strategy demonstrates a practical and reproducible
solution for spectral dimensionality reduction and unmixing in geological HSI
applications.
|
2025-08-01
|
{'id': '2508.00539', 'title': 'Phase-Locked SNR Band Selection for Weak Mineral Signal Detection in Hyperspectral Imagery', 'abstract': 'Hyperspectral imaging offers detailed spectral information for mineral\nmapping; however, weak mineral signatures are often masked by noisy and\nredundant bands, limiting detection performance. To address this, we propose a\ntwo-stage integrated framework for enhanced mineral detection in the Cuprite\nmining district. In the first stage, we compute the signal-to-noise ratio (SNR)\nfor each spectral band and apply a phase-locked thresholding technique to\ndiscard low-SNR bands, effectively removing redundancy and suppressing\nbackground noise. Savitzky-Golay filtering is then employed for spectral\nsmoothing, serving a dual role first to stabilize trends during band selection,\nand second to preserve fine-grained spectral features during preprocessing. In\nthe second stage, the refined HSI data is reintroduced into the model, where\nKMeans clustering is used to extract 12 endmember spectra (W1 custom), followed\nby non negative least squares (NNLS) for abundance unmixing. The resulting\nendmembers are quantitatively compared with laboratory spectra (W1 raw) using\ncosine similarity and RMSE metrics. Experimental results confirm that our\nproposed pipeline improves unmixing accuracy and enhances the detection of weak\nmineral zones. This two-pass strategy demonstrates a practical and reproducible\nsolution for spectral dimensionality reduction and unmixing in geological HSI\napplications.', 'authors': ['Judy X Yang'], 'published': '2025-08-01 11:26:49+00:00', 'categories': ['cs.LG', 'Cs'], 'url': 'http://arxiv.org/abs/2508.00539v2'}
|
14
|
2508.00476
| 2,508.00476
| null |
GETALP@AutoMin 2025: Leveraging RAG to Answer Questions based on Meeting Transcripts
|
This paper documents GETALP's submission to the Third Run of the Automatic
Minuting Shared Task at SIGDial 2025. We participated in Task B:
question-answering based on meeting transcripts. Our method is based on a
retrieval augmented generation (RAG) system and Abstract Meaning
Representations (AMR). We propose three systems combining these two approaches.
Our results show that incorporating AMR leads to high-quality responses for
approximately 35% of the questions and provides notable improvements in
answering questions that involve distinguishing between different participants
(e.g., who questions).
|
2025-08-01
|
{'id': '2508.00476', 'title': 'GETALP@AutoMin 2025: Leveraging RAG to Answer Questions based on Meeting Transcripts', 'abstract': "This paper documents GETALP's submission to the Third Run of the Automatic\nMinuting Shared Task at SIGDial 2025. We participated in Task B:\nquestion-answering based on meeting transcripts. Our method is based on a\nretrieval augmented generation (RAG) system and Abstract Meaning\nRepresentations (AMR). We propose three systems combining these two approaches.\nOur results show that incorporating AMR leads to high-quality responses for\napproximately 35% of the questions and provides notable improvements in\nanswering questions that involve distinguishing between different participants\n(e.g., who questions).", 'authors': ['Jeongwoo Kang', 'Markarit Vartampetian', 'Felix Herron', 'Yongxin Zhou', 'Diandra Fabre', 'Gabriela Gonzalez-Saez'], 'published': '2025-08-01 09:51:05+00:00', 'categories': ['cs.CL'], 'url': 'http://arxiv.org/abs/2508.00476v1'}
|
15
|
2508.00721
| 2,508.00721
| null |
FMPlug: Plug-In Foundation Flow-Matching Priors for Inverse Problems
|
We present FMPlug, a novel plug-in framework that enhances foundation
flow-matching (FM) priors for solving ill-posed inverse problems. Unlike
traditional approaches that rely on domain-specific or untrained priors, FMPlug
smartly leverages two simple but powerful insights: the similarity between
observed and desired objects and the Gaussianity of generative flows. By
introducing a time-adaptive warm-up strategy and sharp Gaussianity
regularization, FMPlug unlocks the true potential of domain-agnostic foundation
models. Our method beats state-of-the-art methods that use foundation FM priors
by significant margins, on image super-resolution and Gaussian deblurring.
|
2025-08-01
|
{'id': '2508.00721', 'title': 'FMPlug: Plug-In Foundation Flow-Matching Priors for Inverse Problems', 'abstract': 'We present FMPlug, a novel plug-in framework that enhances foundation\nflow-matching (FM) priors for solving ill-posed inverse problems. Unlike\ntraditional approaches that rely on domain-specific or untrained priors, FMPlug\nsmartly leverages two simple but powerful insights: the similarity between\nobserved and desired objects and the Gaussianity of generative flows. By\nintroducing a time-adaptive warm-up strategy and sharp Gaussianity\nregularization, FMPlug unlocks the true potential of domain-agnostic foundation\nmodels. Our method beats state-of-the-art methods that use foundation FM priors\nby significant margins, on image super-resolution and Gaussian deblurring.', 'authors': ['Yuxiang Wan', 'Ryan Devera', 'Wenjie Zhang', 'Ju Sun'], 'published': '2025-08-01 15:40:37+00:00', 'categories': ['eess.IV', 'cs.CV', 'cs.LG', 'eess.SP'], 'url': 'http://arxiv.org/abs/2508.00721v1'}
|
16
|
2508.00700
| 2,508.007
| null |
Is LLM-Generated Code More Maintainable \& Reliable than Human-Written Code?
|
Background: The rise of Large Language Models (LLMs) in software development
has opened new possibilities for code generation. Despite the widespread use of
this technology, it remains unclear how well LLMs generate code solutions in
terms of software quality and how they compare to human-written code. Aims:
This study compares the internal quality attributes of LLM-generated and
human-written code. Method: Our empirical study integrates datasets of coding
tasks, three LLM configurations (zero-shot, few-shot, and fine-tuning), and
SonarQube to assess software quality. The dataset comprises Python code
solutions across three difficulty levels: introductory, interview, and
competition. We analyzed key code quality metrics, including maintainability
and reliability, and the estimated effort required to resolve code issues.
Results: Our analysis shows that LLM-generated code has fewer bugs and requires
less effort to fix them overall. Interestingly, fine-tuned models reduced the
prevalence of high-severity issues, such as blocker and critical bugs, and
shifted them to lower-severity categories, but decreased the model's
performance. In competition-level problems, the LLM solutions sometimes
introduce structural issues that are not present in human-written code.
Conclusion: Our findings provide valuable insights into the quality of
LLM-generated code; however, the introduction of critical issues in more
complex scenarios highlights the need for a systematic evaluation and
validation of LLM solutions. Our work deepens the understanding of the
strengths and limitations of LLMs for code generation.
|
2025-08-01
|
{'id': '2508.00700', 'title': 'Is LLM-Generated Code More Maintainable \\& Reliable than Human-Written Code?', 'abstract': "Background: The rise of Large Language Models (LLMs) in software development\nhas opened new possibilities for code generation. Despite the widespread use of\nthis technology, it remains unclear how well LLMs generate code solutions in\nterms of software quality and how they compare to human-written code. Aims:\nThis study compares the internal quality attributes of LLM-generated and\nhuman-written code. Method: Our empirical study integrates datasets of coding\ntasks, three LLM configurations (zero-shot, few-shot, and fine-tuning), and\nSonarQube to assess software quality. The dataset comprises Python code\nsolutions across three difficulty levels: introductory, interview, and\ncompetition. We analyzed key code quality metrics, including maintainability\nand reliability, and the estimated effort required to resolve code issues.\nResults: Our analysis shows that LLM-generated code has fewer bugs and requires\nless effort to fix them overall. Interestingly, fine-tuned models reduced the\nprevalence of high-severity issues, such as blocker and critical bugs, and\nshifted them to lower-severity categories, but decreased the model's\nperformance. In competition-level problems, the LLM solutions sometimes\nintroduce structural issues that are not present in human-written code.\nConclusion: Our findings provide valuable insights into the quality of\nLLM-generated code; however, the introduction of critical issues in more\ncomplex scenarios highlights the need for a systematic evaluation and\nvalidation of LLM solutions. Our work deepens the understanding of the\nstrengths and limitations of LLMs for code generation.", 'authors': ['Alfred Santa Molison', 'Marcia Moraes', 'Glaucia Melo', 'Fabio Santos', 'Wesley K. G. Assuncao'], 'published': '2025-08-01 15:17:34+00:00', 'categories': ['cs.SE'], 'url': 'http://arxiv.org/abs/2508.00700v1'}
|
17
|
2508.00288
| 2,508.00288
| null |
UAV-ON: A Benchmark for Open-World Object Goal Navigation with Aerial Agents
|
Aerial navigation is a fundamental yet underexplored capability in embodied
intelligence, enabling agents to operate in large-scale, unstructured
environments where traditional navigation paradigms fall short. However, most
existing research follows the Vision-and-Language Navigation (VLN) paradigm,
which heavily depends on sequential linguistic instructions, limiting its
scalability and autonomy. To address this gap, we introduce UAV-ON, a benchmark
for large-scale Object Goal Navigation (ObjectNav) by aerial agents in
open-world environments, where agents operate based on high-level semantic
goals without relying on detailed instructional guidance as in VLN. UAV-ON
comprises 14 high-fidelity Unreal Engine environments with diverse semantic
regions and complex spatial layouts, covering urban, natural, and mixed-use
settings. It defines 1270 annotated target objects, each characterized by an
instance-level instruction that encodes category, physical footprint, and
visual descriptors, allowing grounded reasoning. These instructions serve as
semantic goals, introducing realistic ambiguity and complex reasoning
challenges for aerial agents. To evaluate the benchmark, we implement several
baseline methods, including Aerial ObjectNav Agent (AOA), a modular policy that
integrates instruction semantics with egocentric observations for long-horizon,
goal-directed exploration. Empirical results show that all baselines struggle
in this setting, highlighting the compounded challenges of aerial navigation
and semantic goal grounding. UAV-ON aims to advance research on scalable UAV
autonomy driven by semantic goal descriptions in complex real-world
environments.
|
2025-08-01
|
{'id': '2508.00288', 'title': 'UAV-ON: A Benchmark for Open-World Object Goal Navigation with Aerial Agents', 'abstract': 'Aerial navigation is a fundamental yet underexplored capability in embodied\nintelligence, enabling agents to operate in large-scale, unstructured\nenvironments where traditional navigation paradigms fall short. However, most\nexisting research follows the Vision-and-Language Navigation (VLN) paradigm,\nwhich heavily depends on sequential linguistic instructions, limiting its\nscalability and autonomy. To address this gap, we introduce UAV-ON, a benchmark\nfor large-scale Object Goal Navigation (ObjectNav) by aerial agents in\nopen-world environments, where agents operate based on high-level semantic\ngoals without relying on detailed instructional guidance as in VLN. UAV-ON\ncomprises 14 high-fidelity Unreal Engine environments with diverse semantic\nregions and complex spatial layouts, covering urban, natural, and mixed-use\nsettings. It defines 1270 annotated target objects, each characterized by an\ninstance-level instruction that encodes category, physical footprint, and\nvisual descriptors, allowing grounded reasoning. These instructions serve as\nsemantic goals, introducing realistic ambiguity and complex reasoning\nchallenges for aerial agents. To evaluate the benchmark, we implement several\nbaseline methods, including Aerial ObjectNav Agent (AOA), a modular policy that\nintegrates instruction semantics with egocentric observations for long-horizon,\ngoal-directed exploration. Empirical results show that all baselines struggle\nin this setting, highlighting the compounded challenges of aerial navigation\nand semantic goal grounding. UAV-ON aims to advance research on scalable UAV\nautonomy driven by semantic goal descriptions in complex real-world\nenvironments.', 'authors': ['Jianqiang Xiao', 'Yuexuan Sun', 'Yixin Shao', 'Boxi Gan', 'Rongqiang Liu', 'Yanjing Wu', 'Weili Gua', 'Xiang Deng'], 'published': '2025-08-01 03:23:06+00:00', 'categories': ['cs.RO', 'cs.CV'], 'url': 'http://arxiv.org/abs/2508.00288v1'}
|
18
|
2508.00668
| 2,508.00668
| null |
Advancing Quantum Information Science Pre-College Education: The Case for Learning Sciences Collaboration
|
As quantum information science advances and the need for pre-college
engagement grows, a critical question remains: How can young learners be
prepared to participate in a field so radically different from what they have
encountered before? This paper argues that meeting this challenge will require
strong interdisciplinary collaboration with the Learning Sciences (LS), a field
dedicated to understanding how people learn and designing theory-guided
environments to support learning. Drawing on lessons from previous STEM
education efforts, we discuss two key contributions of the learning sciences to
quantum information science (QIS) education. The first is design-based
research, the signature methodology of learning sciences, which can inform the
development, refinement, and scaling of effective QIS learning experiences. The
second is a framework for reshaping how learners reason about, learn and
participate in QIS practices through shifts in knowledge representations that
provide new forms of engagement and associated learning. We call for a two-way
partnership between quantum information science and the learning sciences, one
that not only supports learning in quantum concepts and practices but also
improves our understanding of how to teach and support learning in highly
complex domains. We also consider potential questions involved in bridging
these disciplinary communities and argue that the theoretical and practical
benefits justify the effort.
|
2025-08-01
|
{'id': '2508.00668', 'title': 'Advancing Quantum Information Science Pre-College Education: The Case for Learning Sciences Collaboration', 'abstract': 'As quantum information science advances and the need for pre-college\nengagement grows, a critical question remains: How can young learners be\nprepared to participate in a field so radically different from what they have\nencountered before? This paper argues that meeting this challenge will require\nstrong interdisciplinary collaboration with the Learning Sciences (LS), a field\ndedicated to understanding how people learn and designing theory-guided\nenvironments to support learning. Drawing on lessons from previous STEM\neducation efforts, we discuss two key contributions of the learning sciences to\nquantum information science (QIS) education. The first is design-based\nresearch, the signature methodology of learning sciences, which can inform the\ndevelopment, refinement, and scaling of effective QIS learning experiences. The\nsecond is a framework for reshaping how learners reason about, learn and\nparticipate in QIS practices through shifts in knowledge representations that\nprovide new forms of engagement and associated learning. We call for a two-way\npartnership between quantum information science and the learning sciences, one\nthat not only supports learning in quantum concepts and practices but also\nimproves our understanding of how to teach and support learning in highly\ncomplex domains. We also consider potential questions involved in bridging\nthese disciplinary communities and argue that the theoretical and practical\nbenefits justify the effort.', 'authors': ['Raquel Coelho', 'Roy Pea', 'Christian Schunn', 'Jinglei Cheng', 'Junyu Liu'], 'published': '2025-08-01 14:41:18+00:00', 'categories': ['physics.ed-ph', 'cs.AI', 'cs.CY', 'quant-ph'], 'url': 'http://arxiv.org/abs/2508.00668v1'}
|
19
|
2508.00725
| 2,508.00725
| null |
The SN 1987A Cooling Bound on Dark Matter Absorption in Electron Targets
|
We present new supernova (SN 1987A) cooling bounds on sub-MeV fermionic dark
matter with effective couplings to electrons. These bounds probe the parameter
space relevant for direct detection experiments in which dark matter can be
absorbed by the target material, showing strong complementarity with indirect
searches and constraints from dark matter overproduction. Crucially, our limits
exclude the projected sensitivity regions of current and upcoming direct
detection experiments. Since these conclusions are a priori not valid for light
mediators, we extend our analysis to this case. We show that sub-GeV mediators
can be produced resonantly both in supernova cores and in the early Universe,
altering the SN 1987A analysis for effective couplings. Still, a combination of
supernova cooling constraints and limits from dark matter overproduction
excludes the entire parameter space relevant for direct detection in this case.
|
2025-08-01
|
{'id': '2508.00725', 'title': 'The SN 1987A Cooling Bound on Dark Matter Absorption in Electron Targets', 'abstract': 'We present new supernova (SN 1987A) cooling bounds on sub-MeV fermionic dark\nmatter with effective couplings to electrons. These bounds probe the parameter\nspace relevant for direct detection experiments in which dark matter can be\nabsorbed by the target material, showing strong complementarity with indirect\nsearches and constraints from dark matter overproduction. Crucially, our limits\nexclude the projected sensitivity regions of current and upcoming direct\ndetection experiments. Since these conclusions are a priori not valid for light\nmediators, we extend our analysis to this case. We show that sub-GeV mediators\ncan be produced resonantly both in supernova cores and in the early Universe,\naltering the SN 1987A analysis for effective couplings. Still, a combination of\nsupernova cooling constraints and limits from dark matter overproduction\nexcludes the entire parameter space relevant for direct detection in this case.', 'authors': ['Claudio Andrea Manzari', 'Jorge Martin Camalich', 'Jonas Spinner', 'Robert Ziegler'], 'published': '2025-08-01 15:48:31+00:00', 'categories': ['hep-ph', 'astro-ph.HE', 'hep-ex'], 'url': 'http://arxiv.org/abs/2508.00725v1'}
|
20
|
2508.00633
| 2,508.00633
| null |
Dynamics and Coherence for the Free Cornering with Protocol Choice
|
We present a term rewriting system that models the dynamic aspects of the
free cornering with protocol choice of a monoidal category, which has been
proposed as a categorical model of process interaction. This term rewriting
system is confluent and terminating in an appropriate sense. We use this
machinery to prove a coherence theorem for the free cornering with protocol
choice.
|
2025-08-01
|
{'id': '2508.00633', 'title': 'Dynamics and Coherence for the Free Cornering with Protocol Choice', 'abstract': 'We present a term rewriting system that models the dynamic aspects of the\nfree cornering with protocol choice of a monoidal category, which has been\nproposed as a categorical model of process interaction. This term rewriting\nsystem is confluent and terminating in an appropriate sense. We use this\nmachinery to prove a coherence theorem for the free cornering with protocol\nchoice.', 'authors': ['Chad Nester', 'Niels Voorneveld'], 'published': '2025-08-01 13:45:24+00:00', 'categories': ['math.CT', 'cs.LO'], 'url': 'http://arxiv.org/abs/2508.00633v1'}
|
21
|
2508.00587
| 2,508.00587
| null |
Uncertainty-Aware Likelihood Ratio Estimation for Pixel-Wise Out-of-Distribution Detection
|
Semantic segmentation models trained on known object classes often fail in
real-world autonomous driving scenarios by confidently misclassifying unknown
objects. While pixel-wise out-of-distribution detection can identify unknown
objects, existing methods struggle in complex scenes where rare object classes
are often confused with truly unknown objects. We introduce an
uncertainty-aware likelihood ratio estimation method that addresses these
limitations. Our approach uses an evidential classifier within a likelihood
ratio test to distinguish between known and unknown pixel features from a
semantic segmentation model, while explicitly accounting for uncertainty.
Instead of producing point estimates, our method outputs probability
distributions that capture uncertainty from both rare training examples and
imperfect synthetic outliers. We show that by incorporating uncertainty in this
way, outlier exposure can be leveraged more effectively. Evaluated on five
standard benchmark datasets, our method achieves the lowest average false
positive rate (2.5%) among state-of-the-art while maintaining high average
precision (90.91%) and incurring only negligible computational overhead. Code
is available at https://github.com/glasbruch/ULRE.
|
2025-08-01
|
{'id': '2508.00587', 'title': 'Uncertainty-Aware Likelihood Ratio Estimation for Pixel-Wise Out-of-Distribution Detection', 'abstract': 'Semantic segmentation models trained on known object classes often fail in\nreal-world autonomous driving scenarios by confidently misclassifying unknown\nobjects. While pixel-wise out-of-distribution detection can identify unknown\nobjects, existing methods struggle in complex scenes where rare object classes\nare often confused with truly unknown objects. We introduce an\nuncertainty-aware likelihood ratio estimation method that addresses these\nlimitations. Our approach uses an evidential classifier within a likelihood\nratio test to distinguish between known and unknown pixel features from a\nsemantic segmentation model, while explicitly accounting for uncertainty.\nInstead of producing point estimates, our method outputs probability\ndistributions that capture uncertainty from both rare training examples and\nimperfect synthetic outliers. We show that by incorporating uncertainty in this\nway, outlier exposure can be leveraged more effectively. Evaluated on five\nstandard benchmark datasets, our method achieves the lowest average false\npositive rate (2.5%) among state-of-the-art while maintaining high average\nprecision (90.91%) and incurring only negligible computational overhead. Code\nis available at https://github.com/glasbruch/ULRE.', 'authors': ['Marc Hölle', 'Walter Kellermann', 'Vasileios Belagiannis'], 'published': '2025-08-01 12:39:16+00:00', 'categories': ['cs.CV'], 'url': 'http://arxiv.org/abs/2508.00587v1'}
|
22
|
2508.00607
| 2,508.00607
| null |
Two-photon-assisted collisions in ultracold gases of polar molecules
|
We present a theoretical formalism to treat the ultracold dynamics of a pair
of colliding polar molecules submitted to two laser fields. We express the
dressed Hamiltonian including the dipole-dipole interaction of the colliding
molecular pair, both in their ground and electronic excited states, as well as
their interaction with the two laser fields. We apply adiabatic elimination of
the electronic excited state to reduce the size of the dressed-state basis in
which the dressed Hamiltoninan is expressed. In an application, we investigate
the feasibility of two-photon collisional shielding between two molecules,
which could be favored by the Raman resonance condition suppressing unwanted
spontaneous emission and photon scattering. We demonstrate the influence of the
laser Rabi frequencies on the dynamics through the computation of elastic,
inelastic, and reactive collision rates.
|
2025-08-01
|
{'id': '2508.00607', 'title': 'Two-photon-assisted collisions in ultracold gases of polar molecules', 'abstract': 'We present a theoretical formalism to treat the ultracold dynamics of a pair\nof colliding polar molecules submitted to two laser fields. We express the\ndressed Hamiltonian including the dipole-dipole interaction of the colliding\nmolecular pair, both in their ground and electronic excited states, as well as\ntheir interaction with the two laser fields. We apply adiabatic elimination of\nthe electronic excited state to reduce the size of the dressed-state basis in\nwhich the dressed Hamiltoninan is expressed. In an application, we investigate\nthe feasibility of two-photon collisional shielding between two \\NaK molecules,\nwhich could be favored by the Raman resonance condition suppressing unwanted\nspontaneous emission and photon scattering. We demonstrate the influence of the\nlaser Rabi frequencies on the dynamics through the computation of elastic,\ninelastic, and reactive collision rates.', 'authors': ['Charbel Karam', 'Gohar Hovhannesyan', 'Romain Vexiau', 'Maxence Lepers', 'Nadia Bouloufa-Maafa', 'Olivier Dulieu'], 'published': '2025-08-01 13:09:01+00:00', 'categories': ['cond-mat.quant-gas'], 'url': 'http://arxiv.org/abs/2508.00607v1'}
|
3,509
|
2501.17351
| 2,501.17351
| null |
Realtime Limb Trajectory Optimization for Humanoid Running Through Centroidal Angular Momentum Dynamics
|
One of the essential aspects of humanoid robot running is determining the
limb-swinging trajectories. During the flight phases, where the ground reaction
forces are not available for regulation, the limb swinging trajectories are
significant for the stability of the next stance phase. Due to the conservation
of angular momentum, improper leg and arm swinging results in highly tilted and
unsustainable body configurations at the next stance phase landing. In such
cases, the robotic system fails to maintain locomotion independent of the
stability of the center of mass trajectories. This problem is more apparent for
fast and high flight time trajectories. This paper proposes a real-time
nonlinear limb trajectory optimization problem for humanoid running. The
optimization problem is tested on two different humanoid robot models, and the
generated trajectories are verified using a running algorithm for both robots
in a simulation environment.
|
2025-01-29
|
{'id': '2501.17351', 'title': 'Realtime Limb Trajectory Optimization for Humanoid Running Through Centroidal Angular Momentum Dynamics', 'abstract': 'One of the essential aspects of humanoid robot running is determining the\nlimb-swinging trajectories. During the flight phases, where the ground reaction\nforces are not available for regulation, the limb swinging trajectories are\nsignificant for the stability of the next stance phase. Due to the conservation\nof angular momentum, improper leg and arm swinging results in highly tilted and\nunsustainable body configurations at the next stance phase landing. In such\ncases, the robotic system fails to maintain locomotion independent of the\nstability of the center of mass trajectories. This problem is more apparent for\nfast and high flight time trajectories. This paper proposes a real-time\nnonlinear limb trajectory optimization problem for humanoid running. The\noptimization problem is tested on two different humanoid robot models, and the\ngenerated trajectories are verified using a running algorithm for both robots\nin a simulation environment.', 'authors': ['Sait Sovukluk', 'Robert Schuller', 'Johannes Englsberger', 'Christian Ott'], 'published': '2025-01-29 00:06:01+00:00', 'categories': ['cs.RO'], 'url': 'http://arxiv.org/abs/2501.17351v3'}
|
23
|
2508.05616v1
| 2,508.05616
| 1
|
TrajEvo: Trajectory Prediction Heuristics Design via LLM-driven Evolution
|
Trajectory prediction is a critical task in modeling human behavior,
especially in safety-critical domains such as social robotics and autonomous
vehicle navigation. Traditional heuristics based on handcrafted rules often
lack accuracy and generalizability. Although deep learning approaches offer
improved performance, they typically suffer from high computational cost,
limited explainability, and, importantly, poor generalization to
out-of-distribution (OOD) scenarios. In this paper, we introduce TrajEvo, a
framework that leverages Large Language Models (LLMs) to automatically design
trajectory prediction heuristics. TrajEvo employs an evolutionary algorithm to
generate and refine prediction heuristics from past trajectory data. We propose
two key innovations: Cross-Generation Elite Sampling to encourage population
diversity, and a Statistics Feedback Loop that enables the LLM to analyze and
improve alternative predictions. Our evaluations demonstrate that TrajEvo
outperforms existing heuristic methods across multiple real-world datasets, and
notably surpasses both heuristic and deep learning methods in generalizing to
an unseen OOD real-world dataset. TrajEvo marks a promising step toward the
automated design of fast, explainable, and generalizable trajectory prediction
heuristics. We release our source code to facilitate future research at
https://github.com/ai4co/trajevo.
|
2025-08-07
|
{'id': '2508.05616v1', 'title': 'TrajEvo: Trajectory Prediction Heuristics Design via LLM-driven Evolution', 'abstract': 'Trajectory prediction is a critical task in modeling human behavior,\nespecially in safety-critical domains such as social robotics and autonomous\nvehicle navigation. Traditional heuristics based on handcrafted rules often\nlack accuracy and generalizability. Although deep learning approaches offer\nimproved performance, they typically suffer from high computational cost,\nlimited explainability, and, importantly, poor generalization to\nout-of-distribution (OOD) scenarios. In this paper, we introduce TrajEvo, a\nframework that leverages Large Language Models (LLMs) to automatically design\ntrajectory prediction heuristics. TrajEvo employs an evolutionary algorithm to\ngenerate and refine prediction heuristics from past trajectory data. We propose\ntwo key innovations: Cross-Generation Elite Sampling to encourage population\ndiversity, and a Statistics Feedback Loop that enables the LLM to analyze and\nimprove alternative predictions. Our evaluations demonstrate that TrajEvo\noutperforms existing heuristic methods across multiple real-world datasets, and\nnotably surpasses both heuristic and deep learning methods in generalizing to\nan unseen OOD real-world dataset. TrajEvo marks a promising step toward the\nautomated design of fast, explainable, and generalizable trajectory prediction\nheuristics. We release our source code to facilitate future research at\nhttps://github.com/ai4co/trajevo.', 'authors': ['Zhikai Zhao', 'Chuanbo Hua', 'Federico Berto', 'Kanghoon Lee', 'Zihan Ma', 'Jiachen Li', 'Jinkyoo Park'], 'published': '2025-08-07 17:55:10+00:00', 'categories': ['cs.LG', 'cs.AI', 'cs.NE', 'cs.RO'], 'url': 'http://arxiv.org/abs/2508.05616v1'}
|
24
|
2508.00466
| 2,508.00466
| null |
Rare few-body decays of the Standard Model Higgs boson
|
We present a survey of rare and exclusive few-body decays of the standard
model (SM) Higgs boson, defined as those into two to four final particles with
branching fractions $B10^-5$. Studies of such decays can
be exploited to constrain Yukawa couplings of quarks and leptons, probe
flavour-changing Higgs decays, estimate backgrounds for exotic Higgs decays
into beyond-SM particles, and/or confirm quantum chromodynamics factorization
with small nonperturbative corrections. We collect the theoretical
$B$ values for about 70 unobserved Higgs rare decay channels,
indicating their current experimental limits, and estimating their expected
bounds in p-p collisions at the HL-LHC. Among those, we include 20 new decay
channels computed for the first time for ultrarare Higgs boson decays into
photons and/or neutrinos, radiative quark-flavour-changing exclusive decays,
and radiative decays into leptonium states. This survey can help guide and
prioritize upcoming experimental and theoretical studies of unobserved Higgs
boson decays.
|
2025-08-01
|
{'id': '2508.00466', 'title': 'Rare few-body decays of the Standard Model Higgs boson', 'abstract': 'We present a survey of rare and exclusive few-body decays of the standard\nmodel (SM) Higgs boson, defined as those into two to four final particles with\nbranching fractions $\\mathcal{B}\\lesssim 10^{-5}$. Studies of such decays can\nbe exploited to constrain Yukawa couplings of quarks and leptons, probe\nflavour-changing Higgs decays, estimate backgrounds for exotic Higgs decays\ninto beyond-SM particles, and/or confirm quantum chromodynamics factorization\nwith small nonperturbative corrections. We collect the theoretical\n$\\mathcal{B}$ values for about 70 unobserved Higgs rare decay channels,\nindicating their current experimental limits, and estimating their expected\nbounds in p-p collisions at the HL-LHC. Among those, we include 20 new decay\nchannels computed for the first time for ultrarare Higgs boson decays into\nphotons and/or neutrinos, radiative quark-flavour-changing exclusive decays,\nand radiative decays into leptonium states. This survey can help guide and\nprioritize upcoming experimental and theoretical studies of unobserved Higgs\nboson decays.', 'authors': ["David d'Enterria", 'Van Dung Le'], 'published': '2025-08-01 09:39:06+00:00', 'categories': ['hep-ph', 'hep-ex'], 'url': 'http://arxiv.org/abs/2508.00466v1'}
|
25
|
2508.00667
| 2,508.00667
| null |
Dependence of halo properties on central-satellite magnitude gaps through weak lensing measurements
|
The magnitude gap between the central and satellite galaxies encodes
information about the mass accretion history of a dark matter halo, and serves
as a useful observational probe for the mass distribution in a halo. In this
work, we perform the first weak lensing test of the connections between the
magnitude gap and the halo profile. We measure the halo profiles of isolated
central galaxies (ICGs) selected primarily from the SDSS Main Galaxy Sample.
Halo mass and concentration are inferred by fitting stacked lensing profiles in
bins of central luminosity, $L_c$, and the central-satellite magnitude
gap, $L_gap$. We detect dependence on the magnitude gap in both halo
properties. The dependence is the strongest in the ICG luminosity range of
$10^10.3<L_c[h^-2L_]10^10.7$, where halos with smaller
gaps have higher masses and lower concentrations. When $10^10.7
<L_c[h^-2L_] 10^11.1$, however, no significant gap dependence is
detected. In the range of $10^9.9<L_c[h^-2L_] 10^10.3$, a disordering of the gap dependence is marginally observable. We
compare the observational results with predictions by two lightcone catalogs
built from the Illustris TNG300 and the Millennium simulations. The gap
dependence in the two mock samples show overall consistency with observations,
but neither matches them in all $L_c$ bins to a quantitative level. We
also compare the significance of the gap dependence on halo mass and
concentration and find that our measurement prefers gap dependence in both
parameters, while the halo mass dependence is preferred over the concentration
if only one of the two dependencies is allowed.
|
2025-08-01
|
{'id': '2508.00667', 'title': 'Dependence of halo properties on central-satellite magnitude gaps through weak lensing measurements', 'abstract': 'The magnitude gap between the central and satellite galaxies encodes\ninformation about the mass accretion history of a dark matter halo, and serves\nas a useful observational probe for the mass distribution in a halo. In this\nwork, we perform the first weak lensing test of the connections between the\nmagnitude gap and the halo profile. We measure the halo profiles of isolated\ncentral galaxies (ICGs) selected primarily from the SDSS Main Galaxy Sample.\nHalo mass and concentration are inferred by fitting stacked lensing profiles in\nbins of central luminosity, $L_\\mathrm{c}$, and the central-satellite magnitude\ngap, $L_\\mathrm{gap}$. We detect dependence on the magnitude gap in both halo\nproperties. The dependence is the strongest in the ICG luminosity range of\n$10^{10.3}<L_\\mathrm{c}[h^{-2}L_\\odot]\\leq 10^{10.7}$, where halos with smaller\ngaps have higher masses and lower concentrations. When $10^{10.7}\n<L_c[h^{-2}L_\\odot] \\leq 10^{11.1}$, however, no significant gap dependence is\ndetected. In the range of $10^{9.9}<L_\\mathrm{c}[h^{-2}L_\\odot] \\leq\n10^{10.3}$, a disordering of the gap dependence is marginally observable. We\ncompare the observational results with predictions by two lightcone catalogs\nbuilt from the Illustris TNG300 and the Millennium simulations. The gap\ndependence in the two mock samples show overall consistency with observations,\nbut neither matches them in all $L_\\mathrm{c}$ bins to a quantitative level. We\nalso compare the significance of the gap dependence on halo mass and\nconcentration and find that our measurement prefers gap dependence in both\nparameters, while the halo mass dependence is preferred over the concentration\nif only one of the two dependencies is allowed.', 'authors': ['Mingtao Yang', 'Jiaxin Han', 'Wenting Wang', 'Hekun Li', 'Cong Liu', 'Jun Zhang', 'Shuai Feng', 'Shiyin Shen', 'Zhenjie Liu', 'Xiaohu Yang', 'Yi Lu', 'Surhud More'], 'published': '2025-08-01 14:40:45+00:00', 'categories': ['astro-ph.CO', 'astro-ph.GA'], 'url': 'http://arxiv.org/abs/2508.00667v1'}
|
26
|
2508.00658
| 2,508.00658
| null |
Multi-Band Variable-Lag Granger Causality: A Unified Framework for Causal Time Series Inference across Frequencies
|
Understanding causal relationships in time series is fundamental to many
domains, including neuroscience, economics, and behavioral science. Granger
causality is one of the well-known techniques for inferring causality in time
series. Typically, Granger causality frameworks have a strong fix-lag
assumption between cause and effect, which is often unrealistic in complex
systems. While recent work on variable-lag Granger causality (VLGC) addresses
this limitation by allowing a cause to influence an effect with different time
lags at each time point, it fails to account for the fact that causal
interactions may vary not only in time delay but also across frequency bands.
For example, in brain signals, alpha-band activity may influence another region
with a shorter delay than slower delta-band oscillations. In this work, we
formalize Multi-Band Variable-Lag Granger Causality (MB-VLGC) and propose a
novel framework that generalizes traditional VLGC by explicitly modeling
frequency-dependent causal delays. We provide a formal definition of MB-VLGC,
demonstrate its theoretical soundness, and propose an efficient inference
pipeline. Extensive experiments across multiple domains demonstrate that our
framework significantly outperforms existing methods on both synthetic and
real-world datasets, confirming its broad applicability to any type of time
series data. Code and datasets are publicly available.
|
2025-08-01
|
{'id': '2508.00658', 'title': 'Multi-Band Variable-Lag Granger Causality: A Unified Framework for Causal Time Series Inference across Frequencies', 'abstract': 'Understanding causal relationships in time series is fundamental to many\ndomains, including neuroscience, economics, and behavioral science. Granger\ncausality is one of the well-known techniques for inferring causality in time\nseries. Typically, Granger causality frameworks have a strong fix-lag\nassumption between cause and effect, which is often unrealistic in complex\nsystems. While recent work on variable-lag Granger causality (VLGC) addresses\nthis limitation by allowing a cause to influence an effect with different time\nlags at each time point, it fails to account for the fact that causal\ninteractions may vary not only in time delay but also across frequency bands.\nFor example, in brain signals, alpha-band activity may influence another region\nwith a shorter delay than slower delta-band oscillations. In this work, we\nformalize Multi-Band Variable-Lag Granger Causality (MB-VLGC) and propose a\nnovel framework that generalizes traditional VLGC by explicitly modeling\nfrequency-dependent causal delays. We provide a formal definition of MB-VLGC,\ndemonstrate its theoretical soundness, and propose an efficient inference\npipeline. Extensive experiments across multiple domains demonstrate that our\nframework significantly outperforms existing methods on both synthetic and\nreal-world datasets, confirming its broad applicability to any type of time\nseries data. Code and datasets are publicly available.', 'authors': ['Chakattrai Sookkongwaree', 'Tattep Lakmuang', 'Chainarong Amornbunchornvej'], 'published': '2025-08-01 14:22:51+00:00', 'categories': ['cs.AI', 'cs.LG', 'econ.EM', 'stat.ME'], 'url': 'http://arxiv.org/abs/2508.00658v1'}
|
27
|
2508.00342
| 2,508.00342
| null |
Uniform electronic states and $s$-wave superconductivity in a strongly disordered high-entropy compound (RuRhPdIr)$_{0.6}$Pt$_{0.4}$Sb
|
High-entropy compounds, where multiple elements occupy a single
crystallographic site in a highly disordered manner, challenge conventional
understandings of electronic structure based on periodicity and well-defined
band dispersion. Here, we report a detailed nuclear magnetic resonance study of
the high-entropy superconductor (RuRhPdIr)$_0.6$Pt$_0.4$Sb, revealing a
spatially homogeneous electronic environment in the normal state, in stark
contrast to its crystallographically disordered lattice. The superconducting
state exhibits a small but solid Hebel-Slichter coherence peak followed by a
significant decrease in the nuclear spin-lattice relaxation rate, providing
compelling evidence for fully gapped $s$-wave pairing. Our findings not only
deepen the understanding of superconductivity in highly disordered quantum
materials but also open a new pathway for exploring novel superconducting
states in entropy-stabilized systems.
|
2025-08-01
|
{'id': '2508.00342', 'title': 'Uniform electronic states and $s$-wave superconductivity in a strongly disordered high-entropy compound (RuRhPdIr)$_{0.6}$Pt$_{0.4}$Sb', 'abstract': 'High-entropy compounds, where multiple elements occupy a single\ncrystallographic site in a highly disordered manner, challenge conventional\nunderstandings of electronic structure based on periodicity and well-defined\nband dispersion. Here, we report a detailed nuclear magnetic resonance study of\nthe high-entropy superconductor (RuRhPdIr)$_{0.6}$Pt$_{0.4}$Sb, revealing a\nspatially homogeneous electronic environment in the normal state, in stark\ncontrast to its crystallographically disordered lattice. The superconducting\nstate exhibits a small but solid Hebel-Slichter coherence peak followed by a\nsignificant decrease in the nuclear spin-lattice relaxation rate, providing\ncompelling evidence for fully gapped $s$-wave pairing. Our findings not only\ndeepen the understanding of superconductivity in highly disordered quantum\nmaterials but also open a new pathway for exploring novel superconducting\nstates in entropy-stabilized systems.', 'authors': ['Yufu Yamada', 'Shunsaku Kitagawa', 'Taishi Ihara', 'Kenji Ishida', 'Naoto Uematsu', 'Daigorou Hirai', 'Koshi Takenaka'], 'published': '2025-08-01 06:08:12+00:00', 'categories': ['cond-mat.supr-con', 'cond-mat.mtrl-sci', 'cond-mat.str-el'], 'url': 'http://arxiv.org/abs/2508.00342v1'}
|
28
|
2508.00720
| 2,508.0072
| null |
Continuum damping of topologically-protected edge modes at the boundary of non-uniformly magnetized cylindrical plasmas
|
We present an extension of recently discovered collisionlessly damped yet
topologically protected surface plasma waves (TSPWs), from a simplified slab
geometry to a cylindrical plasma-vacuum interface. A distinctive feature of
these modes-emerging above the electron cyclotron frequency-is their
collisionless damping, which arises from resonant coupling to a continuum of
upper-hybrid modes localized within a smooth plasma-vacuum transition layer. We
demonstrate both temporal and spatial damping of TSPWs at the boundary of a
cylindrical magnetized plasma. Furthermore, we show that in the presence of a
non-uniform magnetic field, the TSPW exhibits a smooth and continuous
transition, highlighting its topological robustness under magnetic
inhomogeneity.
|
2025-08-01
|
{'id': '2508.00720', 'title': 'Continuum damping of topologically-protected edge modes at the boundary of non-uniformly magnetized cylindrical plasmas', 'abstract': 'We present an extension of recently discovered collisionlessly damped yet\ntopologically protected surface plasma waves (TSPWs), from a simplified slab\ngeometry to a cylindrical plasma-vacuum interface. A distinctive feature of\nthese modes-emerging above the electron cyclotron frequency-is their\ncollisionless damping, which arises from resonant coupling to a continuum of\nupper-hybrid modes localized within a smooth plasma-vacuum transition layer. We\ndemonstrate both temporal and spatial damping of TSPWs at the boundary of a\ncylindrical magnetized plasma. Furthermore, we show that in the presence of a\nnon-uniform magnetic field, the TSPW exhibits a smooth and continuous\ntransition, highlighting its topological robustness under magnetic\ninhomogeneity.', 'authors': ['Roopendra Singh Rajawat', 'Gennady Shvets'], 'published': '2025-08-01 15:40:33+00:00', 'categories': ['physics.plasm-ph'], 'url': 'http://arxiv.org/abs/2508.00720v2'}
|
29
|
2508.00747
| 2,508.00747
| null |
Zero Probability of the Cut Locus of a Fréchet Mean on a Riemannian Manifold
|
We show that the cut locus of a Fr\'echet mean of a random variable on a
connected and complete Riemanian manifold has zero probability, a result known
previously in special cases and conjectured in general. In application, we rule
out stickiness, while providing examples of nowhere smooth Fr\'echet functions
and we discuss extensions of the statement to Fr\'echet $p$-means, for $p2$, as well as to noncomplete manifolds and more general metric spaces.
|
2025-08-01
|
{'id': '2508.00747', 'title': 'Zero Probability of the Cut Locus of a Fréchet Mean on a Riemannian Manifold', 'abstract': "We show that the cut locus of a Fr\\'echet mean of a random variable on a\nconnected and complete Riemanian manifold has zero probability, a result known\npreviously in special cases and conjectured in general. In application, we rule\nout stickiness, while providing examples of nowhere smooth Fr\\'echet functions\nand we discuss extensions of the statement to Fr\\'echet $p$-means, for $p\\neq\n2$, as well as to noncomplete manifolds and more general metric spaces.", 'authors': ['Alexander Lytchak', 'Stephan F. Huckemann'], 'published': '2025-08-01 16:21:28+00:00', 'categories': ['math.PR', 'math.DG', '60D05, 53C20'], 'url': 'http://arxiv.org/abs/2508.00747v1'}
|
30
|
2508.00523
| 2,508.00523
| null |
Online Nonsubmodular Optimization with Delayed Feedback in the Bandit Setting
|
We investigate the online nonsubmodular optimization with delayed feedback in
the bandit setting, where the loss function is $$-weakly DR-submodular
and $$-weakly DR-supermodular. Previous work has established an
$(,)$-regret bound of $O(nd^1/3T^2/3)$, where $n$ is
the dimensionality and $d$ is the maximum delay. However, its regret bound
relies on the maximum delay and is thus sensitive to irregular delays.
Additionally, it couples the effects of delays and bandit feedback as its bound
is the product of the delay term and the $O(nT^2/3)$ regret bound
in the bandit setting without delayed feedback. In this paper, we develop two
algorithms to address these limitations, respectively. Firstly, we propose a
novel method, namely DBGD-NF, which employs the one-point gradient estimator
and utilizes all the available estimated gradients in each round to update the
decision. It achieves a better $O(nd^1/3T^2/3)$ regret
bound, which is relevant to the average delay $d =
1T_t=1^T d_td$. Secondly, we extend DBGD-NF by employing a
blocking update mechanism to decouple the joint effect of the delays and bandit
feedback, which enjoys an $O(n(T^2/3 + dT))$ regret bound.
When $d = O(T^1/3)$, our regret bound matches the
$O(nT^2/3)$ bound in the bandit setting without delayed feedback.
Compared to our first $O(nd^1/3T^2/3)$ bound, it is more
advantageous when the maximum delay $d = o(d^2/3T^1/3)$. Finally, we
conduct experiments on structured sparse learning to demonstrate the
superiority of our methods.
|
2025-08-01
|
{'id': '2508.00523', 'title': 'Online Nonsubmodular Optimization with Delayed Feedback in the Bandit Setting', 'abstract': 'We investigate the online nonsubmodular optimization with delayed feedback in\nthe bandit setting, where the loss function is $\\alpha$-weakly DR-submodular\nand $\\beta$-weakly DR-supermodular. Previous work has established an\n$(\\alpha,\\beta)$-regret bound of $\\mathcal{O}(nd^{1/3}T^{2/3})$, where $n$ is\nthe dimensionality and $d$ is the maximum delay. However, its regret bound\nrelies on the maximum delay and is thus sensitive to irregular delays.\nAdditionally, it couples the effects of delays and bandit feedback as its bound\nis the product of the delay term and the $\\mathcal{O}(nT^{2/3})$ regret bound\nin the bandit setting without delayed feedback. In this paper, we develop two\nalgorithms to address these limitations, respectively. Firstly, we propose a\nnovel method, namely DBGD-NF, which employs the one-point gradient estimator\nand utilizes all the available estimated gradients in each round to update the\ndecision. It achieves a better $\\mathcal{O}(n\\bar{d}^{1/3}T^{2/3})$ regret\nbound, which is relevant to the average delay $\\bar{d} =\n\\frac{1}{T}\\sum_{t=1}^T d_t\\leq d$. Secondly, we extend DBGD-NF by employing a\nblocking update mechanism to decouple the joint effect of the delays and bandit\nfeedback, which enjoys an $\\mathcal{O}(n(T^{2/3} + \\sqrt{dT}))$ regret bound.\nWhen $d = \\mathcal{O}(T^{1/3})$, our regret bound matches the\n$\\mathcal{O}(nT^{2/3})$ bound in the bandit setting without delayed feedback.\nCompared to our first $\\mathcal{O}(n\\bar{d}^{1/3}T^{2/3})$ bound, it is more\nadvantageous when the maximum delay $d = o(\\bar{d}^{2/3}T^{1/3})$. Finally, we\nconduct experiments on structured sparse learning to demonstrate the\nsuperiority of our methods.', 'authors': ['Sifan Yang', 'Yuanyu Wan', 'Lijun Zhang'], 'published': '2025-08-01 11:00:19+00:00', 'categories': ['cs.LG'], 'url': 'http://arxiv.org/abs/2508.00523v1'}
|
31
|
2508.00701
| 2,508.00701
| null |
D3: Training-Free AI-Generated Video Detection Using Second-Order Features
|
The evolution of video generation techniques, such as Sora, has made it
increasingly easy to produce high-fidelity AI-generated videos, raising public
concern over the dissemination of synthetic content. However, existing
detection methodologies remain limited by their insufficient exploration of
temporal artifacts in synthetic videos. To bridge this gap, we establish a
theoretical framework through second-order dynamical analysis under Newtonian
mechanics, subsequently extending the Second-order Central Difference features
tailored for temporal artifact detection. Building on this theoretical
foundation, we reveal a fundamental divergence in second-order feature
distributions between real and AI-generated videos. Concretely, we propose
Detection by Difference of Differences (D3), a novel training-free detection
method that leverages the above second-order temporal discrepancies. We
validate the superiority of our D3 on 4 open-source datasets (Gen-Video,
VideoPhy, EvalCrafter, VidProM), 40 subsets in total. For example, on GenVideo,
D3 outperforms the previous best method by 10.39% (absolute) mean Average
Precision. Additional experiments on time cost and post-processing operations
demonstrate D3's exceptional computational efficiency and strong robust
performance. Our code is available at https://github.com/Zig-HS/D3.
|
2025-08-01
|
{'id': '2508.00701', 'title': 'D3: Training-Free AI-Generated Video Detection Using Second-Order Features', 'abstract': "The evolution of video generation techniques, such as Sora, has made it\nincreasingly easy to produce high-fidelity AI-generated videos, raising public\nconcern over the dissemination of synthetic content. However, existing\ndetection methodologies remain limited by their insufficient exploration of\ntemporal artifacts in synthetic videos. To bridge this gap, we establish a\ntheoretical framework through second-order dynamical analysis under Newtonian\nmechanics, subsequently extending the Second-order Central Difference features\ntailored for temporal artifact detection. Building on this theoretical\nfoundation, we reveal a fundamental divergence in second-order feature\ndistributions between real and AI-generated videos. Concretely, we propose\nDetection by Difference of Differences (D3), a novel training-free detection\nmethod that leverages the above second-order temporal discrepancies. We\nvalidate the superiority of our D3 on 4 open-source datasets (Gen-Video,\nVideoPhy, EvalCrafter, VidProM), 40 subsets in total. For example, on GenVideo,\nD3 outperforms the previous best method by 10.39% (absolute) mean Average\nPrecision. Additional experiments on time cost and post-processing operations\ndemonstrate D3's exceptional computational efficiency and strong robust\nperformance. Our code is available at https://github.com/Zig-HS/D3.", 'authors': ['Chende Zheng', 'Ruiqi suo', 'Chenhao Lin', 'Zhengyu Zhao', 'Le Yang', 'Shuai Liu', 'Minghui Yang', 'Cong Wang', 'Chao Shen'], 'published': '2025-08-01 15:17:51+00:00', 'categories': ['cs.CV', 'cs.AI'], 'url': 'http://arxiv.org/abs/2508.00701v2'}
|
32
|
2508.00345
| 2,508.00345
| null |
Concentration and Markups in International Trade
|
This paper derives a closed-form expression linking aggregate markups on
imported inputs to concentration in a model of firm-to-firm trade with
two-sided market power. Our theory extends standard oligopoly insights in two
dimensions. First, it reveals that markups increase with exporter concentration
and decrease with importer concentration, reflecting the balance of oligopoly
and oligopsony forces. Second, it adapts conventional market definitions to
reflect rigid trading relationships, yielding new concentration measures that
capture competition in firm-to-firm trade. Analysis of Colombian
transaction-level import data shows these differences are key to understanding
markup dynamics in international trade.
|
2025-08-01
|
{'id': '2508.00345', 'title': 'Concentration and Markups in International Trade', 'abstract': 'This paper derives a closed-form expression linking aggregate markups on\nimported inputs to concentration in a model of firm-to-firm trade with\ntwo-sided market power. Our theory extends standard oligopoly insights in two\ndimensions. First, it reveals that markups increase with exporter concentration\nand decrease with importer concentration, reflecting the balance of oligopoly\nand oligopsony forces. Second, it adapts conventional market definitions to\nreflect rigid trading relationships, yielding new concentration measures that\ncapture competition in firm-to-firm trade. Analysis of Colombian\ntransaction-level import data shows these differences are key to understanding\nmarkup dynamics in international trade.', 'authors': ['Alviarez Vanessa', 'Fioretti Michele', 'Kikkawa Ken', 'Morlacco Monica'], 'published': '2025-08-01 06:18:24+00:00', 'categories': ['econ.TH'], 'url': 'http://arxiv.org/abs/2508.00345v1'}
|
33
|
2508.00261
| 2,508.00261
| null |
Energy Efficient Trajectory Control and Resource Allocation in Multi-UAV-assisted MEC via Deep Reinforcement Learning
|
Mobile edge computing (MEC) is a promising technique to improve the
computational capacity of smart devices (SDs) in Internet of Things (IoT).
However, the performance of MEC is restricted due to its fixed location and
limited service scope. Hence, we investigate an unmanned aerial vehicle
(UAV)-assisted MEC system, where multiple UAVs are dispatched and each UAV can
simultaneously provide computing service for multiple SDs. To improve the
performance of system, we formulated a UAV-based trajectory control and
resource allocation multi-objective optimization problem (TCRAMOP) to
simultaneously maximize the offloading number of UAVs and minimize total
offloading delay and total energy consumption of UAVs by optimizing the flight
paths of UAVs as well as the computing resource allocated to served SDs. Then,
consider that the solution of TCRAMOP requires continuous decision-making and
the system is dynamic, we propose an enhanced deep reinforcement learning (DRL)
algorithm, namely, distributed proximal policy optimization with imitation
learning (DPPOIL). This algorithm incorporates the generative adversarial
imitation learning technique to improve the policy performance. Simulation
results demonstrate the effectiveness of our proposed DPPOIL and prove that the
learned strategy of DPPOIL is better compared with other baseline methods.
|
2025-08-01
|
{'id': '2508.00261', 'title': 'Energy Efficient Trajectory Control and Resource Allocation in Multi-UAV-assisted MEC via Deep Reinforcement Learning', 'abstract': 'Mobile edge computing (MEC) is a promising technique to improve the\ncomputational capacity of smart devices (SDs) in Internet of Things (IoT).\nHowever, the performance of MEC is restricted due to its fixed location and\nlimited service scope. Hence, we investigate an unmanned aerial vehicle\n(UAV)-assisted MEC system, where multiple UAVs are dispatched and each UAV can\nsimultaneously provide computing service for multiple SDs. To improve the\nperformance of system, we formulated a UAV-based trajectory control and\nresource allocation multi-objective optimization problem (TCRAMOP) to\nsimultaneously maximize the offloading number of UAVs and minimize total\noffloading delay and total energy consumption of UAVs by optimizing the flight\npaths of UAVs as well as the computing resource allocated to served SDs. Then,\nconsider that the solution of TCRAMOP requires continuous decision-making and\nthe system is dynamic, we propose an enhanced deep reinforcement learning (DRL)\nalgorithm, namely, distributed proximal policy optimization with imitation\nlearning (DPPOIL). This algorithm incorporates the generative adversarial\nimitation learning technique to improve the policy performance. Simulation\nresults demonstrate the effectiveness of our proposed DPPOIL and prove that the\nlearned strategy of DPPOIL is better compared with other baseline methods.', 'authors': ['Saichao Liu', 'Geng Sun', 'Chuang Zhang', 'Xuejie Liu', 'Jiacheng Wang', 'Changyuan Zhao', 'Dusit Niyato'], 'published': '2025-08-01 02:09:40+00:00', 'categories': ['cs.NI', 'eess.SP'], 'url': 'http://arxiv.org/abs/2508.00261v1'}
|
1,543
|
2501.01195
| 2,501.01195
| null |
Data Augmentation Techniques for Chinese Disease Name Normalization
|
Disease name normalization is an important task in the medical domain. It
classifies disease names written in various formats into standardized names,
serving as a fundamental component in smart healthcare systems for various
disease-related functions. Nevertheless, the most significant obstacle to
existing disease name normalization systems is the severe shortage of training
data. Consequently, we present a novel data augmentation approach that includes
a series of data augmentation techniques and some supporting modules to help
mitigate the problem. Through extensive experimentation, we illustrate that our
proposed approach exhibits significant performance improvements across various
baseline models and training objectives, particularly in scenarios with limited
training data
|
2025-01-02
|
{'id': '2501.01195', 'title': 'Data Augmentation Techniques for Chinese Disease Name Normalization', 'abstract': 'Disease name normalization is an important task in the medical domain. It\nclassifies disease names written in various formats into standardized names,\nserving as a fundamental component in smart healthcare systems for various\ndisease-related functions. Nevertheless, the most significant obstacle to\nexisting disease name normalization systems is the severe shortage of training\ndata. Consequently, we present a novel data augmentation approach that includes\na series of data augmentation techniques and some supporting modules to help\nmitigate the problem. Through extensive experimentation, we illustrate that our\nproposed approach exhibits significant performance improvements across various\nbaseline models and training objectives, particularly in scenarios with limited\ntraining data', 'authors': ['Wenqian Cui', 'Xiangling Fu', 'Shaohui Liu', 'Mingjun Gu', 'Xien Liu', 'Ji Wu', 'Irwin King'], 'published': '2025-01-02 11:12:03+00:00', 'categories': ['cs.CL', 'cs.AI'], 'url': 'http://arxiv.org/abs/2501.01195v1'}
|
34
|
2508.00516
| 2,508.00516
| null |
Probing scalar field with generic extreme mass-ratio inspirals around Kerr black holes
|
The future space-based gravitational wave observatories are expected to
provide unprecedented opportunities to explore intricate characteristics of
black hole binaries, particularly for extreme mass-ratio inspirals (EMRIs), in
which a stellar-mass compact object slowly inspirals into a supermassive black
hole. These systems are very prominent sources for testing gravity in the
strong gravity fields and for probing potential deviations from general
relativity, including those arising from the presence of fundamental scalar
fields. In this work, we examine the impact of a scalar charge carried by the
inspiraling object within the context of EMRIs. We focus on generic orbits that
present both eccentricity and inclination to evaluate how these parameters
affect the modifications induced by the scalar charge to the gravitational wave
signal. Our results demonstrate that the inclusion of orbital inclination, in
particular, enhances the detectability of scalar field effects by introducing
richer waveform features that deviate from the purely general relativistic
case. The interplay among scalar charge, eccentricity and inclination provides
a more complete sampling of the black hole spacetime, suggesting that EMRIs
with such generic orbits represent compelling systems for stringently
constraining or discovering new fundamental fields through future gravitational
wave observations.
|
2025-08-01
|
{'id': '2508.00516', 'title': 'Probing scalar field with generic extreme mass-ratio inspirals around Kerr black holes', 'abstract': 'The future space-based gravitational wave observatories are expected to\nprovide unprecedented opportunities to explore intricate characteristics of\nblack hole binaries, particularly for extreme mass-ratio inspirals (EMRIs), in\nwhich a stellar-mass compact object slowly inspirals into a supermassive black\nhole. These systems are very prominent sources for testing gravity in the\nstrong gravity fields and for probing potential deviations from general\nrelativity, including those arising from the presence of fundamental scalar\nfields. In this work, we examine the impact of a scalar charge carried by the\ninspiraling object within the context of EMRIs. We focus on generic orbits that\npresent both eccentricity and inclination to evaluate how these parameters\naffect the modifications induced by the scalar charge to the gravitational wave\nsignal. Our results demonstrate that the inclusion of orbital inclination, in\nparticular, enhances the detectability of scalar field effects by introducing\nricher waveform features that deviate from the purely general relativistic\ncase. The interplay among scalar charge, eccentricity and inclination provides\na more complete sampling of the black hole spacetime, suggesting that EMRIs\nwith such generic orbits represent compelling systems for stringently\nconstraining or discovering new fundamental fields through future gravitational\nwave observations.', 'authors': ['Tieguang Zi', 'Shailesh Kumar'], 'published': '2025-08-01 10:49:56+00:00', 'categories': ['gr-qc', 'astro-ph.HE', 'hep-th'], 'url': 'http://arxiv.org/abs/2508.00516v1'}
|
35
|
2508.00583
| 2,508.00583
| null |
Enhancing Wireless Networks for IoT with Large Vision Models: Foundations and Applications
|
Large vision models (LVMs) have emerged as a foundational paradigm in visual
intelligence, achieving state-of-the-art performance across diverse visual
tasks. Recent advances in LVMs have facilitated their integration into Internet
of Things (IoT) scenarios, offering superior generalization and adaptability
for vision-assisted network optimization. In this paper, we first investigate
the functionalities and core architectures of LVMs, highlighting their
capabilities across classification, segmentation, generation, and multimodal
visual processing. We then explore a variety of LVM applications in wireless
communications, covering representative tasks across the physical layer,
network layer, and application layer. Furthermore, given the substantial model
size of LVMs and the challenges of model retraining in wireless domains, we
propose a progressive fine-tuning framework that incrementally adapts
pretrained LVMs for joint optimization of multiple IoT tasks. A case study in
low-altitude economy networks (LAENets) demonstrates the effectiveness of the
proposed framework over conventional CNNs in joint beamforming and positioning
tasks for Internet of drones, underscoring a promising direction for
integrating LVMs into intelligent wireless systems.
|
2025-08-01
|
{'id': '2508.00583', 'title': 'Enhancing Wireless Networks for IoT with Large Vision Models: Foundations and Applications', 'abstract': 'Large vision models (LVMs) have emerged as a foundational paradigm in visual\nintelligence, achieving state-of-the-art performance across diverse visual\ntasks. Recent advances in LVMs have facilitated their integration into Internet\nof Things (IoT) scenarios, offering superior generalization and adaptability\nfor vision-assisted network optimization. In this paper, we first investigate\nthe functionalities and core architectures of LVMs, highlighting their\ncapabilities across classification, segmentation, generation, and multimodal\nvisual processing. We then explore a variety of LVM applications in wireless\ncommunications, covering representative tasks across the physical layer,\nnetwork layer, and application layer. Furthermore, given the substantial model\nsize of LVMs and the challenges of model retraining in wireless domains, we\npropose a progressive fine-tuning framework that incrementally adapts\npretrained LVMs for joint optimization of multiple IoT tasks. A case study in\nlow-altitude economy networks (LAENets) demonstrates the effectiveness of the\nproposed framework over conventional CNNs in joint beamforming and positioning\ntasks for Internet of drones, underscoring a promising direction for\nintegrating LVMs into intelligent wireless systems.', 'authors': ['Yunting Xu', 'Jiacheng Wang', 'Ruichen Zhang', 'Dusit Niyato', 'Deepu Rajan', 'Liang Yu', 'Haibo Zhou', 'Abbas Jamalipour', 'Xianbin Wang'], 'published': '2025-08-01 12:25:51+00:00', 'categories': ['cs.NI'], 'url': 'http://arxiv.org/abs/2508.00583v1'}
|
36
|
2508.00696
| 2,508.00696
| null |
Online Rolling Controlled Sequential Monte Carlo
|
We introduce methodology for real-time inference in general-state-space
hidden Markov models. Specifically, we extend recent advances in controlled
sequential Monte Carlo (CSMC) methods-originally proposed for offline
smoothing-to the online setting via a rolling window mechanism. Our novel
online rolling controlled sequential Monte Carlo (ORCSMC) algorithm employs two
particle systems to simultaneously estimate twisting functions and perform
filtering, ensuring real-time adaptivity to new observations while maintaining
bounded computational cost. Numerical results on linear-Gaussian, stochastic
volatility, and neuroscience models demonstrate improved estimation accuracy
and robustness in higher dimensions, compared to standard particle filtering
approaches. The method offers a statistically efficient and practical solution
for sequential and real-time inference in complex latent variable models.
|
2025-08-01
|
{'id': '2508.00696', 'title': 'Online Rolling Controlled Sequential Monte Carlo', 'abstract': 'We introduce methodology for real-time inference in general-state-space\nhidden Markov models. Specifically, we extend recent advances in controlled\nsequential Monte Carlo (CSMC) methods-originally proposed for offline\nsmoothing-to the online setting via a rolling window mechanism. Our novel\nonline rolling controlled sequential Monte Carlo (ORCSMC) algorithm employs two\nparticle systems to simultaneously estimate twisting functions and perform\nfiltering, ensuring real-time adaptivity to new observations while maintaining\nbounded computational cost. Numerical results on linear-Gaussian, stochastic\nvolatility, and neuroscience models demonstrate improved estimation accuracy\nand robustness in higher dimensions, compared to standard particle filtering\napproaches. The method offers a statistically efficient and practical solution\nfor sequential and real-time inference in complex latent variable models.', 'authors': ['Liwen Xue', 'Axel Finke', 'Adam M. Johansen'], 'published': '2025-08-01 15:12:58+00:00', 'categories': ['stat.CO'], 'url': 'http://arxiv.org/abs/2508.00696v1'}
|
37
|
2508.00380
| 2,508.0038
| null |
Evolutionary Generative Optimization: Towards Fully Data-Driven Evolutionary Optimization via Generative Learning
|
Recent advances in data-driven evolutionary algorithms (EAs) have
demonstrated the potential of leveraging data to improve optimization accuracy
and adaptability. Nevertheless, most existing approaches remain dependent on
handcrafted heuristics, which limits their generality and automation. To
address this challenge, we propose Evolutionary Generative Optimization
(EvoGO), a fully data-driven framework empowered by generative learning. EvoGO
streamlines the evolutionary optimization process into three stages: data
preparation, model training, and population generation. The data preparation
stage constructs a pairwise dataset to enrich training diversity without
incurring additional evaluation costs. During model training, a tailored
generative model learns to transform inferior solutions into superior ones. In
the population generation stage, EvoGO replaces traditional reproduction
operators with a scalable and parallelizable generative mechanism. Extensive
experiments on numerical benchmarks, classical control problems, and
high-dimensional robotic tasks demonstrate that EvoGO consistently converges
within merely 10 generations and significantly outperforms a wide spectrum of
optimization approaches, including traditional EAs, Bayesian optimization, and
reinforcement learning based methods. Source code will be made publicly
available.
|
2025-08-01
|
{'id': '2508.00380', 'title': 'Evolutionary Generative Optimization: Towards Fully Data-Driven Evolutionary Optimization via Generative Learning', 'abstract': 'Recent advances in data-driven evolutionary algorithms (EAs) have\ndemonstrated the potential of leveraging data to improve optimization accuracy\nand adaptability. Nevertheless, most existing approaches remain dependent on\nhandcrafted heuristics, which limits their generality and automation. To\naddress this challenge, we propose Evolutionary Generative Optimization\n(EvoGO), a fully data-driven framework empowered by generative learning. EvoGO\nstreamlines the evolutionary optimization process into three stages: data\npreparation, model training, and population generation. The data preparation\nstage constructs a pairwise dataset to enrich training diversity without\nincurring additional evaluation costs. During model training, a tailored\ngenerative model learns to transform inferior solutions into superior ones. In\nthe population generation stage, EvoGO replaces traditional reproduction\noperators with a scalable and parallelizable generative mechanism. Extensive\nexperiments on numerical benchmarks, classical control problems, and\nhigh-dimensional robotic tasks demonstrate that EvoGO consistently converges\nwithin merely 10 generations and significantly outperforms a wide spectrum of\noptimization approaches, including traditional EAs, Bayesian optimization, and\nreinforcement learning based methods. Source code will be made publicly\navailable.', 'authors': ['Kebin Sun', 'Tao Jiang', 'Ran Cheng', 'Yaochu Jin', 'Kay Chen Tan'], 'published': '2025-08-01 07:17:57+00:00', 'categories': ['cs.NE'], 'url': 'http://arxiv.org/abs/2508.00380v1'}
|
38
|
2508.00331
| 2,508.00331
| null |
Embryology of a Language Model
|
Understanding how language models develop their internal computational
structure is a central problem in the science of deep learning. While
susceptibilities, drawn from statistical physics, offer a promising analytical
tool, their full potential for visualizing network organization remains
untapped. In this work, we introduce an embryological approach, applying UMAP
to the susceptibility matrix to visualize the model's structural development
over training. Our visualizations reveal the emergence of a clear ``body
plan,'' charting the formation of known features like the induction circuit and
discovering previously unknown structures, such as a ``spacing fin'' dedicated
to counting space tokens. This work demonstrates that susceptibility analysis
can move beyond validation to uncover novel mechanisms, providing a powerful,
holistic lens for studying the developmental principles of complex neural
networks.
|
2025-08-01
|
{'id': '2508.00331', 'title': 'Embryology of a Language Model', 'abstract': "Understanding how language models develop their internal computational\nstructure is a central problem in the science of deep learning. While\nsusceptibilities, drawn from statistical physics, offer a promising analytical\ntool, their full potential for visualizing network organization remains\nuntapped. In this work, we introduce an embryological approach, applying UMAP\nto the susceptibility matrix to visualize the model's structural development\nover training. Our visualizations reveal the emergence of a clear ``body\nplan,'' charting the formation of known features like the induction circuit and\ndiscovering previously unknown structures, such as a ``spacing fin'' dedicated\nto counting space tokens. This work demonstrates that susceptibility analysis\ncan move beyond validation to uncover novel mechanisms, providing a powerful,\nholistic lens for studying the developmental principles of complex neural\nnetworks.", 'authors': ['George Wang', 'Garrett Baker', 'Andrew Gordon', 'Daniel Murfet'], 'published': '2025-08-01 05:39:41+00:00', 'categories': ['cs.LG'], 'url': 'http://arxiv.org/abs/2508.00331v1'}
|
39
|
2508.00458
| 2,508.00458
| null |
LO-Aware Adaptive Modulation for Rydberg Atomic Receivers
|
Rydberg atomic (RA) receivers represent a revolutionary quantum technology
for wireless communications, offering unprecedented sensitivity beyond
conventional radio frequency (RF) antennas. However, these receivers detect
only signal amplitude, losing critical phase information. While reference
signals generated by a local oscillator (LO) can assist in phase recovery,
existing modulation schemes designed for conventional systems perform poorly
with this quantum detection mechanism. This paper introduces a breakthrough
LO-aware adaptive modulation (LOAM) scheme specifically developed for RA
receivers that dynamically adapts to complex fading channel coefficients. LOAM
maximizes the minimum amplitude difference between constellation points,
ensuring optimal detection performance. The innovation employs an adaptive
co-linear constellation architecture aligned with the combined phase of
reference signal and channel coefficient. For strong reference signals, LOAM
generates symmetric constellation points centered at origin; for weak signals,
it adopts non-symmetric distributions. The paper mathematically derives the
threshold governing these operational regimes. Simulation results reveal the
transformative impact of LOAM, demonstrating performance gains exceeding 45 dB
over conventional modulation schemes, including quadrature amplitude modulation
(QAM), phase-shift keying (PSK), and pulse-amplitude modulation (PAM).
|
2025-08-01
|
{'id': '2508.00458', 'title': 'LO-Aware Adaptive Modulation for Rydberg Atomic Receivers', 'abstract': 'Rydberg atomic (RA) receivers represent a revolutionary quantum technology\nfor wireless communications, offering unprecedented sensitivity beyond\nconventional radio frequency (RF) antennas. However, these receivers detect\nonly signal amplitude, losing critical phase information. While reference\nsignals generated by a local oscillator (LO) can assist in phase recovery,\nexisting modulation schemes designed for conventional systems perform poorly\nwith this quantum detection mechanism. This paper introduces a breakthrough\nLO-aware adaptive modulation (LOAM) scheme specifically developed for RA\nreceivers that dynamically adapts to complex fading channel coefficients. LOAM\nmaximizes the minimum amplitude difference between constellation points,\nensuring optimal detection performance. The innovation employs an adaptive\nco-linear constellation architecture aligned with the combined phase of\nreference signal and channel coefficient. For strong reference signals, LOAM\ngenerates symmetric constellation points centered at origin; for weak signals,\nit adopts non-symmetric distributions. The paper mathematically derives the\nthreshold governing these operational regimes. Simulation results reveal the\ntransformative impact of LOAM, demonstrating performance gains exceeding 45 dB\nover conventional modulation schemes, including quadrature amplitude modulation\n(QAM), phase-shift keying (PSK), and pulse-amplitude modulation (PAM).', 'authors': ['Jiuyu Liu', 'Yi Ma', 'Rahim Tafazolli'], 'published': '2025-08-01 09:31:33+00:00', 'categories': ['cs.IT', 'eess.SP', 'math.IT'], 'url': 'http://arxiv.org/abs/2508.00458v1'}
|
40
|
2508.00519
| 2,508.00519
| null |
Localized states and skin effect around non-Hermitian impurities in tight-binding models
|
We use the generalized Bloch theorem formalism of Alase et al. [Phys. Rev. Lett. 117 076804 (2016)] to analyze simple one-dimensional
tight-binding lattice systems connected by Hermitian bonds (all with the same
hopping parameter $t$), but containing one bond impurity which can be either
Hermitian or non-Hermitian. We calculate the band structure, the bulk-boundary
correspondence indicator ($D_L()$) and analyze the eigenvalues of the
lattice translation operator ($z$), for each eigenstate. From the $z$ values
the generalized Brillouin zone can be reconstructed. If the impurity is
Hermitian (and $PT$-symmetric), we find a parameter regime in which
two localized edge states separate from the tight-binding band. We then
simulate a non-Hermitian impurity by keeping hopping in one direction of the
bond impurity the same as the rest of the tight-binding system, and varying
only its reciprocal. Again, we find a region with localized edge states, but in
this case the energy eigenvalues are purely imaginary. We also find that in
this case the two zero energy eigenvectors coalesce, hence this system is an
exceptional line. We then perform an interpolative scan between the above two
scenarios and find that there is an intermediate region exhibiting a
non-Hermitian skin effect. In this region a macroscopic fraction of states
acquire complex energy eigenvalues and exhibit localization towards the
impurity. Our numerical results are supported by a detailed analysis of the
solutions of the boundary/impurity equation.
|
2025-08-01
|
{'id': '2508.00519', 'title': 'Localized states and skin effect around non-Hermitian impurities in tight-binding models', 'abstract': 'We use the generalized Bloch theorem formalism of Alase {\\it et al.} [{\\it\nPhys. Rev. Lett.} {\\bf 117} 076804 (2016)] to analyze simple one-dimensional\ntight-binding lattice systems connected by Hermitian bonds (all with the same\nhopping parameter $t$), but containing one bond impurity which can be either\nHermitian or non-Hermitian. We calculate the band structure, the bulk-boundary\ncorrespondence indicator ($D_L(\\epsilon)$) and analyze the eigenvalues of the\nlattice translation operator ($z$), for each eigenstate. From the $z$ values\nthe generalized Brillouin zone can be reconstructed. If the impurity is\nHermitian (and $\\mathcal{PT}$-symmetric), we find a parameter regime in which\ntwo localized edge states separate from the tight-binding band. We then\nsimulate a non-Hermitian impurity by keeping hopping in one direction of the\nbond impurity the same as the rest of the tight-binding system, and varying\nonly its reciprocal. Again, we find a region with localized edge states, but in\nthis case the energy eigenvalues are purely imaginary. We also find that in\nthis case the two zero energy eigenvectors coalesce, hence this system is an\nexceptional line. We then perform an interpolative scan between the above two\nscenarios and find that there is an intermediate region exhibiting a\nnon-Hermitian skin effect. In this region a macroscopic fraction of states\nacquire complex energy eigenvalues and exhibit localization towards the\nimpurity. Our numerical results are supported by a detailed analysis of the\nsolutions of the boundary/impurity equation.', 'authors': ['Balázs Hetényi', 'Balázs Dóra'], 'published': '2025-08-01 10:53:40+00:00', 'categories': ['cond-mat.mes-hall', 'cond-mat.quant-gas', 'cond-mat.supr-con', 'math-ph', 'math.MP', 'quant-ph'], 'url': 'http://arxiv.org/abs/2508.00519v1'}
|
41
|
2508.00646
| 2,508.00646
| null |
Pull Requests From The Classroom: Co-Developing Curriculum And Code
|
Educational technologies often misalign with instructors' pedagogical goals,
forcing adaptations that compromise teaching efficacy. In this paper, we
present a case study on the co-development of curriculum and technology in the
context of a university course on scientific writing. Specifically, we examine
how a custom-built peer feedback system was iteratively developed alongside the
course to support annotation, feedback exchange, and revision. Results show
that while co-development fostered stronger alignment between software features
and course goals, it also exposed usability limitations and
infrastructure-related frustrations, emphasizing the need for closer
coordination between teaching and technical teams.
|
2025-08-01
|
{'id': '2508.00646', 'title': 'Pull Requests From The Classroom: Co-Developing Curriculum And Code', 'abstract': "Educational technologies often misalign with instructors' pedagogical goals,\nforcing adaptations that compromise teaching efficacy. In this paper, we\npresent a case study on the co-development of curriculum and technology in the\ncontext of a university course on scientific writing. Specifically, we examine\nhow a custom-built peer feedback system was iteratively developed alongside the\ncourse to support annotation, feedback exchange, and revision. Results show\nthat while co-development fostered stronger alignment between software features\nand course goals, it also exposed usability limitations and\ninfrastructure-related frustrations, emphasizing the need for closer\ncoordination between teaching and technical teams.", 'authors': ['Dennis Zyska', 'Ilia Kuznetsov', 'Florian Müller', 'Iryna Gurevych'], 'published': '2025-08-01 14:04:09+00:00', 'categories': ['cs.HC'], 'url': 'http://arxiv.org/abs/2508.00646v1'}
|
42
|
2508.00738
| 2,508.00738
| null |
Tool-Assisted Conformance Checking to Reference Process Models
|
Reference models convey best practices and standards. The reference
frameworks necessitate conformance checks to ensure adherence to established
guidelines and principles, which is crucial for maintaining quality and
consistency in various processes. This paper explores automated conformance
checks for concrete process models against reference models using causal
dependency analysis of tasks and events. Existing notions of conformance
checking for process models focus on verifying process execution traces and
lack the expressiveness and automation needed for semantic model comparison,
leaving this question unresolved. We integrate our approach into a broader
semantic framework for defining reference model conformance. We outline an
algorithm for reference process model conformance checking, evaluate it through
a case study, and discuss its strengths and limitations. Our research provides
a tool-assisted solution enhancing accuracy and flexibility in process model
conformance verification.
|
2025-08-01
|
{'id': '2508.00738', 'title': 'Tool-Assisted Conformance Checking to Reference Process Models', 'abstract': 'Reference models convey best practices and standards. The reference\nframeworks necessitate conformance checks to ensure adherence to established\nguidelines and principles, which is crucial for maintaining quality and\nconsistency in various processes. This paper explores automated conformance\nchecks for concrete process models against reference models using causal\ndependency analysis of tasks and events. Existing notions of conformance\nchecking for process models focus on verifying process execution traces and\nlack the expressiveness and automation needed for semantic model comparison,\nleaving this question unresolved. We integrate our approach into a broader\nsemantic framework for defining reference model conformance. We outline an\nalgorithm for reference process model conformance checking, evaluate it through\na case study, and discuss its strengths and limitations. Our research provides\na tool-assisted solution enhancing accuracy and flexibility in process model\nconformance verification.', 'authors': ['Bernhard Rumpe', 'Max Stachon', 'Sebastian Stüber', 'Valdes Voufo'], 'published': '2025-08-01 16:08:24+00:00', 'categories': ['cs.SE', 'cs.FL', '68N30', 'D.2.4'], 'url': 'http://arxiv.org/abs/2508.00738v2'}
|
43
|
2508.00328
| 2,508.00328
| null |
From Patient Burdens to User Agency: Designing for Real-Time Protection Support in Online Health Consultations
|
Online medical consultation platforms, while convenient, are undermined by
significant privacy risks that erode user trust. We first conducted in-depth
semi-structured interviews with 12 users to understand their perceptions of
security and privacy landscapes on online medical consultation platforms, as
well as their practices, challenges and expectation. Our analysis reveals a
critical disconnect between users' desires for anonymity and control, and
platform realities that offload the responsibility of ``privacy labor''. To
bridge this gap, we present SafeShare, an interaction technique that leverages
localized LLM to redact consultations in real-time. SafeShare balances utility
and privacy through selectively anonymize private information. A technical
evaluation of SafeShare's core PII detection module on 3 dataset demonstrates
high efficacy, achieving 89.64\% accuracy with Qwen3-4B on IMCS21 dataset.
|
2025-08-01
|
{'id': '2508.00328', 'title': 'From Patient Burdens to User Agency: Designing for Real-Time Protection Support in Online Health Consultations', 'abstract': "Online medical consultation platforms, while convenient, are undermined by\nsignificant privacy risks that erode user trust. We first conducted in-depth\nsemi-structured interviews with 12 users to understand their perceptions of\nsecurity and privacy landscapes on online medical consultation platforms, as\nwell as their practices, challenges and expectation. Our analysis reveals a\ncritical disconnect between users' desires for anonymity and control, and\nplatform realities that offload the responsibility of ``privacy labor''. To\nbridge this gap, we present SafeShare, an interaction technique that leverages\nlocalized LLM to redact consultations in real-time. SafeShare balances utility\nand privacy through selectively anonymize private information. A technical\nevaluation of SafeShare's core PII detection module on 3 dataset demonstrates\nhigh efficacy, achieving 89.64\\% accuracy with Qwen3-4B on IMCS21 dataset.", 'authors': ['Shuning Zhang', 'Ying Ma', "Yongquan `Owen' Hu", 'Ting Dang', 'Hong Jia', 'Xin Yi', 'Hewu Li'], 'published': '2025-08-01 05:21:42+00:00', 'categories': ['cs.HC'], 'url': 'http://arxiv.org/abs/2508.00328v1'}
|
44
|
2508.00314
| 2,508.00314
| null |
Hybrid Scandium Aluminum Nitride/Silicon Nitride Integrated Photonic Circuits
|
Scandium-doped aluminum nitride has recently emerged as a promising material
for quantum photonic integrated circuits (PICs) due to its unique combination
of strong second-order nonlinearity, ferroelectricity, piezoelectricity, and
complementary metal-oxide-semiconductor (CMOS) compatibility. However, the
relatively high optical loss reported to date-typically above 2.4 dB/cm-remains
a key challenge that limits its widespread application in low-loss PICs. Here,
we present a monolithically integrated $Si_3N_4$-ScAlN
waveguide platform that overcomes this limitation. By confining light within an
etched $Si_3N_4$ waveguide while preserving the functional
properties of the underlying ScAlN layer, we achieve an intrinsic quality
factor of $Q_i = 3.35 10^5$, corresponding to a propagation
loss of 1.03 dB/cm-comparable to that of commercial single-mode
silicon-on-insulator (SOI) waveguides. This hybrid architecture enables
low-loss and scalable fabrication while retaining the advanced functionalities
offered by ScAlN, such as ferroelectricity and piezoelectricity. Our results
establish a new pathway for ScAlN-based PICs with potential applications in
high-speed optical communication, modulation, sensing, nonlinear optics, and
quantum optics within CMOS-compatible platforms.
|
2025-08-01
|
{'id': '2508.00314', 'title': 'Hybrid Scandium Aluminum Nitride/Silicon Nitride Integrated Photonic Circuits', 'abstract': 'Scandium-doped aluminum nitride has recently emerged as a promising material\nfor quantum photonic integrated circuits (PICs) due to its unique combination\nof strong second-order nonlinearity, ferroelectricity, piezoelectricity, and\ncomplementary metal-oxide-semiconductor (CMOS) compatibility. However, the\nrelatively high optical loss reported to date-typically above 2.4 dB/cm-remains\na key challenge that limits its widespread application in low-loss PICs. Here,\nwe present a monolithically integrated $\\mathrm{Si}_3\\mathrm{N}_4$-ScAlN\nwaveguide platform that overcomes this limitation. By confining light within an\netched $\\mathrm{Si}_3\\mathrm{N}_4$ waveguide while preserving the functional\nproperties of the underlying ScAlN layer, we achieve an intrinsic quality\nfactor of $Q_{\\mathrm{i}} = 3.35 \\times 10^5$, corresponding to a propagation\nloss of 1.03 dB/cm-comparable to that of commercial single-mode\nsilicon-on-insulator (SOI) waveguides. This hybrid architecture enables\nlow-loss and scalable fabrication while retaining the advanced functionalities\noffered by ScAlN, such as ferroelectricity and piezoelectricity. Our results\nestablish a new pathway for ScAlN-based PICs with potential applications in\nhigh-speed optical communication, modulation, sensing, nonlinear optics, and\nquantum optics within CMOS-compatible platforms.', 'authors': ['Jiangnan Liu', 'Shuai Liu', 'Abdur-Raheem Al-Hallak', 'Huabin Yu', 'Zhengwei Ye', 'Yuheng Zhang', 'Zheshen Zhang', 'Zetian Mi'], 'published': '2025-08-01 04:47:48+00:00', 'categories': ['physics.optics', 'cond-mat.mtrl-sci'], 'url': 'http://arxiv.org/abs/2508.00314v1'}
|
50
|
2508.00424
| 2,508.00424
| null |
CrossSet: Unveiling the Complex Interplay of Two Set-typed Dimensions in Multivariate Data
|
The interactive visual analysis of set-typed data, i.e., data with attributes
that are of type set, is a rewarding area of research and applications.
Valuable prior work has contributed solutions that enable the study of such
data with individual set-typed dimensions. In this paper, we present CrossSet,
a novel method for the joint study of two set-typed dimensions and their
interplay. Based on a task analysis, we describe a new, multi-scale approach to
the interactive visual exploration and analysis of such data. Two set-typed
data dimensions are jointly visualized using a hierarchical matrix layout,
enabling the analysis of the interactions between two set-typed attributes at
several levels, in addition to the analysis of individual such dimensions.
CrossSet is anchored at a compact, large-scale overview that is complemented by
drill-down opportunities to study the relations between and within the
set-typed dimensions, enabling an interactive visual multi-scale exploration
and analysis of bivariate set-typed data. Such an interactive approach makes it
possible to study single set-typed dimensions in detail, to gain an overview of
the interaction and association between two such dimensions, to refine one of
the dimensions to gain additional details at several levels, and to drill down
to the specific interactions of individual set-elements from the set-typed
dimensions. To demonstrate the effectiveness and efficiency of CrossSet, we
have evaluated the new method in the context of several application scenarios.
|
2025-08-01
|
{'id': '2508.00424', 'title': 'CrossSet: Unveiling the Complex Interplay of Two Set-typed Dimensions in Multivariate Data', 'abstract': 'The interactive visual analysis of set-typed data, i.e., data with attributes\nthat are of type set, is a rewarding area of research and applications.\nValuable prior work has contributed solutions that enable the study of such\ndata with individual set-typed dimensions. In this paper, we present CrossSet,\na novel method for the joint study of two set-typed dimensions and their\ninterplay. Based on a task analysis, we describe a new, multi-scale approach to\nthe interactive visual exploration and analysis of such data. Two set-typed\ndata dimensions are jointly visualized using a hierarchical matrix layout,\nenabling the analysis of the interactions between two set-typed attributes at\nseveral levels, in addition to the analysis of individual such dimensions.\nCrossSet is anchored at a compact, large-scale overview that is complemented by\ndrill-down opportunities to study the relations between and within the\nset-typed dimensions, enabling an interactive visual multi-scale exploration\nand analysis of bivariate set-typed data. Such an interactive approach makes it\npossible to study single set-typed dimensions in detail, to gain an overview of\nthe interaction and association between two such dimensions, to refine one of\nthe dimensions to gain additional details at several levels, and to drill down\nto the specific interactions of individual set-elements from the set-typed\ndimensions. To demonstrate the effectiveness and efficiency of CrossSet, we\nhave evaluated the new method in the context of several application scenarios.', 'authors': ['Kresimir Matkovic', 'Rainer Splechtna', 'Denis Gracanin', 'Helwig Hauser'], 'published': '2025-08-01 08:25:53+00:00', 'categories': ['cs.GR'], 'url': 'http://arxiv.org/abs/2508.00424v1'}
|
45
|
2508.00256
| 2,508.00256
| null |
Large AI Model-Enabled Secure Communications in Low-Altitude Wireless Networks: Concepts, Perspectives and Case Study
|
Low-altitude wireless networks (LAWNs) have the potential to revolutionize
communications by supporting a range of applications, including urban parcel
delivery, aerial inspections and air taxis. However, compared with traditional
wireless networks, LAWNs face unique security challenges due to low-altitude
operations, frequent mobility and reliance on unlicensed spectrum, making it
more vulnerable to some malicious attacks. In this paper, we investigate some
large artificial intelligence model (LAM)-enabled solutions for secure
communications in LAWNs. Specifically, we first explore the amplified security
risks and important limitations of traditional AI methods in LAWNs. Then, we
introduce the basic concepts of LAMs and delve into the role of LAMs in
addressing these challenges. To demonstrate the practical benefits of LAMs for
secure communications in LAWNs, we propose a novel LAM-based optimization
framework that leverages large language models (LLMs) to generate enhanced
state features on top of handcrafted representations, and to design intrinsic
rewards accordingly, thereby improving reinforcement learning performance for
secure communication tasks. Through a typical case study, simulation results
validate the effectiveness of the proposed framework. Finally, we outline
future directions for integrating LAMs into secure LAWN applications.
|
2025-08-01
|
{'id': '2508.00256', 'title': 'Large AI Model-Enabled Secure Communications in Low-Altitude Wireless Networks: Concepts, Perspectives and Case Study', 'abstract': 'Low-altitude wireless networks (LAWNs) have the potential to revolutionize\ncommunications by supporting a range of applications, including urban parcel\ndelivery, aerial inspections and air taxis. However, compared with traditional\nwireless networks, LAWNs face unique security challenges due to low-altitude\noperations, frequent mobility and reliance on unlicensed spectrum, making it\nmore vulnerable to some malicious attacks. In this paper, we investigate some\nlarge artificial intelligence model (LAM)-enabled solutions for secure\ncommunications in LAWNs. Specifically, we first explore the amplified security\nrisks and important limitations of traditional AI methods in LAWNs. Then, we\nintroduce the basic concepts of LAMs and delve into the role of LAMs in\naddressing these challenges. To demonstrate the practical benefits of LAMs for\nsecure communications in LAWNs, we propose a novel LAM-based optimization\nframework that leverages large language models (LLMs) to generate enhanced\nstate features on top of handcrafted representations, and to design intrinsic\nrewards accordingly, thereby improving reinforcement learning performance for\nsecure communication tasks. Through a typical case study, simulation results\nvalidate the effectiveness of the proposed framework. Finally, we outline\nfuture directions for integrating LAMs into secure LAWN applications.', 'authors': ['Chuang Zhang', 'Geng Sun', 'Jiacheng Wang', 'Yijing Lin', 'Weijie Yuan', 'Sinem Coleri', 'Dusit Niyato', 'Tony Q. S. Quek'], 'published': '2025-08-01 01:53:58+00:00', 'categories': ['cs.NI', 'cs.AI'], 'url': 'http://arxiv.org/abs/2508.00256v1'}
|
46
|
2508.00404
| 2,508.00404
| null |
Quasi-Normal Modes and Nonlinear Electrodynamics in Black Hole Phase Transitions
|
We investigate the connection between thermodynamic phase transitions and
quasi-normal modes (QNMs) in charged black holes with a positive curvature
constant, within the framework of $F(R)$-Euler-Heisenberg gravity. Nonlinear
electromagnetic fields lead to rich thermodynamic phase structures and
significantly affect the QNMs of massless scalar fields. By analyzing the QNMs
spectrum, we find that the transition point marking the disappearance of
divergence in the QNMs slope parameter $K$ aligns with the change of the
thermodynamic phase structure described by the heat capacity, within the bounds
of computational uncertainty. This precise matching holds under variations of
curvature parameters and charge. Furthermore, we show that larger angular
quantum number $l$ diminishes this correspondence, while higher overtone number
$n$ restores it beyond a threshold. These findings demonstrate that
thermodynamic phase transitions of black holes carry embedded dynamical
information, uncovering a fundamental link between black hole thermodynamic and
dynamical properties.
|
2025-08-01
|
{'id': '2508.00404', 'title': 'Quasi-Normal Modes and Nonlinear Electrodynamics in Black Hole Phase Transitions', 'abstract': 'We investigate the connection between thermodynamic phase transitions and\nquasi-normal modes (QNMs) in charged black holes with a positive curvature\nconstant, within the framework of $F(R)$-Euler-Heisenberg gravity. Nonlinear\nelectromagnetic fields lead to rich thermodynamic phase structures and\nsignificantly affect the QNMs of massless scalar fields. By analyzing the QNMs\nspectrum, we find that the transition point marking the disappearance of\ndivergence in the QNMs slope parameter $K$ aligns with the change of the\nthermodynamic phase structure described by the heat capacity, within the bounds\nof computational uncertainty. This precise matching holds under variations of\ncurvature parameters and charge. Furthermore, we show that larger angular\nquantum number $l$ diminishes this correspondence, while higher overtone number\n$n$ restores it beyond a threshold. These findings demonstrate that\nthermodynamic phase transitions of black holes carry embedded dynamical\ninformation, uncovering a fundamental link between black hole thermodynamic and\ndynamical properties.', 'authors': ['Zi-Yu Hou', 'Yu-Qi Lei', 'Xian-Hui Ge'], 'published': '2025-08-01 08:05:46+00:00', 'categories': ['hep-th', 'astro-ph.HE', 'gr-qc'], 'url': 'http://arxiv.org/abs/2508.00404v1'}
|
47
|
2508.00660
| 2,508.0066
| null |
Whitham-Broer-Kaup Systems in Multi-dimensions: Quantum and Resonant NLS Connections
|
An overview is presented of quantum and resonant nonlinear Schr\"odinger
equation links to Whitham-Broer-Kaup type systems. A novel n+1-dimensional
extension of the Whitham-Broer-Kaup hydrodynamic system is constructed with
connection to an equivalent multi-dimensional resonant NLS equation. Hybrid
Ermakov-Painlev\'e II and associated Painlev\'e XXXIV integrable similarity
reductions are derived.
|
2025-08-01
|
{'id': '2508.00660', 'title': 'Whitham-Broer-Kaup Systems in Multi-dimensions: Quantum and Resonant NLS Connections', 'abstract': 'An overview is presented of quantum and resonant nonlinear Schr\\"odinger\nequation links to Whitham-Broer-Kaup type systems. A novel n+1-dimensional\nextension of the Whitham-Broer-Kaup hydrodynamic system is constructed with\nconnection to an equivalent multi-dimensional resonant NLS equation. Hybrid\nErmakov-Painlev\\\'e II and associated Painlev\\\'e XXXIV integrable similarity\nreductions are derived.', 'authors': ['Oktay K Pashaev', 'Colin Rogers'], 'published': '2025-08-01 14:27:23+00:00', 'categories': ['nlin.SI'], 'url': 'http://arxiv.org/abs/2508.00660v1'}
|
48
|
2508.00745
| 2,508.00745
| null |
Irreducible components of Toric Complete Intersections
|
An equivariant linear system on a toric variety is a linear system invariant
under the torus action. We study the number of irreducible components of the
complete intersection of general divisors from a fixed collection of
equivariant linear system on a toric variety $X$. An explicit formula for the
number of components was obtained by Khovanskii in 2016 for the case $X = T^n$
over $C$ and generalized to an algebraically closed field of arbitrary
characteristic the author in 2024. Building on these results, we give a
recursive formula for an arbitrary toric variety.
|
2025-08-01
|
{'id': '2508.00745', 'title': 'Irreducible components of Toric Complete Intersections', 'abstract': 'An equivariant linear system on a toric variety is a linear system invariant\nunder the torus action. We study the number of irreducible components of the\ncomplete intersection of general divisors from a fixed collection of\nequivariant linear system on a toric variety $X$. An explicit formula for the\nnumber of components was obtained by Khovanskii in 2016 for the case $X = T^n$\nover $\\mathbb C$ and generalized to an algebraically closed field of arbitrary\ncharacteristic the author in 2024. Building on these results, we give a\nrecursive formula for an arbitrary toric variety.', 'authors': ['Andrey Zhizhin'], 'published': '2025-08-01 16:20:08+00:00', 'categories': ['math.AG'], 'url': 'http://arxiv.org/abs/2508.00745v1'}
|
49
|
2508.00575
| 2,508.00575
| null |
Analysing Temporal Reasoning in Description Logics Using Formal Grammars
|
We establish a correspondence between (fragments of)
$TEL^$, a temporal extension of the $EL$
description logic with the LTL operator $^k$, and some specific kinds
of formal grammars, in particular, conjunctive grammars (context-free grammars
equipped with the operation of intersection). This connection implies that
$TEL^$ does not possess the property of ultimate periodicity
of models, and further leads to undecidability of query answering in
$TEL^$, closing a question left open since the introduction
of $TEL^$. Moreover, it also allows to establish decidability
of query answering for some new interesting fragments of
$TEL^$, and to reuse for this purpose existing tools and
algorithms for conjunctive grammars.
|
2025-08-01
|
{'id': '2508.00575', 'title': 'Analysing Temporal Reasoning in Description Logics Using Formal Grammars', 'abstract': 'We establish a correspondence between (fragments of)\n$\\mathcal{TEL}^\\bigcirc$, a temporal extension of the $\\mathcal{EL}$\ndescription logic with the LTL operator $\\bigcirc^k$, and some specific kinds\nof formal grammars, in particular, conjunctive grammars (context-free grammars\nequipped with the operation of intersection). This connection implies that\n$\\mathcal{TEL}^\\bigcirc$ does not possess the property of ultimate periodicity\nof models, and further leads to undecidability of query answering in\n$\\mathcal{TEL}^\\bigcirc$, closing a question left open since the introduction\nof $\\mathcal{TEL}^\\bigcirc$. Moreover, it also allows to establish decidability\nof query answering for some new interesting fragments of\n$\\mathcal{TEL}^\\bigcirc$, and to reuse for this purpose existing tools and\nalgorithms for conjunctive grammars.', 'authors': ['Camille Bourgaux', 'Anton Gnatenko', 'Michaël Thomazo'], 'published': '2025-08-01 12:17:49+00:00', 'categories': ['cs.LO', 'cs.AI'], 'url': 'http://arxiv.org/abs/2508.00575v1'}
|
51
|
2508.00736
| 2,508.00736
| null |
A normalizing flow approach for the inference of star cluster properties from unresolved broadband photometry I: Comparison to spectral energy distribution fitting
|
Estimating properties of star clusters from unresolved broadband photometry
is a challenging problem that is classically tackled by spectral energy
distribution (SED) fitting methods that are based on simple stellar population
(SSP) models. However, because of their exponential scaling, grid-based methods
suffer from computational limitations. In addition, stochastic latent variables
in the model can make the computation of the likelihood function intractable.
These limitations can be overcome by modern generative deep learning methods
that offer flexible and powerful tools for modeling high-dimensional posterior
distributions and fast inference from learned data. We present a normalizing
flow approach for the inference of cluster age, mass, and reddening from Hubble
Space Telescope (HST) broadband photometry. In particular, we explore our
network's behavior on an inference problem that has been analyzed in previous
works. We used the SED modeling code CIGALE to create a dataset of synthetic
photometric observations for $5 10^6$ mock star clusters. Subsequently,
this data set was used to train a coupling-based flow in the form of a
conditional invertible neural network (cINN) to predict posterior probability
distributions for cluster age, mass, and reddening from photometric
observations. We predicted cluster parameters for the 'Physics at High Angular
resolution in Nearby GalaxieS' (PHANGS) Data Release 3 catalog. To evaluate the
capabilities of the network, we compared our results to the publicly available
PHANGS estimates and found that the estimates agree reasonably well. We
demonstrate that normalizing flow methods can be a viable tool for the
inference of cluster parameters, and argue that this approach is especially
useful when latent variables make the computation of the likelihood intractable
and in scenarios that require efficient density estimation.
|
2025-08-01
|
{'id': '2508.00736', 'title': 'A normalizing flow approach for the inference of star cluster properties from unresolved broadband photometry I: Comparison to spectral energy distribution fitting', 'abstract': "Estimating properties of star clusters from unresolved broadband photometry\nis a challenging problem that is classically tackled by spectral energy\ndistribution (SED) fitting methods that are based on simple stellar population\n(SSP) models. However, because of their exponential scaling, grid-based methods\nsuffer from computational limitations. In addition, stochastic latent variables\nin the model can make the computation of the likelihood function intractable.\nThese limitations can be overcome by modern generative deep learning methods\nthat offer flexible and powerful tools for modeling high-dimensional posterior\ndistributions and fast inference from learned data. We present a normalizing\nflow approach for the inference of cluster age, mass, and reddening from Hubble\nSpace Telescope (HST) broadband photometry. In particular, we explore our\nnetwork's behavior on an inference problem that has been analyzed in previous\nworks. We used the SED modeling code CIGALE to create a dataset of synthetic\nphotometric observations for $5 \\times 10^6$ mock star clusters. Subsequently,\nthis data set was used to train a coupling-based flow in the form of a\nconditional invertible neural network (cINN) to predict posterior probability\ndistributions for cluster age, mass, and reddening from photometric\nobservations. We predicted cluster parameters for the 'Physics at High Angular\nresolution in Nearby GalaxieS' (PHANGS) Data Release 3 catalog. To evaluate the\ncapabilities of the network, we compared our results to the publicly available\nPHANGS estimates and found that the estimates agree reasonably well. We\ndemonstrate that normalizing flow methods can be a viable tool for the\ninference of cluster parameters, and argue that this approach is especially\nuseful when latent variables make the computation of the likelihood intractable\nand in scenarios that require efficient density estimation.", 'authors': ['Daniel Walter', 'Victor F. Ksoll', 'Ralf S. Klessen', 'Mederic Boquien', 'Aida Wofford', 'Francesco Belfiore', 'Daniel A. Dale', 'Kathryn Grasha', 'David A. Thilker', 'Leonardo Ubeda', 'Thomas G. Williams'], 'published': '2025-08-01 16:06:13+00:00', 'categories': ['astro-ph.GA', 'astro-ph.IM'], 'url': 'http://arxiv.org/abs/2508.00736v1'}
|
52
|
2508.00530
| 2,508.0053
| null |
Quantum repeaters based on stationary and flying Gottesman-Kitaev-Preskill qudits
|
There are various approaches to long-range quantum communication based on
conceptually different forms of quantum repeaters. Here we explore a quantum
repeater scheme that employs quantum error correction (QEC) both on the flying
(light) qubits and on the stationary (matter) qubits. The idea is to combine
the benefits of encoded one-way and two-way schemes where effective channel
transmission and loss scaling are enhanced by means of photon loss codes and
encoded quantum memories, respectively, while sacrificing some of their
advantages such as high clock rates, independent of classical communication
times (one-way), and potentially large segment lengths (two-way). More
specifically, we illustrate, propose, and analyze such a quantum repeater using
the bosonic Gottesman-Kitaev-Preskill (GKP) code which naturally enables
encoding and QEC of qudits, protecting them against transmission and memory
loss, the latter, for instance, occuring on collective spin modes of atomic
ensembles. While the encoded one-way and two-way schemes on their own either
require very high repeater link coupling efficiencies and GKP squeezing or
allow for experimentally more feasible, small values of these parameters,
respectively, we find that there are intermediate parameter regimes where the
combined repeater protocol is superior.
|
2025-08-01
|
{'id': '2508.00530', 'title': 'Quantum repeaters based on stationary and flying Gottesman-Kitaev-Preskill qudits', 'abstract': 'There are various approaches to long-range quantum communication based on\nconceptually different forms of quantum repeaters. Here we explore a quantum\nrepeater scheme that employs quantum error correction (QEC) both on the flying\n(light) qubits and on the stationary (matter) qubits. The idea is to combine\nthe benefits of encoded one-way and two-way schemes where effective channel\ntransmission and loss scaling are enhanced by means of photon loss codes and\nencoded quantum memories, respectively, while sacrificing some of their\nadvantages such as high clock rates, independent of classical communication\ntimes (one-way), and potentially large segment lengths (two-way). More\nspecifically, we illustrate, propose, and analyze such a quantum repeater using\nthe bosonic Gottesman-Kitaev-Preskill (GKP) code which naturally enables\nencoding and QEC of qudits, protecting them against transmission and memory\nloss, the latter, for instance, occuring on collective spin modes of atomic\nensembles. While the encoded one-way and two-way schemes on their own either\nrequire very high repeater link coupling efficiencies and GKP squeezing or\nallow for experimentally more feasible, small values of these parameters,\nrespectively, we find that there are intermediate parameter regimes where the\ncombined repeater protocol is superior.', 'authors': ['Stefan Häussler', 'Peter van Loock'], 'published': '2025-08-01 11:12:39+00:00', 'categories': ['quant-ph'], 'url': 'http://arxiv.org/abs/2508.00530v1'}
|
53
|
2508.00320
| 2,508.0032
| null |
Enhancement of non-Markovianity due to environment-induced indirect interaction
|
Non-Markovian effects are often significant when the system-environment
coupling is not weak. Indeed, we find the non-Markovianity to be negligible for
a single two-level system undergoing pure dephasing via interaction with a
harmonic oscillator environment. In this paper, we examine a natural extension,
namely a pure dephasing model where a collection of two-level systems interacts
with a common environment. We obtain analytically the dynamics of the
collection of the two-level systems, and then take a partial trace over all the
two-level systems except one. This remaining single two-level system is shown
to display markedly non-Markovian dynamics, even in the weak system-environment
coupling regime. This is due to the indirect interaction between two-level
systems induced by their interaction with the common environment. In fact, this
indirect interaction can not only increase the non-Markovianity by orders of
magnitude, but also display qualitatively different characteristics. For
instance, for a single two-level system undergoing pure dephasing, the dynamics
are Markovian for Ohmic and sub-Ohmic environments. This is markedly not the
case when we consider multiple two-level systems. We also show that the
non-Markovianity increases as we increase the number of two-level systems.
These findings provide insights into controlling decoherence in multi-qubit
quantum systems and have implications for quantum technologies where
non-Markovianity can be a resource rather than a limitation.
|
2025-08-01
|
{'id': '2508.00320', 'title': 'Enhancement of non-Markovianity due to environment-induced indirect interaction', 'abstract': 'Non-Markovian effects are often significant when the system-environment\ncoupling is not weak. Indeed, we find the non-Markovianity to be negligible for\na single two-level system undergoing pure dephasing via interaction with a\nharmonic oscillator environment. In this paper, we examine a natural extension,\nnamely a pure dephasing model where a collection of two-level systems interacts\nwith a common environment. We obtain analytically the dynamics of the\ncollection of the two-level systems, and then take a partial trace over all the\ntwo-level systems except one. This remaining single two-level system is shown\nto display markedly non-Markovian dynamics, even in the weak system-environment\ncoupling regime. This is due to the indirect interaction between two-level\nsystems induced by their interaction with the common environment. In fact, this\nindirect interaction can not only increase the non-Markovianity by orders of\nmagnitude, but also display qualitatively different characteristics. For\ninstance, for a single two-level system undergoing pure dephasing, the dynamics\nare Markovian for Ohmic and sub-Ohmic environments. This is markedly not the\ncase when we consider multiple two-level systems. We also show that the\nnon-Markovianity increases as we increase the number of two-level systems.\nThese findings provide insights into controlling decoherence in multi-qubit\nquantum systems and have implications for quantum technologies where\nnon-Markovianity can be a resource rather than a limitation.', 'authors': ['Asif Zaman', 'Muhammad Faryad', 'Adam Zaman Chaudhry'], 'published': '2025-08-01 05:07:25+00:00', 'categories': ['quant-ph'], 'url': 'http://arxiv.org/abs/2508.00320v1'}
|
54
|
2508.00472
| 2,508.00472
| null |
A Conditional GAN for Tabular Data Generation with Probabilistic Sampling of Latent Subspaces
|
The tabular form constitutes the standard way of representing data in
relational database systems and spreadsheets. But, similarly to other forms,
tabular data suffers from class imbalance, a problem that causes serious
performance degradation in a wide variety of machine learning tasks. One of the
most effective solutions dictates the usage of Generative Adversarial Networks
(GANs) in order to synthesize artificial data instances for the
under-represented classes. Despite their good performance, none of the proposed
GAN models takes into account the vector subspaces of the input samples in the
real data space, leading to data generation in arbitrary locations. Moreover,
the class labels are treated in the same manner as the other categorical
variables during training, so conditional sampling by class is rendered less
effective. To overcome these problems, this study presents ctdGAN, a
conditional GAN for alleviating class imbalance in tabular datasets. Initially,
ctdGAN executes a space partitioning step to assign cluster labels to the input
samples. Subsequently, it utilizes these labels to synthesize samples via a
novel probabilistic sampling strategy and a new loss function that penalizes
both cluster and class mis-predictions. In this way, ctdGAN is trained to
generate samples in subspaces that resemble those of the original data
distribution. We also introduce several other improvements, including a simple,
yet effective cluster-wise scaling technique that captures multiple feature
modes without affecting data dimensionality. The exhaustive evaluation of
ctdGAN with 14 imbalanced datasets demonstrated its superiority in generating
high fidelity samples and improving classification accuracy.
|
2025-08-01
|
{'id': '2508.00472', 'title': 'A Conditional GAN for Tabular Data Generation with Probabilistic Sampling of Latent Subspaces', 'abstract': 'The tabular form constitutes the standard way of representing data in\nrelational database systems and spreadsheets. But, similarly to other forms,\ntabular data suffers from class imbalance, a problem that causes serious\nperformance degradation in a wide variety of machine learning tasks. One of the\nmost effective solutions dictates the usage of Generative Adversarial Networks\n(GANs) in order to synthesize artificial data instances for the\nunder-represented classes. Despite their good performance, none of the proposed\nGAN models takes into account the vector subspaces of the input samples in the\nreal data space, leading to data generation in arbitrary locations. Moreover,\nthe class labels are treated in the same manner as the other categorical\nvariables during training, so conditional sampling by class is rendered less\neffective. To overcome these problems, this study presents ctdGAN, a\nconditional GAN for alleviating class imbalance in tabular datasets. Initially,\nctdGAN executes a space partitioning step to assign cluster labels to the input\nsamples. Subsequently, it utilizes these labels to synthesize samples via a\nnovel probabilistic sampling strategy and a new loss function that penalizes\nboth cluster and class mis-predictions. In this way, ctdGAN is trained to\ngenerate samples in subspaces that resemble those of the original data\ndistribution. We also introduce several other improvements, including a simple,\nyet effective cluster-wise scaling technique that captures multiple feature\nmodes without affecting data dimensionality. The exhaustive evaluation of\nctdGAN with 14 imbalanced datasets demonstrated its superiority in generating\nhigh fidelity samples and improving classification accuracy.', 'authors': ['Leonidas Akritidis', 'Panayiotis Bozanis'], 'published': '2025-08-01 09:49:57+00:00', 'categories': ['cs.LG'], 'url': 'http://arxiv.org/abs/2508.00472v1'}
|
55
|
2508.00741
| 2,508.00741
| null |
Out-of-Context Abduction: LLMs Make Inferences About Procedural Data Leveraging Declarative Facts in Earlier Training Data
|
Large language models (LLMs) are trained on large corpora, yet it is unclear
whether they can reason about the information present within their training
data. We design experiments to study out-of-context abduction in LLMs, the
ability to infer the most plausible explanations for observations using
relevant facts present in training data. We train treatment LLMs on names and
behavior descriptions of fictitious chatbots, but not on examples of dialogue
with the chatbots. We find that OpenAI's GPT 4o LLM can correctly infer at
least one chatbot's name after observing example responses characteristic of
that chatbot. We also find that previously training GPT 4o on descriptions of a
chatbot's behavior allows it to display behaviors more characteristic of the
chatbot when iteratively trained to display such behaviors. Our results have
implications for situational awareness in LLMs and, therefore, for AI safety.
|
2025-08-01
|
{'id': '2508.00741', 'title': 'Out-of-Context Abduction: LLMs Make Inferences About Procedural Data Leveraging Declarative Facts in Earlier Training Data', 'abstract': "Large language models (LLMs) are trained on large corpora, yet it is unclear\nwhether they can reason about the information present within their training\ndata. We design experiments to study out-of-context abduction in LLMs, the\nability to infer the most plausible explanations for observations using\nrelevant facts present in training data. We train treatment LLMs on names and\nbehavior descriptions of fictitious chatbots, but not on examples of dialogue\nwith the chatbots. We find that OpenAI's GPT 4o LLM can correctly infer at\nleast one chatbot's name after observing example responses characteristic of\nthat chatbot. We also find that previously training GPT 4o on descriptions of a\nchatbot's behavior allows it to display behaviors more characteristic of the\nchatbot when iteratively trained to display such behaviors. Our results have\nimplications for situational awareness in LLMs and, therefore, for AI safety.", 'authors': ['Sohaib Imran', 'Rob Lamb', 'Peter M. Atkinson'], 'published': '2025-08-01 16:12:23+00:00', 'categories': ['cs.CL', 'cs.AI'], 'url': 'http://arxiv.org/abs/2508.00741v1'}
|
56
|
2508.00682
| 2,508.00682
| null |
Unveiling Dynamic Binary Instrumentation Techniques
|
Dynamic Binary Instrumentation (DBI) is the set of techniques that enable
instrumentation of programs at run-time, making it possible to monitor and
modify the execution of compiled binaries or entire systems. DBI is used for
countless security applications and analyses, and is extensively used across
many fields in both industry and academia. Over the years, several DBI
approaches have been proposed based on different technologies and implementing
diverse techniques. Every solution tries to overcome certain limitations, but
they sometimes bring other shortcomings. Some are specialized for one
particular domain or task, while others have a wider scope.
In this paper, we shed light into the labyrinth of DBI, bringing together
process-level and whole-system approaches. We depict their building blocks and
analyze the underlying instrumentation techniques, comparing their ability to
instrument different primitives and run-time events. Then, we evaluate their
performance when implementing each primitive, and highlight relevant
observations. Our results show that no single technique is better than the rest
in all circumstances.
|
2025-08-01
|
{'id': '2508.00682', 'title': 'Unveiling Dynamic Binary Instrumentation Techniques', 'abstract': 'Dynamic Binary Instrumentation (DBI) is the set of techniques that enable\ninstrumentation of programs at run-time, making it possible to monitor and\nmodify the execution of compiled binaries or entire systems. DBI is used for\ncountless security applications and analyses, and is extensively used across\nmany fields in both industry and academia. Over the years, several DBI\napproaches have been proposed based on different technologies and implementing\ndiverse techniques. Every solution tries to overcome certain limitations, but\nthey sometimes bring other shortcomings. Some are specialized for one\nparticular domain or task, while others have a wider scope.\n In this paper, we shed light into the labyrinth of DBI, bringing together\nprocess-level and whole-system approaches. We depict their building blocks and\nanalyze the underlying instrumentation techniques, comparing their ability to\ninstrument different primitives and run-time events. Then, we evaluate their\nperformance when implementing each primitive, and highlight relevant\nobservations. Our results show that no single technique is better than the rest\nin all circumstances.', 'authors': ['Oscar Llorente-Vazquez', 'Xabier Ugarte-Pedrero', 'Igor Santos-Grueiro', 'Pablo Garcia Bringas'], 'published': '2025-08-01 14:53:03+00:00', 'categories': ['cs.CR', 'cs.SE'], 'url': 'http://arxiv.org/abs/2508.00682v1'}
|
57
|
2508.00398
| 2,508.00398
| null |
Occlusion-robust Stylization for Drawing-based 3D Animation
|
3D animation aims to generate a 3D animated video from an input image and a
target 3D motion sequence. Recent advances in image-to-3D models enable the
creation of animations directly from user-hand drawings. Distinguished from
conventional 3D animation, drawing-based 3D animation is crucial to preserve
artist's unique style properties, such as rough contours and distinct stroke
patterns. However, recent methods still exhibit quality deterioration in style
properties, especially under occlusions caused by overlapping body parts,
leading to contour flickering and stroke blurring. This occurs due to a
`stylization pose gap' between training and inference in stylization networks
designed to preserve drawing styles in drawing-based 3D animation systems. The
stylization pose gap denotes that input target poses used to train the
stylization network are always in occlusion-free poses, while target poses
encountered in an inference include diverse occlusions under dynamic motions.
To this end, we propose Occlusion-robust Stylization Framework (OSF) for
drawing-based 3D animation. We found that while employing object's edge can be
effective input prior for guiding stylization, it becomes notably inaccurate
when occlusions occur at inference. Thus, our proposed OSF provides
occlusion-robust edge guidance for stylization network using optical flow,
ensuring a consistent stylization even under occlusions. Furthermore, OSF
operates in a single run instead of the previous two-stage method, achieving
2.4x faster inference and 2.1x less memory.
|
2025-08-01
|
{'id': '2508.00398', 'title': 'Occlusion-robust Stylization for Drawing-based 3D Animation', 'abstract': "3D animation aims to generate a 3D animated video from an input image and a\ntarget 3D motion sequence. Recent advances in image-to-3D models enable the\ncreation of animations directly from user-hand drawings. Distinguished from\nconventional 3D animation, drawing-based 3D animation is crucial to preserve\nartist's unique style properties, such as rough contours and distinct stroke\npatterns. However, recent methods still exhibit quality deterioration in style\nproperties, especially under occlusions caused by overlapping body parts,\nleading to contour flickering and stroke blurring. This occurs due to a\n`stylization pose gap' between training and inference in stylization networks\ndesigned to preserve drawing styles in drawing-based 3D animation systems. The\nstylization pose gap denotes that input target poses used to train the\nstylization network are always in occlusion-free poses, while target poses\nencountered in an inference include diverse occlusions under dynamic motions.\nTo this end, we propose Occlusion-robust Stylization Framework (OSF) for\ndrawing-based 3D animation. We found that while employing object's edge can be\neffective input prior for guiding stylization, it becomes notably inaccurate\nwhen occlusions occur at inference. Thus, our proposed OSF provides\nocclusion-robust edge guidance for stylization network using optical flow,\nensuring a consistent stylization even under occlusions. Furthermore, OSF\noperates in a single run instead of the previous two-stage method, achieving\n2.4x faster inference and 2.1x less memory.", 'authors': ['Sunjae Yoon', 'Gwanhyeong Koo', 'Younghwan Lee', 'Ji Woo Hong', 'Chang D. Yoo'], 'published': '2025-08-01 07:52:07+00:00', 'categories': ['cs.GR', 'cs.CV'], 'url': 'http://arxiv.org/abs/2508.00398v1'}
|
58
|
2508.00272
| 2,508.00272
| null |
Towards Robust Semantic Correspondence: A Benchmark and Insights
|
Semantic correspondence aims to identify semantically meaningful
relationships between different images and is a fundamental challenge in
computer vision. It forms the foundation for numerous tasks such as 3D
reconstruction, object tracking, and image editing. With the progress of
large-scale vision models, semantic correspondence has achieved remarkable
performance in controlled and high-quality conditions. However, the robustness
of semantic correspondence in challenging scenarios is much less investigated.
In this work, we establish a novel benchmark for evaluating semantic
correspondence in adverse conditions. The benchmark dataset comprises 14
distinct challenging scenarios that reflect commonly encountered imaging
issues, including geometric distortion, image blurring, digital artifacts, and
environmental occlusion. Through extensive evaluations, we provide several key
insights into the robustness of semantic correspondence approaches: (1) All
existing methods suffer from noticeable performance drops under adverse
conditions; (2) Using large-scale vision models can enhance overall robustness,
but fine-tuning on these models leads to a decline in relative robustness; (3)
The DINO model outperforms the Stable Diffusion in relative robustness, and
their fusion achieves better absolute robustness; Moreover, We evaluate common
robustness enhancement strategies for semantic correspondence and find that
general data augmentations are ineffective, highlighting the need for
task-specific designs. These results are consistent across both our dataset and
real-world benchmarks.
|
2025-08-01
|
{'id': '2508.00272', 'title': 'Towards Robust Semantic Correspondence: A Benchmark and Insights', 'abstract': 'Semantic correspondence aims to identify semantically meaningful\nrelationships between different images and is a fundamental challenge in\ncomputer vision. It forms the foundation for numerous tasks such as 3D\nreconstruction, object tracking, and image editing. With the progress of\nlarge-scale vision models, semantic correspondence has achieved remarkable\nperformance in controlled and high-quality conditions. However, the robustness\nof semantic correspondence in challenging scenarios is much less investigated.\nIn this work, we establish a novel benchmark for evaluating semantic\ncorrespondence in adverse conditions. The benchmark dataset comprises 14\ndistinct challenging scenarios that reflect commonly encountered imaging\nissues, including geometric distortion, image blurring, digital artifacts, and\nenvironmental occlusion. Through extensive evaluations, we provide several key\ninsights into the robustness of semantic correspondence approaches: (1) All\nexisting methods suffer from noticeable performance drops under adverse\nconditions; (2) Using large-scale vision models can enhance overall robustness,\nbut fine-tuning on these models leads to a decline in relative robustness; (3)\nThe DINO model outperforms the Stable Diffusion in relative robustness, and\ntheir fusion achieves better absolute robustness; Moreover, We evaluate common\nrobustness enhancement strategies for semantic correspondence and find that\ngeneral data augmentations are ineffective, highlighting the need for\ntask-specific designs. These results are consistent across both our dataset and\nreal-world benchmarks.', 'authors': ['Wenyue Chong'], 'published': '2025-08-01 02:38:39+00:00', 'categories': ['cs.CV'], 'url': 'http://arxiv.org/abs/2508.00272v1'}
|
59
|
2508.00533
| 2,508.00533
| null |
Beyond asymptotic reasoning: a practical ground state projector based on the wall-Chebyshev expansion
|
We introduce a quantum algorithm for ground-state preparation based on a
Chebyshev series approximation to the wall function. This projector can be
efficiently implemented as a product of Hamiltonian operators, enabling a
straightforward realization via the linear combinations of unitaries method. We
analyze the asymptotic scaling and provide numerical benchmarks, demonstrating
that the wall-Chebyshev projector achieves competitive performance with leading
methods based on imaginary time evolution and alternative projector function
approximations. Notably, our approach exhibits superior robustness and
convergence in scenarios where accurate ground-state energy estimates are
unavailable, showing promise for realistic chemistry problems.
|
2025-08-01
|
{'id': '2508.00533', 'title': 'Beyond asymptotic reasoning: a practical ground state projector based on the wall-Chebyshev expansion', 'abstract': 'We introduce a quantum algorithm for ground-state preparation based on a\nChebyshev series approximation to the wall function. This projector can be\nefficiently implemented as a product of Hamiltonian operators, enabling a\nstraightforward realization via the linear combinations of unitaries method. We\nanalyze the asymptotic scaling and provide numerical benchmarks, demonstrating\nthat the wall-Chebyshev projector achieves competitive performance with leading\nmethods based on imaginary time evolution and alternative projector function\napproximations. Notably, our approach exhibits superior robustness and\nconvergence in scenarios where accurate ground-state energy estimates are\nunavailable, showing promise for realistic chemistry problems.', 'authors': ['Maria-Andreea Filip', 'Nathan Fitzpatrick'], 'published': '2025-08-01 11:15:15+00:00', 'categories': ['quant-ph', 'physics.chem-ph'], 'url': 'http://arxiv.org/abs/2508.00533v1'}
|
61
|
2508.00435
| 2,508.00435
| null |
Gapless superconductivity from extremely dilute magnetic disorder in 2H-NbSe2-xSx
|
Most superconducting materials exhibit a vanishing density of states at the
Fermi level and Anderson's theorem posits that the superconducting gap is
robust against nonmagnetic disorder. Although dilute magnetic impurities lead
to localized in-gap states, these states typically have no bearing on the
material's bulk superconducting properties. However, numerous experiments
reveal a finite density of states at the Fermi level in systems with an
apparently negligible number of magnetic impurities. Here, using scanning
tunneling microscopy and self-consistent Bogoliubov-de Gennes calculations, we
find that gapless superconductivity emerges in 2H-NbSe2-xSx at remarkably low
magnetic impurity concentrations. Furthermore, our density functional theory
calculations and in-gap quasiparticle interference measurements demonstrate
that the Se-S substitution significantly modifies the band structure. This
modification favours nesting and dictates the in-gap scattering for x>0, in
stark contrast to the dominant charge density wave interactions in pure
2H-NbSe2. Our findings reveal an unusual superconducting response to disorder
and highlight the importance of incorporating material-specific band structures
in the understanding of a superconductor's response to even very low
concentrations of magnetic impurities.
|
2025-08-01
|
{'id': '2508.00435', 'title': 'Gapless superconductivity from extremely dilute magnetic disorder in 2H-NbSe2-xSx', 'abstract': "Most superconducting materials exhibit a vanishing density of states at the\nFermi level and Anderson's theorem posits that the superconducting gap is\nrobust against nonmagnetic disorder. Although dilute magnetic impurities lead\nto localized in-gap states, these states typically have no bearing on the\nmaterial's bulk superconducting properties. However, numerous experiments\nreveal a finite density of states at the Fermi level in systems with an\napparently negligible number of magnetic impurities. Here, using scanning\ntunneling microscopy and self-consistent Bogoliubov-de Gennes calculations, we\nfind that gapless superconductivity emerges in 2H-NbSe2-xSx at remarkably low\nmagnetic impurity concentrations. Furthermore, our density functional theory\ncalculations and in-gap quasiparticle interference measurements demonstrate\nthat the Se-S substitution significantly modifies the band structure. This\nmodification favours nesting and dictates the in-gap scattering for x>0, in\nstark contrast to the dominant charge density wave interactions in pure\n2H-NbSe2. Our findings reveal an unusual superconducting response to disorder\nand highlight the importance of incorporating material-specific band structures\nin the understanding of a superconductor's response to even very low\nconcentrations of magnetic impurities.", 'authors': ['Jose Antonio Moreno', 'Mercè Roig', 'Víctor Barrena', 'Edwin Herrera', 'Alberto M. Ruiz', 'Samuel Mañas-Valero', 'Antón Fente', 'Anita Smeets', 'Jazmín Aragón', 'Yanina Fasano', 'Beilun Wu', 'Maria N. Gastiasoro', 'Eugenio Coronado', 'José J. Baldoví', 'Brian M. Andersen', 'Isabel Guillamón', 'Hermann Suderow'], 'published': '2025-08-01 08:48:53+00:00', 'categories': ['cond-mat.supr-con', 'cond-mat.mes-hall', 'cond-mat.mtrl-sci', 'cond-mat.str-el'], 'url': 'http://arxiv.org/abs/2508.00435v1'}
|
62
|
2508.00300
| 2,508.003
| null |
MetaExplainer: A Framework to Generate Multi-Type User-Centered Explanations for AI Systems
|
Explanations are crucial for building trustworthy AI systems, but a gap often
exists between the explanations provided by models and those needed by users.
To address this gap, we introduce MetaExplainer, a neuro-symbolic framework
designed to generate user-centered explanations. Our approach employs a
three-stage process: first, we decompose user questions into machine-readable
formats using state-of-the-art large language models (LLM); second, we delegate
the task of generating system recommendations to model explainer methods; and
finally, we synthesize natural language explanations that summarize the
explainer outputs. Throughout this process, we utilize an Explanation Ontology
to guide the language models and explainer methods. By leveraging LLMs and a
structured approach to explanation generation, MetaExplainer aims to enhance
the interpretability and trustworthiness of AI systems across various
applications, providing users with tailored, question-driven explanations that
better meet their needs. Comprehensive evaluations of MetaExplainer demonstrate
a step towards evaluating and utilizing current state-of-the-art explanation
frameworks. Our results show high performance across all stages, with a 59.06%
F1-score in question reframing, 70% faithfulness in model explanations, and 67%
context-utilization in natural language synthesis. User studies corroborate
these findings, highlighting the creativity and comprehensiveness of generated
explanations. Tested on the Diabetes (PIMA Indian) tabular dataset,
MetaExplainer supports diverse explanation types, including Contrastive,
Counterfactual, Rationale, Case-Based, and Data explanations. The framework's
versatility and traceability from using ontology to guide LLMs suggest broad
applicability beyond the tested scenarios, positioning MetaExplainer as a
promising tool for enhancing AI explainability across various domains.
|
2025-08-01
|
{'id': '2508.00300', 'title': 'MetaExplainer: A Framework to Generate Multi-Type User-Centered Explanations for AI Systems', 'abstract': "Explanations are crucial for building trustworthy AI systems, but a gap often\nexists between the explanations provided by models and those needed by users.\nTo address this gap, we introduce MetaExplainer, a neuro-symbolic framework\ndesigned to generate user-centered explanations. Our approach employs a\nthree-stage process: first, we decompose user questions into machine-readable\nformats using state-of-the-art large language models (LLM); second, we delegate\nthe task of generating system recommendations to model explainer methods; and\nfinally, we synthesize natural language explanations that summarize the\nexplainer outputs. Throughout this process, we utilize an Explanation Ontology\nto guide the language models and explainer methods. By leveraging LLMs and a\nstructured approach to explanation generation, MetaExplainer aims to enhance\nthe interpretability and trustworthiness of AI systems across various\napplications, providing users with tailored, question-driven explanations that\nbetter meet their needs. Comprehensive evaluations of MetaExplainer demonstrate\na step towards evaluating and utilizing current state-of-the-art explanation\nframeworks. Our results show high performance across all stages, with a 59.06%\nF1-score in question reframing, 70% faithfulness in model explanations, and 67%\ncontext-utilization in natural language synthesis. User studies corroborate\nthese findings, highlighting the creativity and comprehensiveness of generated\nexplanations. Tested on the Diabetes (PIMA Indian) tabular dataset,\nMetaExplainer supports diverse explanation types, including Contrastive,\nCounterfactual, Rationale, Case-Based, and Data explanations. The framework's\nversatility and traceability from using ontology to guide LLMs suggest broad\napplicability beyond the tested scenarios, positioning MetaExplainer as a\npromising tool for enhancing AI explainability across various domains.", 'authors': ['Shruthi Chari', 'Oshani Seneviratne', 'Prithwish Chakraborty', 'Pablo Meyer', 'Deborah L. McGuinness'], 'published': '2025-08-01 04:01:40+00:00', 'categories': ['cs.HC', 'cs.AI', 'cs.LG'], 'url': 'http://arxiv.org/abs/2508.00300v1'}
|
63
|
2508.00443
| 2,508.00443
| null |
SDMatte: Grafting Diffusion Models for Interactive Matting
|
Recent interactive matting methods have shown satisfactory performance in
capturing the primary regions of objects, but they fall short in extracting
fine-grained details in edge regions. Diffusion models trained on billions of
image-text pairs, demonstrate exceptional capability in modeling highly complex
data distributions and synthesizing realistic texture details, while exhibiting
robust text-driven interaction capabilities, making them an attractive solution
for interactive matting. To this end, we propose SDMatte, a diffusion-driven
interactive matting model, with three key contributions. First, we exploit the
powerful priors of diffusion models and transform the text-driven interaction
capability into visual prompt-driven interaction capability to enable
interactive matting. Second, we integrate coordinate embeddings of visual
prompts and opacity embeddings of target objects into U-Net, enhancing
SDMatte's sensitivity to spatial position information and opacity information.
Third, we propose a masked self-attention mechanism that enables the model to
focus on areas specified by visual prompts, leading to better performance.
Extensive experiments on multiple datasets demonstrate the superior performance
of our method, validating its effectiveness in interactive matting. Our code
and model are available at https://github.com/vivoCameraResearch/SDMatte.
|
2025-08-01
|
{'id': '2508.00443', 'title': 'SDMatte: Grafting Diffusion Models for Interactive Matting', 'abstract': "Recent interactive matting methods have shown satisfactory performance in\ncapturing the primary regions of objects, but they fall short in extracting\nfine-grained details in edge regions. Diffusion models trained on billions of\nimage-text pairs, demonstrate exceptional capability in modeling highly complex\ndata distributions and synthesizing realistic texture details, while exhibiting\nrobust text-driven interaction capabilities, making them an attractive solution\nfor interactive matting. To this end, we propose SDMatte, a diffusion-driven\ninteractive matting model, with three key contributions. First, we exploit the\npowerful priors of diffusion models and transform the text-driven interaction\ncapability into visual prompt-driven interaction capability to enable\ninteractive matting. Second, we integrate coordinate embeddings of visual\nprompts and opacity embeddings of target objects into U-Net, enhancing\nSDMatte's sensitivity to spatial position information and opacity information.\nThird, we propose a masked self-attention mechanism that enables the model to\nfocus on areas specified by visual prompts, leading to better performance.\nExtensive experiments on multiple datasets demonstrate the superior performance\nof our method, validating its effectiveness in interactive matting. Our code\nand model are available at https://github.com/vivoCameraResearch/SDMatte.", 'authors': ['Longfei Huang', 'Yu Liang', 'Hao Zhang', 'Jinwei Chen', 'Wei Dong', 'Lunde Chen', 'Wanyu Liu', 'Bo Li', 'Peng-Tao Jiang'], 'published': '2025-08-01 09:00:48+00:00', 'categories': ['cs.CV'], 'url': 'http://arxiv.org/abs/2508.00443v2'}
|
64
|
2508.00757
| 2,508.00757
| null |
GLiDRE: Generalist Lightweight model for Document-level Relation Extraction
|
Relation Extraction (RE) is a fundamental task in Natural Language
Processing, and its document-level variant poses significant challenges, due to
the need to model complex interactions between entities across sentences.
Current approaches, largely based on the ATLOP architecture, are commonly
evaluated on benchmarks like DocRED and Re-DocRED. However, their performance
in zero-shot or few-shot settings remains largely underexplored due to the
task's complexity. Recently, the GLiNER model has shown that a compact NER
model can outperform much larger Large Language Models. With a similar
motivation, we introduce GLiDRE, a new model for document-level relation
extraction that builds on the key ideas of GliNER. We benchmark GLiDRE against
state-of-the-art models across various data settings on the Re-DocRED dataset.
Our results demonstrate that GLiDRE achieves state-of-the-art performance in
few-shot scenarios. Our code is publicly available.
|
2025-08-01
|
{'id': '2508.00757', 'title': 'GLiDRE: Generalist Lightweight model for Document-level Relation Extraction', 'abstract': "Relation Extraction (RE) is a fundamental task in Natural Language\nProcessing, and its document-level variant poses significant challenges, due to\nthe need to model complex interactions between entities across sentences.\nCurrent approaches, largely based on the ATLOP architecture, are commonly\nevaluated on benchmarks like DocRED and Re-DocRED. However, their performance\nin zero-shot or few-shot settings remains largely underexplored due to the\ntask's complexity. Recently, the GLiNER model has shown that a compact NER\nmodel can outperform much larger Large Language Models. With a similar\nmotivation, we introduce GLiDRE, a new model for document-level relation\nextraction that builds on the key ideas of GliNER. We benchmark GLiDRE against\nstate-of-the-art models across various data settings on the Re-DocRED dataset.\nOur results demonstrate that GLiDRE achieves state-of-the-art performance in\nfew-shot scenarios. Our code is publicly available.", 'authors': ['Robin Armingaud', 'Romaric Besançon'], 'published': '2025-08-01 16:33:13+00:00', 'categories': ['cs.CL'], 'url': 'http://arxiv.org/abs/2508.00757v1'}
|
65
|
2508.00332
| 2,508.00332
| null |
Improving Multimodal Contrastive Learning of Sentence Embeddings with Object-Phrase Alignment
|
Multimodal sentence embedding models typically leverage image-caption pairs
in addition to textual data during training. However, such pairs often contain
noise, including redundant or irrelevant information on either the image or
caption side. To mitigate this issue, we propose MCSEO, a method that enhances
multimodal sentence embeddings by incorporating fine-grained object-phrase
alignment alongside traditional image-caption alignment. Specifically, MCSEO
utilizes existing segmentation and object detection models to extract accurate
object-phrase pairs, which are then used to optimize a contrastive learning
objective tailored to object-phrase correspondence. Experimental results on
semantic textual similarity (STS) tasks across different backbone models
demonstrate that MCSEO consistently outperforms strong baselines, highlighting
the significance of precise object-phrase alignment in multimodal
representation learning.
|
2025-08-01
|
{'id': '2508.00332', 'title': 'Improving Multimodal Contrastive Learning of Sentence Embeddings with Object-Phrase Alignment', 'abstract': 'Multimodal sentence embedding models typically leverage image-caption pairs\nin addition to textual data during training. However, such pairs often contain\nnoise, including redundant or irrelevant information on either the image or\ncaption side. To mitigate this issue, we propose MCSEO, a method that enhances\nmultimodal sentence embeddings by incorporating fine-grained object-phrase\nalignment alongside traditional image-caption alignment. Specifically, MCSEO\nutilizes existing segmentation and object detection models to extract accurate\nobject-phrase pairs, which are then used to optimize a contrastive learning\nobjective tailored to object-phrase correspondence. Experimental results on\nsemantic textual similarity (STS) tasks across different backbone models\ndemonstrate that MCSEO consistently outperforms strong baselines, highlighting\nthe significance of precise object-phrase alignment in multimodal\nrepresentation learning.', 'authors': ['Kaiyan Zhao', 'Zhongtao Miao', 'Yoshimasa Tsuruoka'], 'published': '2025-08-01 05:42:28+00:00', 'categories': ['cs.CL'], 'url': 'http://arxiv.org/abs/2508.00332v1'}
|
66
|
2508.00401
| 2,508.00401
| null |
Theory of Mind Using Active Inference: A Framework for Multi-Agent Cooperation
|
We present a novel approach to multi-agent cooperation by implementing theory
of mind (ToM) within active inference. ToM - the ability to understand that
others can have differing knowledge and goals - enables agents to reason about
others' beliefs while planning their own actions. Unlike previous active
inference approaches to multi-agent cooperation, our method neither relies on
task-specific shared generative models nor requires explicit communication,
while being generalisable. In our framework, the ToM-equipped agent maintains
distinct representations of its own and others' beliefs and goals. We extend
the sophisticated inference tree-based planning algorithm to systematically
explore joint policy spaces through recursive reasoning. Our approach is
evaluated through collision avoidance and foraging task simulations. Results
demonstrate that ToM-equipped agents cooperate better compared to non-ToM
counterparts by being able to avoid collisions and reduce redundant efforts.
Crucially, ToM agents accomplish this by inferring others' beliefs solely from
observable behaviour. This work advances practical applications in artificial
intelligence while providing computational insights into ToM.
|
2025-08-01
|
{'id': '2508.00401', 'title': 'Theory of Mind Using Active Inference: A Framework for Multi-Agent Cooperation', 'abstract': "We present a novel approach to multi-agent cooperation by implementing theory\nof mind (ToM) within active inference. ToM - the ability to understand that\nothers can have differing knowledge and goals - enables agents to reason about\nothers' beliefs while planning their own actions. Unlike previous active\ninference approaches to multi-agent cooperation, our method neither relies on\ntask-specific shared generative models nor requires explicit communication,\nwhile being generalisable. In our framework, the ToM-equipped agent maintains\ndistinct representations of its own and others' beliefs and goals. We extend\nthe sophisticated inference tree-based planning algorithm to systematically\nexplore joint policy spaces through recursive reasoning. Our approach is\nevaluated through collision avoidance and foraging task simulations. Results\ndemonstrate that ToM-equipped agents cooperate better compared to non-ToM\ncounterparts by being able to avoid collisions and reduce redundant efforts.\nCrucially, ToM agents accomplish this by inferring others' beliefs solely from\nobservable behaviour. This work advances practical applications in artificial\nintelligence while providing computational insights into ToM.", 'authors': ['Riddhi J. Pitliya', 'Ozan Catal', 'Toon Van de Maele', 'Corrado Pezzato', 'Tim Verbelen'], 'published': '2025-08-01 08:02:35+00:00', 'categories': ['cs.AI', 'cs.MA'], 'url': 'http://arxiv.org/abs/2508.00401v1'}
|
67
|
2508.00223
| 2,508.00223
| null |
Structural Causal Models for Extremes: an Approach Based on Exponent Measures
|
We introduce a new formulation of structural causal models for extremes,
called the extremal structural causal model (eSCM). Unlike conventional
structural causal models, where randomness is governed by a probability
distribution, eSCMs use an exponent measure--an infinite-mass law that
naturally arises in the analysis of multivariate extremes. Central to this
framework are activation variables, which abstract the single-big-jump
principle, along with additional randomization that enriches the class of eSCM
laws. This formulation encompasses all possible laws of directed graphical
models under the recently introduced notion of extremal conditional
independence. We also identify an inherent asymmetry in eSCMs under natural
assumptions, enabling the identifiability of causal directions, a central
challenge in causal inference. Finally, we propose a method that utilizes this
causal asymmetry and demonstrate its effectiveness in both simulated and real
datasets.
|
2025-08-01
|
{'id': '2508.00223', 'title': 'Structural Causal Models for Extremes: an Approach Based on Exponent Measures', 'abstract': 'We introduce a new formulation of structural causal models for extremes,\ncalled the extremal structural causal model (eSCM). Unlike conventional\nstructural causal models, where randomness is governed by a probability\ndistribution, eSCMs use an exponent measure--an infinite-mass law that\nnaturally arises in the analysis of multivariate extremes. Central to this\nframework are activation variables, which abstract the single-big-jump\nprinciple, along with additional randomization that enriches the class of eSCM\nlaws. This formulation encompasses all possible laws of directed graphical\nmodels under the recently introduced notion of extremal conditional\nindependence. We also identify an inherent asymmetry in eSCMs under natural\nassumptions, enabling the identifiability of causal directions, a central\nchallenge in causal inference. Finally, we propose a method that utilizes this\ncausal asymmetry and demonstrate its effectiveness in both simulated and real\ndatasets.', 'authors': ['Fei Fang', 'Shuyang Bai', 'Tiandong Wang'], 'published': '2025-08-01 00:01:23+00:00', 'categories': ['math.ST', 'stat.ME', 'stat.TH'], 'url': 'http://arxiv.org/abs/2508.00223v1'}
|
End of preview. Expand
in Data Studio
README.md exists but content is empty.
- Downloads last month
- 21
Size of downloaded dataset files:
72.4 MB
Size of the auto-converted Parquet files:
72.4 MB
Number of rows:
42,118