bibtex_url
stringlengths 41
50
| bibtext
stringlengths 693
2.88k
| abstract
stringlengths 0
2k
| authors
listlengths 1
45
| title
stringlengths 21
206
| id
stringlengths 7
16
| type
stringclasses 2
values | arxiv_id
stringlengths 9
12
⌀ |
|---|---|---|---|---|---|---|---|
https://aclanthology.org/2024.findings-acl.68.bib | @inproceedings{shafayat-etal-2024-benqa,
title = "{BE}n{QA}: A Question Answering Benchmark for {B}engali and {E}nglish",
author = "Shafayat, Sheikh and
Hasan, H and
Mahim, Minhajur and
Putri, Rifki and
Thorne, James and
Oh, Alice",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.68",
pages = "1158--1177",
abstract = "In this study, we introduce BEnQA, a dataset comprising parallel Bengali and English exam questions for middle and high school levels in Bangladesh. Our dataset consists of approximately 5K questions covering several subjects in science with different types of questions, including factual, application, and reasoning-based questions. We benchmark several Large Language Models (LLMs) with our parallel dataset and observe a notable performance disparity between the models in Bengali and English. We also investigate some prompting methods, and find that Chain-of-Thought prompting is beneficial mostly on reasoning questions, but not so much on factual ones. We also find that appending English translation helps to answer questions in Bengali. Our findings point to promising future research directions for improving the performance of LLMs in Bengali and more generally in low-resource languages.",
}
| In this study, we introduce BEnQA, a dataset comprising parallel Bengali and English exam questions for middle and high school levels in Bangladesh. Our dataset consists of approximately 5K questions covering several subjects in science with different types of questions, including factual, application, and reasoning-based questions. We benchmark several Large Language Models (LLMs) with our parallel dataset and observe a notable performance disparity between the models in Bengali and English. We also investigate some prompting methods, and find that Chain-of-Thought prompting is beneficial mostly on reasoning questions, but not so much on factual ones. We also find that appending English translation helps to answer questions in Bengali. Our findings point to promising future research directions for improving the performance of LLMs in Bengali and more generally in low-resource languages. | [
"Shafayat, Sheikh",
"Hasan, H",
"Mahim, Minhajur",
"Putri, Rifki",
"Thorne, James",
"Oh, Alice"
]
| {BE}n{QA}: A Question Answering Benchmark for {B}engali and {E}nglish | findings-acl.68 | Poster | 2403.10900v1 |
https://aclanthology.org/2024.findings-acl.69.bib | @inproceedings{cui-etal-2024-multi,
title = "{MORE}: Multi-m{O}dal {RE}trieval Augmented Generative Commonsense Reasoning",
author = "Cui, Wanqing and
Bi, Keping and
Guo, Jiafeng and
Cheng, Xueqi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.69",
pages = "1178--1192",
abstract = "Since commonsense information has been recorded significantly less frequently than its existence, language models pre-trained by text generation have difficulty to learn sufficient commonsense knowledge. Several studies have leveraged text retrieval to augment the models{'} commonsense ability. Unlike text, images capture commonsense information inherently but little effort has been paid to effectively utilize them. In this work, we propose a novel $\textbf{M}$ulti-m$\textbf{O}$dal $\textbf{RE}$trieval (MORE) augmentation framework, to leverage both text and images to enhance the commonsense ability of language models. Extensive experiments on the Common-Gen task have demonstrated the efficacy of MORE based on the pre-trained models of both single and multiple modalities.",
}
| Since commonsense information has been recorded significantly less frequently than its existence, language models pre-trained by text generation have difficulty to learn sufficient commonsense knowledge. Several studies have leveraged text retrieval to augment the models{'} commonsense ability. Unlike text, images capture commonsense information inherently but little effort has been paid to effectively utilize them. In this work, we propose a novel $\textbf{M}$ulti-m$\textbf{O}$dal $\textbf{RE}$trieval (MORE) augmentation framework, to leverage both text and images to enhance the commonsense ability of language models. Extensive experiments on the Common-Gen task have demonstrated the efficacy of MORE based on the pre-trained models of both single and multiple modalities. | [
"Cui, Wanqing",
"Bi, Keping",
"Guo, Jiafeng",
"Cheng, Xueqi"
]
| {MORE}: Multi-m{O}dal {RE}trieval Augmented Generative Commonsense Reasoning | findings-acl.69 | Poster | 2210.12887v1 |
https://aclanthology.org/2024.findings-acl.70.bib | @inproceedings{jin-etal-2024-cutting,
title = "Cutting Off the Head Ends the Conflict: A Mechanism for Interpreting and Mitigating Knowledge Conflicts in Language Models",
author = "Jin, Zhuoran and
Cao, Pengfei and
Yuan, Hongbang and
Chen, Yubo and
Xu, Jiexin and
Li, Huaijun and
Jiang, Xiaojian and
Liu, Kang and
Zhao, Jun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.70",
pages = "1193--1215",
abstract = "Recently, retrieval augmentation and tool augmentation have demonstrated a remarkable capability to expand the internal memory boundaries of language models (LMs) by providing external context. However, internal memory and external context inevitably clash, leading to knowledge conflicts within LMs. In this paper, we aim to interpret the mechanism of knowledge conflicts through the lens of information flow, and then mitigate conflicts by precise interventions at the pivotal point. We find there are some attention heads with opposite effects in the later layers, where memory heads can recall knowledge from internal memory, and context heads can retrieve knowledge from external context. Moreover, we reveal that the pivotal point at which knowledge conflicts emerge in LMs is the integration of inconsistent information flows by memory heads and context heads. Inspired by the insights, we propose a novel method called Pruning Head via PatH PatcHing (PH3), which can efficiently mitigate knowledge conflicts by pruning conflicting attention heads without updating model parameters. PH3 can flexibly control eight LMs to use internal memory ($\uparrow$ 44.0{\%}) or external context ($\uparrow$ 38.5{\%}). Moreover, PH3 can also improve the performance of LMs on open-domain QA tasks. We also conduct extensive experiments to demonstrate the cross-model, cross-relation, and cross-format generalization of our method. Our code is publicly available at https://github.com/jinzhuoran/MConflict/.",
}
| Recently, retrieval augmentation and tool augmentation have demonstrated a remarkable capability to expand the internal memory boundaries of language models (LMs) by providing external context. However, internal memory and external context inevitably clash, leading to knowledge conflicts within LMs. In this paper, we aim to interpret the mechanism of knowledge conflicts through the lens of information flow, and then mitigate conflicts by precise interventions at the pivotal point. We find there are some attention heads with opposite effects in the later layers, where memory heads can recall knowledge from internal memory, and context heads can retrieve knowledge from external context. Moreover, we reveal that the pivotal point at which knowledge conflicts emerge in LMs is the integration of inconsistent information flows by memory heads and context heads. Inspired by the insights, we propose a novel method called Pruning Head via PatH PatcHing (PH3), which can efficiently mitigate knowledge conflicts by pruning conflicting attention heads without updating model parameters. PH3 can flexibly control eight LMs to use internal memory ($\uparrow$ 44.0{\%}) or external context ($\uparrow$ 38.5{\%}). Moreover, PH3 can also improve the performance of LMs on open-domain QA tasks. We also conduct extensive experiments to demonstrate the cross-model, cross-relation, and cross-format generalization of our method. Our code is publicly available at https://github.com/jinzhuoran/MConflict/. | [
"Jin, Zhuoran",
"Cao, Pengfei",
"Yuan, Hongbang",
"Chen, Yubo",
"Xu, Jiexin",
"Li, Huaijun",
"Jiang, Xiaojian",
"Liu, Kang",
"Zhao, Jun"
]
| Cutting Off the Head Ends the Conflict: A Mechanism for Interpreting and Mitigating Knowledge Conflicts in Language Models | findings-acl.70 | Poster | 2402.18154v1 |
https://aclanthology.org/2024.findings-acl.71.bib | @inproceedings{pei-etal-2024-biot5,
title = "{B}io{T}5+: Towards Generalized Biological Understanding with {IUPAC} Integration and Multi-task Tuning",
author = "Pei, Qizhi and
Wu, Lijun and
Gao, Kaiyuan and
Liang, Xiaozhuan and
Fang, Yin and
Zhu, Jinhua and
Xie, Shufang and
Qin, Tao and
Yan, Rui",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.71",
pages = "1216--1240",
abstract = "Recent research trends in computational biology have increasingly focused on integrating text and bio-entity modeling, especially in the context of molecules and proteins. However, previous efforts like BioT5 faced challenges in generalizing across diverse tasks and lacked a nuanced understanding of molecular structures, particularly in their textual representations (e.g., IUPAC). This paper introduces BioT5+, an extension of the BioT5 framework, tailored to enhance biological research and drug discovery. BioT5+ incorporates several novel features: integration of IUPAC names for molecular understanding, inclusion of extensive bio-text and molecule data from sources like bioRxiv and PubChem, the multi-task instruction tuning for generality across tasks, and a numerical tokenization technique for improved processing of numerical data. These enhancements allow BioT5+ to bridge the gap between molecular representations and their textual descriptions, providing a more holistic understanding of biological entities, and largely improving the grounded reasoning of bio-text and bio-sequences. The model is pre-trained and fine-tuned with a large number of experiments, including \textit{3 types of problems (classification, regression, generation), 15 kinds of tasks, and 21 total benchmark datasets}, demonstrating the remarkable performance and state-of-the-art results in most cases. BioT5+ stands out for its ability to capture intricate relationships in biological data, thereby contributing significantly to bioinformatics and computational biology. Our code is available at https://github.com/QizhiPei/BioT5.",
}
| Recent research trends in computational biology have increasingly focused on integrating text and bio-entity modeling, especially in the context of molecules and proteins. However, previous efforts like BioT5 faced challenges in generalizing across diverse tasks and lacked a nuanced understanding of molecular structures, particularly in their textual representations (e.g., IUPAC). This paper introduces BioT5+, an extension of the BioT5 framework, tailored to enhance biological research and drug discovery. BioT5+ incorporates several novel features: integration of IUPAC names for molecular understanding, inclusion of extensive bio-text and molecule data from sources like bioRxiv and PubChem, the multi-task instruction tuning for generality across tasks, and a numerical tokenization technique for improved processing of numerical data. These enhancements allow BioT5+ to bridge the gap between molecular representations and their textual descriptions, providing a more holistic understanding of biological entities, and largely improving the grounded reasoning of bio-text and bio-sequences. The model is pre-trained and fine-tuned with a large number of experiments, including \textit{3 types of problems (classification, regression, generation), 15 kinds of tasks, and 21 total benchmark datasets}, demonstrating the remarkable performance and state-of-the-art results in most cases. BioT5+ stands out for its ability to capture intricate relationships in biological data, thereby contributing significantly to bioinformatics and computational biology. Our code is available at https://github.com/QizhiPei/BioT5. | [
"Pei, Qizhi",
"Wu, Lijun",
"Gao, Kaiyuan",
"Liang, Xiaozhuan",
"Fang, Yin",
"Zhu, Jinhua",
"Xie, Shufang",
"Qin, Tao",
"Yan, Rui"
]
| {B}io{T}5+: Towards Generalized Biological Understanding with {IUPAC} Integration and Multi-task Tuning | findings-acl.71 | Poster | 2402.17810v2 |
https://aclanthology.org/2024.findings-acl.72.bib | @inproceedings{wen-etal-2024-sibo,
title = "{SIBO}: A Simple Booster for Parameter-Efficient Fine-Tuning",
author = "Wen, Zhihao and
Zhang, Jie and
Fang, Yuan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.72",
pages = "1241--1257",
abstract = "Fine-tuning all parameters of large language models (LLMs) necessitates substantial computational power and extended time. Latest advancements in parameter-efficient fine-tuning (PEFT) techniques, such as Adapter tuning and LoRA, allow for adjustments to only a minor fraction of the parameters of these LLMs. Concurrently, it has been noted that the issue of over-smoothing diminishes the effectiveness of these Transformer-based LLMs, resulting in suboptimal performances in downstream tasks. In this paper, we present SIBO, which is a SImple BOoster to enhance PEFT, by injecting an initial residual. SIBO is straightforward and readily extensible to a range of state-of-the-art PEFT techniques to alleviate over-smoothing and enhance performance. Extensive experiments on 22 benchmark datasets demonstrate that SIBO significantly enhances the performance of various strong baselines, achieving up to 15.7{\%} and 23.5{\%} improvement over existing PEFT methods on the arithmetic and commonsense reasoning tasks, respectively.",
}
| Fine-tuning all parameters of large language models (LLMs) necessitates substantial computational power and extended time. Latest advancements in parameter-efficient fine-tuning (PEFT) techniques, such as Adapter tuning and LoRA, allow for adjustments to only a minor fraction of the parameters of these LLMs. Concurrently, it has been noted that the issue of over-smoothing diminishes the effectiveness of these Transformer-based LLMs, resulting in suboptimal performances in downstream tasks. In this paper, we present SIBO, which is a SImple BOoster to enhance PEFT, by injecting an initial residual. SIBO is straightforward and readily extensible to a range of state-of-the-art PEFT techniques to alleviate over-smoothing and enhance performance. Extensive experiments on 22 benchmark datasets demonstrate that SIBO significantly enhances the performance of various strong baselines, achieving up to 15.7{\%} and 23.5{\%} improvement over existing PEFT methods on the arithmetic and commonsense reasoning tasks, respectively. | [
"Wen, Zhihao",
"Zhang, Jie",
"Fang, Yuan"
]
| {SIBO}: A Simple Booster for Parameter-Efficient Fine-Tuning | findings-acl.72 | Poster | 2011.02022v2 |
https://aclanthology.org/2024.findings-acl.73.bib | @inproceedings{zhang-etal-2024-geoeval,
title = "{G}eo{E}val: Benchmark for Evaluating {LLM}s and Multi-Modal Models on Geometry Problem-Solving",
author = "Zhang, Jiaxin and
Li, Zhong-Zhi and
Zhang, Ming-Liang and
Yin, Fei and
Liu, Cheng-Lin and
Moshfeghi, Yashar",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.73",
pages = "1258--1276",
abstract = "Recent advancements in large language models (LLMs) and multi-modal models (MMs) have demonstrated their remarkable capabilities in problem-solving. Yet, their proficiency in tackling geometry math problems, which necessitates an integrated understanding of both textual and visual information, has not been thoroughly evaluated. To address this gap, we introduce the GeoEval benchmark, a comprehensive collection that includes a main subset of 2,000 problems, a 750 problems subset focusing on backward reasoning, an augmented sub- set of 2,000 problems, and a hard subset of 300 problems. This benchmark facilitates a deeper investigation into the performance of LLMs and MMs in solving geometry math problems. Our evaluation of ten LLMs and MMs across these varied subsets reveals that the WizardMath model excels, achieving a 55.67{\%} accuracy rate on the main subset but only a 6.00{\%} accuracy on the hard subset. This highlights the critical need for testing models against datasets on which they have not been pre-trained. Additionally, our findings indicate that GPT-series models perform more effectively on problems they have rephrased, suggesting a promising method for enhancing model capabilities.",
}
| Recent advancements in large language models (LLMs) and multi-modal models (MMs) have demonstrated their remarkable capabilities in problem-solving. Yet, their proficiency in tackling geometry math problems, which necessitates an integrated understanding of both textual and visual information, has not been thoroughly evaluated. To address this gap, we introduce the GeoEval benchmark, a comprehensive collection that includes a main subset of 2,000 problems, a 750 problems subset focusing on backward reasoning, an augmented sub- set of 2,000 problems, and a hard subset of 300 problems. This benchmark facilitates a deeper investigation into the performance of LLMs and MMs in solving geometry math problems. Our evaluation of ten LLMs and MMs across these varied subsets reveals that the WizardMath model excels, achieving a 55.67{\%} accuracy rate on the main subset but only a 6.00{\%} accuracy on the hard subset. This highlights the critical need for testing models against datasets on which they have not been pre-trained. Additionally, our findings indicate that GPT-series models perform more effectively on problems they have rephrased, suggesting a promising method for enhancing model capabilities. | [
"Zhang, Jiaxin",
"Li, Zhong-Zhi",
"Zhang, Ming-Liang",
"Yin, Fei",
"Liu, Cheng-Lin",
"Moshfeghi, Yashar"
]
| {G}eo{E}val: Benchmark for Evaluating {LLM}s and Multi-Modal Models on Geometry Problem-Solving | findings-acl.73 | Poster | 2309.04369v1 |
https://aclanthology.org/2024.findings-acl.74.bib | @inproceedings{wang-etal-2024-boosting,
title = "Boosting Textural {NER} with Synthetic Image and Instructive Alignment",
author = "Wang, Jiahao and
Ke, Wenjun and
Wang, Peng and
Zhang, Hang and
Nie, Dong and
Liu, Jiajun and
Li, Guozheng and
Shang, Ziyu",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.74",
pages = "1277--1287",
abstract = "Named entity recognition (NER) is a pivotal task reliant on textual data, often impeding the disambiguation of entities due to the absence of context. To tackle this challenge, conventional methods often incorporate images crawled from the internet as auxiliary information. However, the images often lack sufficient entities or would introduce noise. Even with high-quality images, it is still challenging to efficiently use images as auxiliaries (i.e., fine-grained alignment with texts). We introduce a novel method named InstructNER to address these issues. Leveraging the rich real-world knowledge and image synthesis capabilities of a large pre-trained stable diffusion (SD) model, InstructNER transforms the text-only NER into a multimodal NER (MNER) task. A selection process automatically identifies the best synthetic image by comparing fine-grained similarities with internet-crawled images through a visual bag-of-words strategy. Note, during the image synthesis, a cross-attention matrix between synthetic images and raw text emerges, which inspires a soft attention guidance alignment (AGA) mechanism. AGA optimizes the MNER task and concurrently facilitates instructive alignment in MNER. Empirical experiments on prominent MNER datasets show that our method surpasses all text-only baselines, improving F1-score by 1.4{\%} to 2.3{\%}. Remarkably, even when compared to fully multimodal baselines, our approach maintains competitive. Furthermore, we open-source a comprehensive synthetic image dataset and the code to supplement existing raw dataset. The code and datasets are available in https://github.com/Heyest/InstructNER.",
}
| Named entity recognition (NER) is a pivotal task reliant on textual data, often impeding the disambiguation of entities due to the absence of context. To tackle this challenge, conventional methods often incorporate images crawled from the internet as auxiliary information. However, the images often lack sufficient entities or would introduce noise. Even with high-quality images, it is still challenging to efficiently use images as auxiliaries (i.e., fine-grained alignment with texts). We introduce a novel method named InstructNER to address these issues. Leveraging the rich real-world knowledge and image synthesis capabilities of a large pre-trained stable diffusion (SD) model, InstructNER transforms the text-only NER into a multimodal NER (MNER) task. A selection process automatically identifies the best synthetic image by comparing fine-grained similarities with internet-crawled images through a visual bag-of-words strategy. Note, during the image synthesis, a cross-attention matrix between synthetic images and raw text emerges, which inspires a soft attention guidance alignment (AGA) mechanism. AGA optimizes the MNER task and concurrently facilitates instructive alignment in MNER. Empirical experiments on prominent MNER datasets show that our method surpasses all text-only baselines, improving F1-score by 1.4{\%} to 2.3{\%}. Remarkably, even when compared to fully multimodal baselines, our approach maintains competitive. Furthermore, we open-source a comprehensive synthetic image dataset and the code to supplement existing raw dataset. The code and datasets are available in https://github.com/Heyest/InstructNER. | [
"Wang, Jiahao",
"Ke, Wenjun",
"Wang, Peng",
"Zhang, Hang",
"Nie, Dong",
"Liu, Jiajun",
"Li, Guozheng",
"Shang, Ziyu"
]
| Boosting Textural {NER} with Synthetic Image and Instructive Alignment | findings-acl.74 | Poster | 2401.13545v1 |
https://aclanthology.org/2024.findings-acl.75.bib | @inproceedings{voita-etal-2024-neurons,
title = "Neurons in Large Language Models: Dead, N-gram, Positional",
author = "Voita, Elena and
Ferrando, Javier and
Nalmpantis, Christoforos",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.75",
pages = "1288--1301",
abstract = "We analyze a family of large language models in such a lightweight manner that can be done on a single GPU. Specifically, we focus on the OPT family of models ranging from 125m to 66b parameters and rely only on whether an FFN neuron is activated or not. First, we find that the early part of the network is sparse and represents many discrete features. Here, many neurons (more than in some layers of the 66b model) are {``}dead{''}, i.e. they never activate on a large collection of diverse data. At the same time, many of the alive neurons are reserved for discrete features and act as token and n-gram detectors. Interestingly, their corresponding FFN updates not only promote next token candidates as could be expected, but also explicitly focus on removing the information about triggering them tokens, i.e., current input. To the best of our knowledge, this is the first example of mechanisms specialized at removing (rather than adding) information from the residual stream. With scale, models become more sparse in a sense that they have more dead neurons and token detectors. Finally, some neurons are positional: them being activated or not depends largely (or solely) on position and less so (or not at all) on textual data. We find that smaller models have sets of neurons acting as position range indicators while larger models operate in a less explicit manner.",
}
| We analyze a family of large language models in such a lightweight manner that can be done on a single GPU. Specifically, we focus on the OPT family of models ranging from 125m to 66b parameters and rely only on whether an FFN neuron is activated or not. First, we find that the early part of the network is sparse and represents many discrete features. Here, many neurons (more than in some layers of the 66b model) are {``}dead{''}, i.e. they never activate on a large collection of diverse data. At the same time, many of the alive neurons are reserved for discrete features and act as token and n-gram detectors. Interestingly, their corresponding FFN updates not only promote next token candidates as could be expected, but also explicitly focus on removing the information about triggering them tokens, i.e., current input. To the best of our knowledge, this is the first example of mechanisms specialized at removing (rather than adding) information from the residual stream. With scale, models become more sparse in a sense that they have more dead neurons and token detectors. Finally, some neurons are positional: them being activated or not depends largely (or solely) on position and less so (or not at all) on textual data. We find that smaller models have sets of neurons acting as position range indicators while larger models operate in a less explicit manner. | [
"Voita, Elena",
"Ferr",
"o, Javier",
"Nalmpantis, Christoforos"
]
| Neurons in Large Language Models: Dead, N-gram, Positional | findings-acl.75 | Poster | 2309.04827v1 |
https://aclanthology.org/2024.findings-acl.76.bib | @inproceedings{li-etal-2024-llms,
title = "{LLM}s as Bridges: Reformulating Grounded Multimodal Named Entity Recognition",
author = "Li, Jinyuan and
Li, Han and
Sun, Di and
Wang, Jiahao and
Zhang, Wenkun and
Wang, Zan and
Pan, Gang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.76",
pages = "1302--1318",
abstract = "Grounded Multimodal Named Entity Recognition (GMNER) is a nascent multimodal task that aims to identify named entities, entity types and their corresponding visual regions. GMNER task exhibits two challenging properties: 1) The weak correlation between image-text pairs in social media results in a significant portion of named entities being ungroundable. 2) There exists a distinction between coarse-grained referring expressions commonly used in similar tasks (e.g., phrase localization, referring expression comprehension) and fine-grained named entities. In this paper, we propose RiVEG, a unified framework that reformulates GMNER into a joint MNER-VE-VG task by leveraging large language models (LLMs) as a connecting bridge. This reformulation brings two benefits: 1) It maintains the optimal MNER performance and eliminates the need for employing object detection methods to pre-extract regional features, thereby naturally addressing two major limitations of existing GMNER methods. 2) The introduction of entity expansion expression and Visual Entailment (VE) module unifies Visual Grounding (VG) and Entity Grounding (EG). It enables RiVEG to effortlessly inherit the Visual Entailment and Visual Grounding capabilities of any current or prospective multimodal pretraining models. Extensive experiments demonstrate that RiVEG outperforms state-of-the-art methods on the existing GMNER dataset and achieves absolute leads of 10.65{\%}, 6.21{\%}, and 8.83{\%} in all three subtasks.",
}
| Grounded Multimodal Named Entity Recognition (GMNER) is a nascent multimodal task that aims to identify named entities, entity types and their corresponding visual regions. GMNER task exhibits two challenging properties: 1) The weak correlation between image-text pairs in social media results in a significant portion of named entities being ungroundable. 2) There exists a distinction between coarse-grained referring expressions commonly used in similar tasks (e.g., phrase localization, referring expression comprehension) and fine-grained named entities. In this paper, we propose RiVEG, a unified framework that reformulates GMNER into a joint MNER-VE-VG task by leveraging large language models (LLMs) as a connecting bridge. This reformulation brings two benefits: 1) It maintains the optimal MNER performance and eliminates the need for employing object detection methods to pre-extract regional features, thereby naturally addressing two major limitations of existing GMNER methods. 2) The introduction of entity expansion expression and Visual Entailment (VE) module unifies Visual Grounding (VG) and Entity Grounding (EG). It enables RiVEG to effortlessly inherit the Visual Entailment and Visual Grounding capabilities of any current or prospective multimodal pretraining models. Extensive experiments demonstrate that RiVEG outperforms state-of-the-art methods on the existing GMNER dataset and achieves absolute leads of 10.65{\%}, 6.21{\%}, and 8.83{\%} in all three subtasks. | [
"Li, Jinyuan",
"Li, Han",
"Sun, Di",
"Wang, Jiahao",
"Zhang, Wenkun",
"Wang, Zan",
"Pan, Gang"
]
| {LLM}s as Bridges: Reformulating Grounded Multimodal Named Entity Recognition | findings-acl.76 | Poster | 2402.09989v4 |
https://aclanthology.org/2024.findings-acl.77.bib | @inproceedings{laosaengpha-etal-2024-learning,
title = "Learning Job Title Representation from Job Description Aggregation Network",
author = "Laosaengpha, Napat and
Tativannarat, Thanit and
Piansaddhayanon, Chawan and
Rutherford, Attapol and
Chuangsuwanich, Ekapol",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.77",
pages = "1319--1329",
abstract = "Learning job title representation is a vital process for developing automatic human resource tools. To do so, existing methods primarily rely on learning the title representation through skills extracted from the job description, neglecting the rich and diverse content within. Thus, we propose an alternative framework for learning job titles through their respective job description (JD) and utilize a Job Description Aggregator component to handle the lengthy description and bidirectional contrastive loss to account for the bidirectional relationship between the job title and its description. We evaluated the performance of our method on both in-domain and out-of-domain settings, achieving a superior performance over the skill-based approach.",
}
| Learning job title representation is a vital process for developing automatic human resource tools. To do so, existing methods primarily rely on learning the title representation through skills extracted from the job description, neglecting the rich and diverse content within. Thus, we propose an alternative framework for learning job titles through their respective job description (JD) and utilize a Job Description Aggregator component to handle the lengthy description and bidirectional contrastive loss to account for the bidirectional relationship between the job title and its description. We evaluated the performance of our method on both in-domain and out-of-domain settings, achieving a superior performance over the skill-based approach. | [
"Laosaengpha, Napat",
"Tativannarat, Thanit",
"Piansaddhayanon, Chawan",
"Rutherford, Attapol",
"Chuangsuwanich, Ekapol"
]
| Learning Job Title Representation from Job Description Aggregation Network | findings-acl.77 | Poster | 2406.08055v1 |
https://aclanthology.org/2024.findings-acl.78.bib | @inproceedings{singh-etal-2024-flowvqa,
title = "{F}low{VQA}: Mapping Multimodal Logic in Visual Question Answering with Flowcharts",
author = "Singh, Shubhankar and
Chaurasia, Purvi and
Varun, Yerram and
Pandya, Pranshu and
Gupta, Vatsal and
Gupta, Vivek and
Roth, Dan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.78",
pages = "1330--1350",
abstract = "Existing benchmarks for visual question answering lack in visual grounding and complexity, particularly in evaluating spatial reasoning skills. We introduce FlowVQA, a novel benchmark aimed at assessing the capabilities of visual question-answering multimodal language models in reasoning with flowcharts as visual contexts. FlowVQA comprises 2,272 carefully generated and human-verified flowchart images from three distinct content sources, along with 22,413 diverse question-answer pairs, to test a spectrum of reasoning tasks, including information localization, decision-making, and logical progression. We conduct a thorough baseline evaluation on a suite of both open-source and proprietary multimodal language models using various strategies, followed by an analysis of directional bias. The results underscore the benchmark{'}s potential as a vital tool for advancing the field of multimodal modeling, providing a focused and challenging environment for enhancing model performance in visual and logical reasoning tasks.",
}
| Existing benchmarks for visual question answering lack in visual grounding and complexity, particularly in evaluating spatial reasoning skills. We introduce FlowVQA, a novel benchmark aimed at assessing the capabilities of visual question-answering multimodal language models in reasoning with flowcharts as visual contexts. FlowVQA comprises 2,272 carefully generated and human-verified flowchart images from three distinct content sources, along with 22,413 diverse question-answer pairs, to test a spectrum of reasoning tasks, including information localization, decision-making, and logical progression. We conduct a thorough baseline evaluation on a suite of both open-source and proprietary multimodal language models using various strategies, followed by an analysis of directional bias. The results underscore the benchmark{'}s potential as a vital tool for advancing the field of multimodal modeling, providing a focused and challenging environment for enhancing model performance in visual and logical reasoning tasks. | [
"Singh, Shubhankar",
"Chaurasia, Purvi",
"Varun, Yerram",
"P",
"ya, Pranshu",
"Gupta, Vatsal",
"Gupta, Vivek",
"Roth, Dan"
]
| {F}low{VQA}: Mapping Multimodal Logic in Visual Question Answering with Flowcharts | findings-acl.78 | Poster | 2406.19237v2 |
https://aclanthology.org/2024.findings-acl.79.bib | @inproceedings{yu-etal-2024-flexible,
title = "Flexible Weight Tuning and Weight Fusion Strategies for Continual Named Entity Recognition",
author = "Yu, Yahan and
Zhang, Duzhen and
Chen, Xiuyi and
Chu, Chenhui",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.79",
pages = "1351--1358",
abstract = "Continual Named Entity Recognition (CNER) is dedicated to sequentially learning new entity types while mitigating catastrophic forgetting of old entity types. Traditional CNER approaches commonly employ knowledge distillation to retain old knowledge within the current model. However, because only the representations of old and new models are constrained to be consistent, the reliance solely on distillation in existing methods still suffers from catastrophic forgetting. To further alleviate the forgetting issue of old entity types, this paper introduces flexible Weight Tuning (WT) and Weight Fusion (WF) strategies for CNER. The WT strategy, applied at each training step, employs a learning rate schedule on the parameters of the current model. After learning the current task, the WF strategy dynamically integrates knowledge from both the current and previous models for inference. Notably, these two strategies are model-agnostic and seamlessly integrate with existing State-Of-The-Art (SOTA) models. Extensive experiments demonstrate that the WT and WF strategies consistently enhance the performance of previous SOTA methods across ten CNER settings in three datasets.",
}
| Continual Named Entity Recognition (CNER) is dedicated to sequentially learning new entity types while mitigating catastrophic forgetting of old entity types. Traditional CNER approaches commonly employ knowledge distillation to retain old knowledge within the current model. However, because only the representations of old and new models are constrained to be consistent, the reliance solely on distillation in existing methods still suffers from catastrophic forgetting. To further alleviate the forgetting issue of old entity types, this paper introduces flexible Weight Tuning (WT) and Weight Fusion (WF) strategies for CNER. The WT strategy, applied at each training step, employs a learning rate schedule on the parameters of the current model. After learning the current task, the WF strategy dynamically integrates knowledge from both the current and previous models for inference. Notably, these two strategies are model-agnostic and seamlessly integrate with existing State-Of-The-Art (SOTA) models. Extensive experiments demonstrate that the WT and WF strategies consistently enhance the performance of previous SOTA methods across ten CNER settings in three datasets. | [
"Yu, Yahan",
"Zhang, Duzhen",
"Chen, Xiuyi",
"Chu, Chenhui"
]
| Flexible Weight Tuning and Weight Fusion Strategies for Continual Named Entity Recognition | findings-acl.79 | Poster | 2308.08625v1 |
https://aclanthology.org/2024.findings-acl.80.bib | @inproceedings{chen-etal-2024-unveiling,
title = "Unveiling the Achilles{'} Heel of {NLG} Evaluators: A Unified Adversarial Framework Driven by Large Language Models",
author = "Chen, Yiming and
Zhang, Chen and
Luo, Danqing and
D{'}Haro, Luis Fernando and
Tan, Robby and
Li, Haizhou",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.80",
pages = "1359--1375",
abstract = "The automatic evaluation of natural language generation (NLG) systems presents a long-lasting challenge. Recent studies have highlighted various neural metrics that align well with human evaluations. Yet, the robustness of these evaluators against adversarial perturbations remains largely under-explored due to the unique challenges in obtaining adversarial data for different NLG evaluation tasks. To address the problem, we introduce AdvEval, a novel black-box adversarial framework against NLG evaluators. AdvEval is specially tailored to generate data that yield strong disagreements between human and victim evaluators. Specifically, inspired by the recent success of large language models (LLMs) in text generation and evaluation, we adopt strong LLMs as both the data generator and gold evaluator. Adversarial data are automatically optimized with feedback from the gold and victim evaluator. We conduct experiments on 12 victim evaluators and 11 NLG datasets, spanning tasks including dialogue, summarization, and question evaluation. The results show that AdvEval can lead to significant performance degradation of various victim metrics, thereby validating its efficacy.",
}
| The automatic evaluation of natural language generation (NLG) systems presents a long-lasting challenge. Recent studies have highlighted various neural metrics that align well with human evaluations. Yet, the robustness of these evaluators against adversarial perturbations remains largely under-explored due to the unique challenges in obtaining adversarial data for different NLG evaluation tasks. To address the problem, we introduce AdvEval, a novel black-box adversarial framework against NLG evaluators. AdvEval is specially tailored to generate data that yield strong disagreements between human and victim evaluators. Specifically, inspired by the recent success of large language models (LLMs) in text generation and evaluation, we adopt strong LLMs as both the data generator and gold evaluator. Adversarial data are automatically optimized with feedback from the gold and victim evaluator. We conduct experiments on 12 victim evaluators and 11 NLG datasets, spanning tasks including dialogue, summarization, and question evaluation. The results show that AdvEval can lead to significant performance degradation of various victim metrics, thereby validating its efficacy. | [
"Chen, Yiming",
"Zhang, Chen",
"Luo, Danqing",
"D{'}Haro, Luis Fern",
"o",
"Tan, Robby",
"Li, Haizhou"
]
| Unveiling the Achilles{'} Heel of {NLG} Evaluators: A Unified Adversarial Framework Driven by Large Language Models | findings-acl.80 | Poster | 2405.14646v1 |
https://aclanthology.org/2024.findings-acl.81.bib | @inproceedings{liusie-etal-2024-teacher,
title = "Teacher-Student Training for Debiasing: General Permutation Debiasing for Large Language Models",
author = "Liusie, Adian and
Fathullah, Yassir and
Gales, Mark",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.81",
pages = "1376--1387",
abstract = "Large Language Models (LLMs) have demonstrated impressive zero-shot capabilities and versatility in NLP tasks, however they sometimes fail to maintain crucial invariances for specific tasks. One example is permutation sensitivity, where LLMs{'} outputs may significantly vary depending on the order of the input options. While debiasing techniques can mitigate these issues, and yield better performance and reliability, they often come with a high computational cost at inference. This paper addresses this inefficiency at inference time. The aim is to distill the capabilities of a computationally intensive, debiased, teacher model into a more compact student model. We explore two variants of student models: one based on pure distillation, and the other on an error-correction approach for more complex tasks, where the student corrects a single biased decision from the teacher to achieve a debiased output. Our approach is general and can be applied to both black-box and white-box LLMs. Furthermore, we demonstrate that our compact, encoder-only student models can outperform their larger, biased teacher counterparts, achieving better results with significantly fewer parameters.",
}
| Large Language Models (LLMs) have demonstrated impressive zero-shot capabilities and versatility in NLP tasks, however they sometimes fail to maintain crucial invariances for specific tasks. One example is permutation sensitivity, where LLMs{'} outputs may significantly vary depending on the order of the input options. While debiasing techniques can mitigate these issues, and yield better performance and reliability, they often come with a high computational cost at inference. This paper addresses this inefficiency at inference time. The aim is to distill the capabilities of a computationally intensive, debiased, teacher model into a more compact student model. We explore two variants of student models: one based on pure distillation, and the other on an error-correction approach for more complex tasks, where the student corrects a single biased decision from the teacher to achieve a debiased output. Our approach is general and can be applied to both black-box and white-box LLMs. Furthermore, we demonstrate that our compact, encoder-only student models can outperform their larger, biased teacher counterparts, achieving better results with significantly fewer parameters. | [
"Liusie, Adian",
"Fathullah, Yassir",
"Gales, Mark"
]
| Teacher-Student Training for Debiasing: General Permutation Debiasing for Large Language Models | findings-acl.81 | Poster | 2403.13590v1 |
https://aclanthology.org/2024.findings-acl.82.bib | @inproceedings{pang-etal-2024-uncovering,
title = "Uncovering Limitations of Large Language Models in Information Seeking from Tables",
author = "Pang, Chaoxu and
Cao, Yixuan and
Yang, Chunhao and
Luo, Ping",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.82",
pages = "1388--1409",
abstract = "Tables are recognized for their high information density and widespread usage, serving as essential sources of information. Seeking information from tables (TIS) is a crucial capability for Large Language Models (LLMs), serving as the foundation of knowledge-based Q{\&}A systems. However, this field presently suffers from an absence of thorough and reliable evaluation. This paper introduces a more reliable benchmark for Table Information Seeking (TabIS). To avoid the unreliable evaluation caused by text similarity-based metrics, TabIS adopts a single-choice question format (with two options per question) instead of a text generation format. We establish an effective pipeline for generating options, ensuring their difficulty and quality. Experiments conducted on 12 LLMs reveal that while the performance of GPT-4-turbo is marginally satisfactory, both other proprietary and open-source models perform inadequately. Further analysis shows that LLMs exhibit a poor understanding of table structures, and struggle to balance between TIS performance and robustness against pseudo-relevant tables (common in retrieval-augmented systems). These findings uncover the limitations and potential challenges of LLMs in seeking information from tables. We release our data and code to facilitate further research in this field.",
}
| Tables are recognized for their high information density and widespread usage, serving as essential sources of information. Seeking information from tables (TIS) is a crucial capability for Large Language Models (LLMs), serving as the foundation of knowledge-based Q{\&}A systems. However, this field presently suffers from an absence of thorough and reliable evaluation. This paper introduces a more reliable benchmark for Table Information Seeking (TabIS). To avoid the unreliable evaluation caused by text similarity-based metrics, TabIS adopts a single-choice question format (with two options per question) instead of a text generation format. We establish an effective pipeline for generating options, ensuring their difficulty and quality. Experiments conducted on 12 LLMs reveal that while the performance of GPT-4-turbo is marginally satisfactory, both other proprietary and open-source models perform inadequately. Further analysis shows that LLMs exhibit a poor understanding of table structures, and struggle to balance between TIS performance and robustness against pseudo-relevant tables (common in retrieval-augmented systems). These findings uncover the limitations and potential challenges of LLMs in seeking information from tables. We release our data and code to facilitate further research in this field. | [
"Pang, Chaoxu",
"Cao, Yixuan",
"Yang, Chunhao",
"Luo, Ping"
]
| Uncovering Limitations of Large Language Models in Information Seeking from Tables | findings-acl.82 | Poster | 2406.04113v1 |
https://aclanthology.org/2024.findings-acl.83.bib | @inproceedings{zhou-etal-2024-ensemble,
title = "An Ensemble-of-Experts Framework for Rehearsal-free Continual Relation Extraction",
author = "Zhou, Shen and
Li, Yongqi and
Miao, Xin and
Qian, Tieyun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.83",
pages = "1410--1423",
abstract = "Continual relation extraction (CRE) aims to continuously learn relations in new tasks without forgetting old relations in previous tasks.Current CRE methods are all rehearsal-based which need to store samples and thus may encounter privacy and security issues.This paper targets rehearsal-free continual relation extraction for the first time and decomposes it into task identification and within-task prediction sub-problems. Existing rehearsal-free methods focus on training a model (expert) for within-task prediction yet neglect to enhance models{'} capability of task identification.In this paper, we propose an Ensemble-of-Experts (EoE) framework for rehearsal-free continual relation extraction. Specifically, we first discriminatively train each expert by augmenting analogous relations across tasks to enhance the expert{'}s task identification ability. We then propose a cascade voting mechanism to form an ensemble of experts for effectively aggregating their abilities.Extensive experiments demonstrate that our method outperforms current rehearsal-free methods and is even better than rehearsal-based CRE methods.",
}
| Continual relation extraction (CRE) aims to continuously learn relations in new tasks without forgetting old relations in previous tasks.Current CRE methods are all rehearsal-based which need to store samples and thus may encounter privacy and security issues.This paper targets rehearsal-free continual relation extraction for the first time and decomposes it into task identification and within-task prediction sub-problems. Existing rehearsal-free methods focus on training a model (expert) for within-task prediction yet neglect to enhance models{'} capability of task identification.In this paper, we propose an Ensemble-of-Experts (EoE) framework for rehearsal-free continual relation extraction. Specifically, we first discriminatively train each expert by augmenting analogous relations across tasks to enhance the expert{'}s task identification ability. We then propose a cascade voting mechanism to form an ensemble of experts for effectively aggregating their abilities.Extensive experiments demonstrate that our method outperforms current rehearsal-free methods and is even better than rehearsal-based CRE methods. | [
"Zhou, Shen",
"Li, Yongqi",
"Miao, Xin",
"Qian, Tieyun"
]
| An Ensemble-of-Experts Framework for Rehearsal-free Continual Relation Extraction | findings-acl.83 | Poster | 2305.03503v1 |
https://aclanthology.org/2024.findings-acl.84.bib | @inproceedings{wenzel-jatowt-2024-temporal,
title = "Temporal Validity Change Prediction",
author = "Wenzel, Georg and
Jatowt, Adam",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.84",
pages = "1424--1446",
abstract = "Temporal validity is an important property of text that has many downstream applications, such as recommender systems, conversational AI, and user status tracking. Existing benchmarking tasks often require models to identify the temporal validity duration of a single statement. However, many data sources contain additional context, such as successive sentences in a story or posts on a social media profile. This context may alter the duration for which the originally collected statement is expected to be valid. We propose Temporal Validity Change Prediction, a natural language processing task benchmarking the capability of machine learning models to detect context statements that induce such change. We create a dataset consisting of temporal target statements sourced from Twitter and crowdsource corresponding context statements. We then benchmark a set of transformer-based language models on our dataset. Finally, we experiment with a multitasking approach to improve the state-of-the-art performance.",
}
| Temporal validity is an important property of text that has many downstream applications, such as recommender systems, conversational AI, and user status tracking. Existing benchmarking tasks often require models to identify the temporal validity duration of a single statement. However, many data sources contain additional context, such as successive sentences in a story or posts on a social media profile. This context may alter the duration for which the originally collected statement is expected to be valid. We propose Temporal Validity Change Prediction, a natural language processing task benchmarking the capability of machine learning models to detect context statements that induce such change. We create a dataset consisting of temporal target statements sourced from Twitter and crowdsource corresponding context statements. We then benchmark a set of transformer-based language models on our dataset. Finally, we experiment with a multitasking approach to improve the state-of-the-art performance. | [
"Wenzel, Georg",
"Jatowt, Adam"
]
| Temporal Validity Change Prediction | findings-acl.84 | Poster | 2401.00779v1 |
https://aclanthology.org/2024.findings-acl.85.bib | @inproceedings{najafi-fyshe-2024-riff,
title = "{RIFF}: Learning to Rephrase Inputs for Few-shot Fine-tuning of Language Models",
author = "Najafi, Saeed and
Fyshe, Alona",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.85",
pages = "1447--1466",
abstract = "Pre-trained Language Models (PLMs) can be accurately fine-tuned for downstream text processing tasks. Recently, researchers have introduced several parameter-efficient fine-tuning methods that optimize input prompts or adjust a small number of model parameters (e.g LoRA). In this study, we explore the impact of altering the input text of the original task in conjunction with parameter-efficient fine-tuning methods. To most effectively rewrite the input text, we train a few-shot paraphrase model with a Maximum-Marginal Likelihood objective. Using six few-shot text classification datasets, we show that enriching data with paraphrases at train and test time enhances the performance beyond what can be achieved with parameter-efficient fine-tuning alone. The code used for our experiments can be found at https://github.com/SaeedNajafi/RIFF.",
}
| Pre-trained Language Models (PLMs) can be accurately fine-tuned for downstream text processing tasks. Recently, researchers have introduced several parameter-efficient fine-tuning methods that optimize input prompts or adjust a small number of model parameters (e.g LoRA). In this study, we explore the impact of altering the input text of the original task in conjunction with parameter-efficient fine-tuning methods. To most effectively rewrite the input text, we train a few-shot paraphrase model with a Maximum-Marginal Likelihood objective. Using six few-shot text classification datasets, we show that enriching data with paraphrases at train and test time enhances the performance beyond what can be achieved with parameter-efficient fine-tuning alone. The code used for our experiments can be found at https://github.com/SaeedNajafi/RIFF. | [
"Najafi, Saeed",
"Fyshe, Alona"
]
| {RIFF}: Learning to Rephrase Inputs for Few-shot Fine-tuning of Language Models | findings-acl.85 | Poster | 2408.03732v1 |
https://aclanthology.org/2024.findings-acl.86.bib | @inproceedings{kteich-etal-2024-modelling,
title = "Modelling Commonsense Commonalities with Multi-Facet Concept Embeddings",
author = "Kteich, Hanane and
Li, Na and
Chatterjee, Usashi and
Bouraoui, Zied and
Schockaert, Steven",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.86",
pages = "1467--1480",
abstract = "Concept embeddings offer a practical and efficient mechanism for injecting commonsense knowledge into downstream tasks. Their core purpose is often not to predict the commonsense properties of concepts themselves, but rather to identify commonalities, i.e. sets of concepts which share some property of interest. Such commonalities are the basis for inductive generalisation, hence high-quality concept embeddings can make learning easier and more robust. Unfortunately, standard embeddings primarily reflect basic taxonomic categories, making them unsuitable for finding commonalities that refer to more specific aspects (e.g. the colour of objects or the materials they are made of). In this paper, we address this limitation by explicitly modelling the different facets of interest when learning concept embeddings. We show that this leads to embeddings which capture a more diverse range of commonsense properties, and consistently improves results in downstream tasks such as ultra-fine entity typing and ontology completion.",
}
| Concept embeddings offer a practical and efficient mechanism for injecting commonsense knowledge into downstream tasks. Their core purpose is often not to predict the commonsense properties of concepts themselves, but rather to identify commonalities, i.e. sets of concepts which share some property of interest. Such commonalities are the basis for inductive generalisation, hence high-quality concept embeddings can make learning easier and more robust. Unfortunately, standard embeddings primarily reflect basic taxonomic categories, making them unsuitable for finding commonalities that refer to more specific aspects (e.g. the colour of objects or the materials they are made of). In this paper, we address this limitation by explicitly modelling the different facets of interest when learning concept embeddings. We show that this leads to embeddings which capture a more diverse range of commonsense properties, and consistently improves results in downstream tasks such as ultra-fine entity typing and ontology completion. | [
"Kteich, Hanane",
"Li, Na",
"Chatterjee, Usashi",
"Bouraoui, Zied",
"Schockaert, Steven"
]
| Modelling Commonsense Commonalities with Multi-Facet Concept Embeddings | findings-acl.86 | Poster | 2403.16984v2 |
https://aclanthology.org/2024.findings-acl.87.bib | @inproceedings{bonnier-2024-revisiting,
title = "Revisiting Multimodal Transformers for Tabular Data with Text Fields",
author = "Bonnier, Thomas",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.87",
pages = "1481--1500",
abstract = "Tabular data with text fields can be leveraged in applications such as financial risk assessment or medical diagnosis prediction. When employing multimodal approaches to make predictions based on these modalities, it is crucial to make the most appropriate modeling choices in terms of numerical feature encoding or fusion strategy. In this paper, we focus on multimodal classification tasks based on tabular datasets with text fields. We build on multimodal Transformers to propose the Tabular-Text Transformer (TTT), a tabular/text dual-stream Transformer network. This architecture includes a distance-to-quantile embedding scheme for numerical features and an overall attention module which concurrently considers self-attention and cross-modal attention. Further, we leverage the two well-informed modality streams to estimate whether a prediction is uncertain or not. To explain uncertainty in terms of feature values, we use a sampling-based approximation of Shapley values in a bimodal context, with two options for the value function. To show the efficacy and relevance of this approach, we compare it to six baselines and measure its ability to quantify and explain uncertainty against various methods. Our code is available at https://github.com/thomas-bonnier/TabularTextTransformer.",
}
| Tabular data with text fields can be leveraged in applications such as financial risk assessment or medical diagnosis prediction. When employing multimodal approaches to make predictions based on these modalities, it is crucial to make the most appropriate modeling choices in terms of numerical feature encoding or fusion strategy. In this paper, we focus on multimodal classification tasks based on tabular datasets with text fields. We build on multimodal Transformers to propose the Tabular-Text Transformer (TTT), a tabular/text dual-stream Transformer network. This architecture includes a distance-to-quantile embedding scheme for numerical features and an overall attention module which concurrently considers self-attention and cross-modal attention. Further, we leverage the two well-informed modality streams to estimate whether a prediction is uncertain or not. To explain uncertainty in terms of feature values, we use a sampling-based approximation of Shapley values in a bimodal context, with two options for the value function. To show the efficacy and relevance of this approach, we compare it to six baselines and measure its ability to quantify and explain uncertainty against various methods. Our code is available at https://github.com/thomas-bonnier/TabularTextTransformer. | [
"Bonnier, Thomas"
]
| Revisiting Multimodal Transformers for Tabular Data with Text Fields | findings-acl.87 | Poster | 2111.02705v1 |
https://aclanthology.org/2024.findings-acl.88.bib | @inproceedings{sadhu-etal-2024-empirical-study,
title = "An Empirical Study on the Characteristics of Bias upon Context Length Variation for {B}angla",
author = "Sadhu, Jayanta and
Khan, Ayan and
Bhattacharjee, Abhik and
Shahriyar, Rifat",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.88",
pages = "1501--1520",
abstract = "Pretrained language models inherently exhibit various social biases, prompting a crucial examination of their social impact across various linguistic contexts due to their widespread usage. Previous studies have provided numerous methods for intrinsic bias measurements, predominantly focused on high-resource languages. In this work, we aim to extend these investigations to Bangla, a low-resource language. Specifically, in this study, we (1) create a dataset for intrinsic gender bias measurement in Bangla, (2) discuss necessary adaptations to apply existing bias measurement methods for Bangla, and (3) examine the impact of context length variation on bias measurement, a factor that has been overlooked in previous studies. Through our experiments, we demonstrate a clear dependency of bias metrics on context length, highlighting the need for nuanced considerations in Bangla bias analysis. We consider our work as a stepping stone for bias measurement in the Bangla Language and make all of our resources publicly available to support future research.",
}
| Pretrained language models inherently exhibit various social biases, prompting a crucial examination of their social impact across various linguistic contexts due to their widespread usage. Previous studies have provided numerous methods for intrinsic bias measurements, predominantly focused on high-resource languages. In this work, we aim to extend these investigations to Bangla, a low-resource language. Specifically, in this study, we (1) create a dataset for intrinsic gender bias measurement in Bangla, (2) discuss necessary adaptations to apply existing bias measurement methods for Bangla, and (3) examine the impact of context length variation on bias measurement, a factor that has been overlooked in previous studies. Through our experiments, we demonstrate a clear dependency of bias metrics on context length, highlighting the need for nuanced considerations in Bangla bias analysis. We consider our work as a stepping stone for bias measurement in the Bangla Language and make all of our resources publicly available to support future research. | [
"Sadhu, Jayanta",
"Khan, Ayan",
"Bhattacharjee, Abhik",
"Shahriyar, Rifat"
]
| An Empirical Study on the Characteristics of Bias upon Context Length Variation for {B}angla | findings-acl.88 | Poster | 2406.17375v1 |
https://aclanthology.org/2024.findings-acl.89.bib | @inproceedings{niu-etal-2024-contempo,
title = "{C}on{T}empo: A Unified Temporally Contrastive Framework for Temporal Relation Extraction",
author = "Niu, Jingcheng and
Liao, Saifei and
Ng, Victoria and
De Montigny, Simon and
Penn, Gerald",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.89",
pages = "1521--1533",
abstract = "The task of temporal relation extraction (TRE) involves identifying and extracting temporal relations between events from narratives. We identify two primary issues with TRE systems. First, by formulating TRE as a simple text classification task where every temporal relation is independent, it is hard to enhance the TRE model{'}s representation of meaning of temporal relations, and its facility with the underlying temporal calculus. We solve the issue by proposing a novel Temporally Contrastive learning model (ConTempo) that increase the model{'}s awareness of the meaning of temporal relations by leveraging their symmetric or antisymmetric properties. Second, the reusability of innovations has been limited due to incompatibilities in model architectures. Therefore, we propose a unified framework and show that ConTempo is compatible with all three main branches of TRE research. Our results demonstrate that the performance gains of ConTempo are more pronounced, with the total combination achieving state-of-the-art performance on the widely used MATRES and TBD corpora. We furthermore identified and corrected a large number of annotation errors present in the test set of MATRES, after which the performance increase brought by ConTempo becomes more apparent.",
}
| The task of temporal relation extraction (TRE) involves identifying and extracting temporal relations between events from narratives. We identify two primary issues with TRE systems. First, by formulating TRE as a simple text classification task where every temporal relation is independent, it is hard to enhance the TRE model{'}s representation of meaning of temporal relations, and its facility with the underlying temporal calculus. We solve the issue by proposing a novel Temporally Contrastive learning model (ConTempo) that increase the model{'}s awareness of the meaning of temporal relations by leveraging their symmetric or antisymmetric properties. Second, the reusability of innovations has been limited due to incompatibilities in model architectures. Therefore, we propose a unified framework and show that ConTempo is compatible with all three main branches of TRE research. Our results demonstrate that the performance gains of ConTempo are more pronounced, with the total combination achieving state-of-the-art performance on the widely used MATRES and TBD corpora. We furthermore identified and corrected a large number of annotation errors present in the test set of MATRES, after which the performance increase brought by ConTempo becomes more apparent. | [
"Niu, Jingcheng",
"Liao, Saifei",
"Ng, Victoria",
"De Montigny, Simon",
"Penn, Gerald"
]
| {C}on{T}empo: A Unified Temporally Contrastive Framework for Temporal Relation Extraction | findings-acl.89 | Poster | 2305.17607v1 |
https://aclanthology.org/2024.findings-acl.90.bib | @inproceedings{ghaddar-etal-2024-charp,
title = "{CHARP}: Conversation History {A}wa{R}eness Probing for Knowledge-grounded Dialogue Systems",
author = "Ghaddar, Abbas and
Alfonso-Hermelo, David and
Langlais, Philippe and
Rezagholizadeh, Mehdi and
Chen, Boxing and
Parthasarathi, Prasanna",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.90",
pages = "1534--1551",
abstract = "In this work, we dive deep into one of the popular knowledge-grounded dialogue benchmarks that focus on faithfulness, FaithDial. We show that a significant portion of the FaithDial data contains annotation artifacts, which may bias models towards completely ignoring the conversation history. We therefore introduce CHARP, a testbed, designed for evaluating supposedly non-hallucinatory models trained on the FaithDial dataset. Our extensive analysis reveals that models primarily exhibit poor performance on CHARP due to their inability to effectively attend to and reason over the conversation history. Furthermore, the evaluation methods of FaithDial fail to capture these shortcomings, neglecting the conversational history. Our findings indicate that there is substantial room for contribution in both dataset creation and hallucination evaluation for knowledge-grounded dialogue, and that CHARP can serve as a tool for monitoring the progress in this particular research area. Data, models, and source code will be publicly available upon acceptance.",
}
| In this work, we dive deep into one of the popular knowledge-grounded dialogue benchmarks that focus on faithfulness, FaithDial. We show that a significant portion of the FaithDial data contains annotation artifacts, which may bias models towards completely ignoring the conversation history. We therefore introduce CHARP, a testbed, designed for evaluating supposedly non-hallucinatory models trained on the FaithDial dataset. Our extensive analysis reveals that models primarily exhibit poor performance on CHARP due to their inability to effectively attend to and reason over the conversation history. Furthermore, the evaluation methods of FaithDial fail to capture these shortcomings, neglecting the conversational history. Our findings indicate that there is substantial room for contribution in both dataset creation and hallucination evaluation for knowledge-grounded dialogue, and that CHARP can serve as a tool for monitoring the progress in this particular research area. Data, models, and source code will be publicly available upon acceptance. | [
"Ghaddar, Abbas",
"Alfonso-Hermelo, David",
"Langlais, Philippe",
"Rezagholizadeh, Mehdi",
"Chen, Boxing",
"Parthasarathi, Prasanna"
]
| {CHARP}: Conversation History {A}wa{R}eness Probing for Knowledge-grounded Dialogue Systems | findings-acl.90 | Poster | 2302.00907v1 |
https://aclanthology.org/2024.findings-acl.91.bib | @inproceedings{lin-etal-2024-criticbench,
title = "{C}ritic{B}ench: Benchmarking {LLM}s for Critique-Correct Reasoning",
author = "Lin, Zicheng and
Gou, Zhibin and
Liang, Tian and
Luo, Ruilin and
Liu, Haowei and
Yang, Yujiu",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.91",
pages = "1552--1587",
abstract = "The ability of Large Language Models (LLMs) to critique and refine their reasoning is crucial for their application in evaluation, feedback provision, and self-improvement. This paper introduces CriticBench, a comprehensive benchmark designed to assess LLMs{'} abilities to critique and rectify their reasoning across a variety of tasks. CriticBench encompasses five reasoning domains: mathematical, commonsense, symbolic, coding, and algorithmic. It compiles 15 datasets and incorporates responses from three LLM families. Utilizing CriticBench, we evaluate and dissect the performance of 17 LLMs in generation, critique, and correction reasoning, i.e., GQC reasoning. Our findings reveal: (1) a linear relationship in GQC capabilities, with critique-focused training markedly enhancing performance; (2) a task-dependent variation in correction effectiveness, with logic-oriented tasks being more amenable to correction; (3) GQC knowledge inconsistencies that decrease as model size increases; and (4) an intriguing inter-model critiquing dynamic, where stronger models are better at critiquing weaker ones, while weaker models can surprisingly surpass stronger ones in their self-critique. We hope these insights into the nuanced critique-correct reasoning of LLMs will foster further research in LLM critique and self-improvement.",
}
| The ability of Large Language Models (LLMs) to critique and refine their reasoning is crucial for their application in evaluation, feedback provision, and self-improvement. This paper introduces CriticBench, a comprehensive benchmark designed to assess LLMs{'} abilities to critique and rectify their reasoning across a variety of tasks. CriticBench encompasses five reasoning domains: mathematical, commonsense, symbolic, coding, and algorithmic. It compiles 15 datasets and incorporates responses from three LLM families. Utilizing CriticBench, we evaluate and dissect the performance of 17 LLMs in generation, critique, and correction reasoning, i.e., GQC reasoning. Our findings reveal: (1) a linear relationship in GQC capabilities, with critique-focused training markedly enhancing performance; (2) a task-dependent variation in correction effectiveness, with logic-oriented tasks being more amenable to correction; (3) GQC knowledge inconsistencies that decrease as model size increases; and (4) an intriguing inter-model critiquing dynamic, where stronger models are better at critiquing weaker ones, while weaker models can surprisingly surpass stronger ones in their self-critique. We hope these insights into the nuanced critique-correct reasoning of LLMs will foster further research in LLM critique and self-improvement. | [
"Lin, Zicheng",
"Gou, Zhibin",
"Liang, Tian",
"Luo, Ruilin",
"Liu, Haowei",
"Yang, Yujiu"
]
| {C}ritic{B}ench: Benchmarking {LLM}s for Critique-Correct Reasoning | findings-acl.91 | Poster | 2303.12810v1 |
https://aclanthology.org/2024.findings-acl.92.bib | @inproceedings{zhang-etal-2024-dafnet,
title = "{DAFN}et: Dynamic Auxiliary Fusion for Sequential Model Editing in Large Language Models",
author = "Zhang, Taolin and
Chen, Qizhou and
Li, Dongyang and
Wang, Chengyu and
He, Xiaofeng and
Huang, Longtao and
Xue{'}, Hui and
Huang, Jun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.92",
pages = "1588--1602",
abstract = "Recently, while large language models (LLMs) have demonstrated impressive results, they still suffer from hallucination, i.e., the generation of false information. Model editing is the task of fixing factual mistakes in LLMs; yet, most previous works treat it as a one-time task, paying little attention to ever-emerging mistakes generated by LLMs. We address the task of sequential model editing (SME) that aims to rectify mistakes continuously. A Dynamic Auxiliary Fusion Network (DAFNet) is designed to enhance the semantic interaction among the factual knowledge within the entire sequence, preventing catastrophic forgetting during the editing process of multiple knowledge triples.Specifically, (1) for semantic fusion within a relation triple, we aggregate the intra-editing attention flow into auto-regressive self-attention with token-level granularity in LLMs. We further leverage multi-layer diagonal inter-editing attention flow to update the weighted representations of the entire sequence-level granularity. (2) Considering that auxiliary parameters are required to store the knowledge for sequential editing, we construct a new dataset named DAFSet, fulfilling recent, popular, long-tail and robust properties to enhance the generality of sequential editing. Experiments show DAFNet significantly outperforms strong baselines in single-turn and sequential editing. The usage of DAFSet also consistently improves the performance of other auxiliary network-based methods in various scenarios.",
}
| Recently, while large language models (LLMs) have demonstrated impressive results, they still suffer from hallucination, i.e., the generation of false information. Model editing is the task of fixing factual mistakes in LLMs; yet, most previous works treat it as a one-time task, paying little attention to ever-emerging mistakes generated by LLMs. We address the task of sequential model editing (SME) that aims to rectify mistakes continuously. A Dynamic Auxiliary Fusion Network (DAFNet) is designed to enhance the semantic interaction among the factual knowledge within the entire sequence, preventing catastrophic forgetting during the editing process of multiple knowledge triples.Specifically, (1) for semantic fusion within a relation triple, we aggregate the intra-editing attention flow into auto-regressive self-attention with token-level granularity in LLMs. We further leverage multi-layer diagonal inter-editing attention flow to update the weighted representations of the entire sequence-level granularity. (2) Considering that auxiliary parameters are required to store the knowledge for sequential editing, we construct a new dataset named DAFSet, fulfilling recent, popular, long-tail and robust properties to enhance the generality of sequential editing. Experiments show DAFNet significantly outperforms strong baselines in single-turn and sequential editing. The usage of DAFSet also consistently improves the performance of other auxiliary network-based methods in various scenarios. | [
"Zhang, Taolin",
"Chen, Qizhou",
"Li, Dongyang",
"Wang, Chengyu",
"He, Xiaofeng",
"Huang, Longtao",
"Xue{'}, Hui",
"Huang, Jun"
]
| {DAFN}et: Dynamic Auxiliary Fusion for Sequential Model Editing in Large Language Models | findings-acl.92 | Poster | 2405.20588v1 |
https://aclanthology.org/2024.findings-acl.93.bib | @inproceedings{urlana-etal-2024-controllable,
title = "Controllable Text Summarization: Unraveling Challenges, Approaches, and Prospects - A Survey",
author = "Urlana, Ashok and
Mishra, Pruthwik and
Roy, Tathagato and
Mishra, Rahul",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.93",
pages = "1603--1623",
abstract = "Generic text summarization approaches often fail to address the specific intent and needs of individual users. Recently, scholarly attention has turned to the development of summarization methods that are more closely tailored and controlled to align with specific objectives and user needs. Despite a growing corpus of controllable summarization research, there is no comprehensive survey available that thoroughly explores the diverse controllable attributes employed in this context, delves into the associated challenges, and investigates the existing solutions. In this survey, we formalize the Controllable Text Summarization (CTS) task, categorize controllable attributes according to their shared characteristics and objectives, and present a thorough examination of existing datasets and methods within each category. Moreover, based on our findings, we uncover limitations and research gaps, while also exploring potential solutions and future directions for CTS. We release our detailed analysis of CTS papers at https://github.com/ashokurlana/controllable{\_}text{\_}summarization{\_}survey.",
}
| Generic text summarization approaches often fail to address the specific intent and needs of individual users. Recently, scholarly attention has turned to the development of summarization methods that are more closely tailored and controlled to align with specific objectives and user needs. Despite a growing corpus of controllable summarization research, there is no comprehensive survey available that thoroughly explores the diverse controllable attributes employed in this context, delves into the associated challenges, and investigates the existing solutions. In this survey, we formalize the Controllable Text Summarization (CTS) task, categorize controllable attributes according to their shared characteristics and objectives, and present a thorough examination of existing datasets and methods within each category. Moreover, based on our findings, we uncover limitations and research gaps, while also exploring potential solutions and future directions for CTS. We release our detailed analysis of CTS papers at https://github.com/ashokurlana/controllable{\_}text{\_}summarization{\_}survey. | [
"Urlana, Ashok",
"Mishra, Pruthwik",
"Roy, Tathagato",
"Mishra, Rahul"
]
| Controllable Text Summarization: Unraveling Challenges, Approaches, and Prospects - A Survey | findings-acl.93 | Poster | 2307.08411v1 |
https://aclanthology.org/2024.findings-acl.94.bib | @inproceedings{huang-etal-2024-benchmarking,
title = "Benchmarking Large Language Models on Communicative Medical Coaching: A Dataset and a Novel System",
author = "Huang, Hengguan and
Wang, Songtao and
Liu, Hongfu and
Wang, Hao and
Wang, Ye",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.94",
pages = "1624--1637",
abstract = "Traditional applications of natural language processing (NLP) in healthcare have predominantly focused on patient-centered services, enhancing patient interactions and care delivery, such as through medical dialogue systems. However, the potential of NLP to benefit inexperienced doctors, particularly in areas such as communicative medical coaching, remains largely unexplored. We introduce {``}ChatCoach{''}, a human-AI cooperative framework designed to assist medical learners in practicing their communication skills during patient consultations. ChatCoach differentiates itself from conventional dialogue systems by offering a simulated environment where medical learners can practice dialogues with a patient agent, while a coach agent provides immediate, structured feedback. This is facilitated by our proposed Generalized Chain-of-Thought (GCoT) approach, which fosters the generation of structured feedback and enhances the utilization of external knowledge sources. Additionally, we have developed a dataset specifically for evaluating Large Language Models (LLMs) within the ChatCoach framework on communicative medical coaching tasks. Our empirical results validate the effectiveness of ChatCoach.",
}
| Traditional applications of natural language processing (NLP) in healthcare have predominantly focused on patient-centered services, enhancing patient interactions and care delivery, such as through medical dialogue systems. However, the potential of NLP to benefit inexperienced doctors, particularly in areas such as communicative medical coaching, remains largely unexplored. We introduce {``}ChatCoach{''}, a human-AI cooperative framework designed to assist medical learners in practicing their communication skills during patient consultations. ChatCoach differentiates itself from conventional dialogue systems by offering a simulated environment where medical learners can practice dialogues with a patient agent, while a coach agent provides immediate, structured feedback. This is facilitated by our proposed Generalized Chain-of-Thought (GCoT) approach, which fosters the generation of structured feedback and enhances the utilization of external knowledge sources. Additionally, we have developed a dataset specifically for evaluating Large Language Models (LLMs) within the ChatCoach framework on communicative medical coaching tasks. Our empirical results validate the effectiveness of ChatCoach. | [
"Huang, Hengguan",
"Wang, Songtao",
"Liu, Hongfu",
"Wang, Hao",
"Wang, Ye"
]
| Benchmarking Large Language Models on Communicative Medical Coaching: A Dataset and a Novel System | findings-acl.94 | Poster | 2402.05547v2 |
https://aclanthology.org/2024.findings-acl.95.bib | @inproceedings{ding-etal-2024-everything,
title = "Everything of Thoughts: Defying the Law of Penrose Triangle for Thought Generation",
author = "Ding, Ruomeng and
Zhang, Chaoyun and
Wang, Lu and
Xu, Yong and
Ma, Minghua and
Zhang, Wei and
Qin, Si and
Rajmohan, Saravan and
Lin, Qingwei and
Zhang, Dongmei",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.95",
pages = "1638--1662",
abstract = "This paper introduce a novel thought prompting approach called {''}Everything of Thoughts{''} (XoT) for Large Language Models (LLMs) to defy the law of {''}Penrose triangle{''} of existing thought paradigms, to achieve three key perspectives in thought generation simultaneously: performance, efficiency, and flexibility. XoT leverages pretrained reinforcement learning and Monte Carlo Tree Search (MCTS) to incorporate external domain knowledge and planning capability into thoughts, thereby enhancing LLMs{'} decision-making capabilities. Through the MCTS-LLM collaborative thought revision framework, XoT autonomously produces high-quality comprehensive cognitive mappings with minimal LLM interactions. Additionally, XoT empowers LLMs to utilize flexible cognitive mappings for solving problems with multiple solutions.We evaluate XoT on several challenging problem-solving tasks, including Game of 24, 8-Puzzle, and Pocket Cube. Our results demonstrate that XoT significantly outperforms existing approaches in various dimensions, showcasing its remarkable proficiency in addressing complex problems across diverse domains. The data and code are available at https://github.com/microsoft/Everything-of-Thoughts-XoT.",
}
| This paper introduce a novel thought prompting approach called {''}Everything of Thoughts{''} (XoT) for Large Language Models (LLMs) to defy the law of {''}Penrose triangle{''} of existing thought paradigms, to achieve three key perspectives in thought generation simultaneously: performance, efficiency, and flexibility. XoT leverages pretrained reinforcement learning and Monte Carlo Tree Search (MCTS) to incorporate external domain knowledge and planning capability into thoughts, thereby enhancing LLMs{'} decision-making capabilities. Through the MCTS-LLM collaborative thought revision framework, XoT autonomously produces high-quality comprehensive cognitive mappings with minimal LLM interactions. Additionally, XoT empowers LLMs to utilize flexible cognitive mappings for solving problems with multiple solutions.We evaluate XoT on several challenging problem-solving tasks, including Game of 24, 8-Puzzle, and Pocket Cube. Our results demonstrate that XoT significantly outperforms existing approaches in various dimensions, showcasing its remarkable proficiency in addressing complex problems across diverse domains. The data and code are available at https://github.com/microsoft/Everything-of-Thoughts-XoT. | [
"Ding, Ruomeng",
"Zhang, Chaoyun",
"Wang, Lu",
"Xu, Yong",
"Ma, Minghua",
"Zhang, Wei",
"Qin, Si",
"Rajmohan, Saravan",
"Lin, Qingwei",
"Zhang, Dongmei"
]
| Everything of Thoughts: Defying the Law of Penrose Triangle for Thought Generation | findings-acl.95 | Poster | 2311.04254v3 |
https://aclanthology.org/2024.findings-acl.96.bib | @inproceedings{zhang-etal-2024-spaghetti,
title = "{SPAGHETTI}: Open-Domain Question Answering from Heterogeneous Data Sources with Retrieval and Semantic Parsing",
author = "Zhang, Heidi and
Semnani, Sina and
Ghassemi, Farhad and
Xu, Jialiang and
Liu, Shicheng and
Lam, Monica",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.96",
pages = "1663--1678",
abstract = "We introduce SPAGHETTI: Semantic Parsing Augmented Generation for Hybrid English information from Text Tables and Infoboxes, a hybrid question-answering (QA) pipeline that utilizes information from heterogeneous knowledge sources, including knowledge base, text, tables, and infoboxes. Our LLM-augmented approach achieves state-of-the-art performance on the Compmix dataset, the most comprehensive heterogeneous open-domain QA dataset, with 56.5{\%} exact match (EM) rate. More importantly, manual analysis on a sample of the dataset suggests that SPAGHETTI is more than 90{\%} accurate, indicating that EM is no longer suitable for assessing the capabilities of QA systems today.",
}
| We introduce SPAGHETTI: Semantic Parsing Augmented Generation for Hybrid English information from Text Tables and Infoboxes, a hybrid question-answering (QA) pipeline that utilizes information from heterogeneous knowledge sources, including knowledge base, text, tables, and infoboxes. Our LLM-augmented approach achieves state-of-the-art performance on the Compmix dataset, the most comprehensive heterogeneous open-domain QA dataset, with 56.5{\%} exact match (EM) rate. More importantly, manual analysis on a sample of the dataset suggests that SPAGHETTI is more than 90{\%} accurate, indicating that EM is no longer suitable for assessing the capabilities of QA systems today. | [
"Zhang, Heidi",
"Semnani, Sina",
"Ghassemi, Farhad",
"Xu, Jialiang",
"Liu, Shicheng",
"Lam, Monica"
]
| {SPAGHETTI}: Open-Domain Question Answering from Heterogeneous Data Sources with Retrieval and Semantic Parsing | findings-acl.96 | Poster | 2406.00562v1 |
https://aclanthology.org/2024.findings-acl.97.bib | @inproceedings{ding-etal-2024-data,
title = "Data Augmentation using {LLM}s: Data Perspectives, Learning Paradigms and Challenges",
author = "Ding, Bosheng and
Qin, Chengwei and
Zhao, Ruochen and
Luo, Tianze and
Li, Xinze and
Chen, Guizhen and
Xia, Wenhan and
Hu, Junjie and
Luu, Anh Tuan and
Joty, Shafiq",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.97",
pages = "1679--1705",
abstract = "In the rapidly evolving field of large language models (LLMs), data augmentation (DA) has emerged as a pivotal technique for enhancing model performance by diversifying training examples without the need for additional data collection. This survey explores the transformative impact of LLMs on DA, particularly addressing the unique challenges and opportunities they present in the context of natural language processing (NLP) and beyond. From both data and learning perspectives, we examine various strategies that utilize LLMs for data augmentation, including a novel exploration of learning paradigms where LLM-generated data is used for diverse forms of further training. Additionally, this paper highlights the primary open challenges faced in this domain, ranging from controllable data augmentation to multi-modal data augmentation. This survey highlights a paradigm shift introduced by LLMs in DA, and aims to serve as a comprehensive guide for researchers and practitioners.",
}
| In the rapidly evolving field of large language models (LLMs), data augmentation (DA) has emerged as a pivotal technique for enhancing model performance by diversifying training examples without the need for additional data collection. This survey explores the transformative impact of LLMs on DA, particularly addressing the unique challenges and opportunities they present in the context of natural language processing (NLP) and beyond. From both data and learning perspectives, we examine various strategies that utilize LLMs for data augmentation, including a novel exploration of learning paradigms where LLM-generated data is used for diverse forms of further training. Additionally, this paper highlights the primary open challenges faced in this domain, ranging from controllable data augmentation to multi-modal data augmentation. This survey highlights a paradigm shift introduced by LLMs in DA, and aims to serve as a comprehensive guide for researchers and practitioners. | [
"Ding, Bosheng",
"Qin, Chengwei",
"Zhao, Ruochen",
"Luo, Tianze",
"Li, Xinze",
"Chen, Guizhen",
"Xia, Wenhan",
"Hu, Junjie",
"Luu, Anh Tuan",
"Joty, Shafiq"
]
| Data Augmentation using {LLM}s: Data Perspectives, Learning Paradigms and Challenges | findings-acl.97 | Poster | 2403.02990v4 |
https://aclanthology.org/2024.findings-acl.98.bib | @inproceedings{hou-etal-2024-k,
title = "k-{S}em{S}tamp: A Clustering-Based Semantic Watermark for Detection of Machine-Generated Text",
author = "Hou, Abe and
Zhang, Jingyu and
Wang, Yichen and
Khashabi, Daniel and
He, Tianxing",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.98",
pages = "1706--1715",
abstract = "Recent watermarked generation algorithms inject detectable signatures during language generation to facilitate post-hoc detection. While token-level watermarks are vulnerable to paraphrase attacks, SemStamp (Hou et al., 2023) applies watermark on the semantic representation of sentences and demonstrates promising robustness. SemStamp employs locality-sensitive hashing (LSH) to partition the semantic space with arbitrary hyperplanes, which results in a suboptimal tradeoff between robustness and speed. We propose k-SemStamp, a simple yet effective enhancement of SemStamp, utilizing k-means clustering as an alternative of LSH to partition the embedding space with awareness of inherent semantic structure. Experimental results indicate that k-SemStamp saliently improves its robustness and sampling efficiency while preserving the generation quality, advancing a more effective tool for machine-generated text detection.",
}
| Recent watermarked generation algorithms inject detectable signatures during language generation to facilitate post-hoc detection. While token-level watermarks are vulnerable to paraphrase attacks, SemStamp (Hou et al., 2023) applies watermark on the semantic representation of sentences and demonstrates promising robustness. SemStamp employs locality-sensitive hashing (LSH) to partition the semantic space with arbitrary hyperplanes, which results in a suboptimal tradeoff between robustness and speed. We propose k-SemStamp, a simple yet effective enhancement of SemStamp, utilizing k-means clustering as an alternative of LSH to partition the embedding space with awareness of inherent semantic structure. Experimental results indicate that k-SemStamp saliently improves its robustness and sampling efficiency while preserving the generation quality, advancing a more effective tool for machine-generated text detection. | [
"Hou, Abe",
"Zhang, Jingyu",
"Wang, Yichen",
"Khashabi, Daniel",
"He, Tianxing"
]
| k-{S}em{S}tamp: A Clustering-Based Semantic Watermark for Detection of Machine-Generated Text | findings-acl.98 | Poster | 2402.18059v3 |
https://aclanthology.org/2024.findings-acl.99.bib | @inproceedings{burapacheep-etal-2024-colorswap,
title = "{C}olor{S}wap: A Color and Word Order Dataset for Multimodal Evaluation",
author = "Burapacheep, Jirayu and
Gaur, Ishan and
Bhatia, Agam and
Thrush, Tristan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.99",
pages = "1716--1726",
abstract = "This paper introduces the ColorSwap dataset, designed to assess and improve the proficiency of multimodal models in matching objects with their colors. The dataset is comprised of 2,000 unique image-caption pairs, grouped into 1,000 examples. Each example includes a caption-image pair, along with a {``}color-swapped{''} pair. We follow the Winoground schema: the two captions in an example have the same words, but the color words have been rearranged to modify different objects. The dataset was created through a novel blend of automated caption and image generation with humans in the loop. We evaluate image-text matching (ITM) and visual language models (VLMs) and find that even the latest ones are still not robust at this task. GPT-4V and LLaVA score 72{\%} and 42{\%} on our main VLM metric, although they may improve with more advanced prompting techniques. On the main ITM metric, contrastive models such as CLIP and SigLIP perform close to chance (at 12{\%} and 30{\%}, respectively), although the non-contrastive BLIP ITM model is stronger (87{\%}). We also find that finetuning on fewer than 2,000 examples yields significant performance gains on this out-of-distribution word-order understanding task.",
}
| This paper introduces the ColorSwap dataset, designed to assess and improve the proficiency of multimodal models in matching objects with their colors. The dataset is comprised of 2,000 unique image-caption pairs, grouped into 1,000 examples. Each example includes a caption-image pair, along with a {``}color-swapped{''} pair. We follow the Winoground schema: the two captions in an example have the same words, but the color words have been rearranged to modify different objects. The dataset was created through a novel blend of automated caption and image generation with humans in the loop. We evaluate image-text matching (ITM) and visual language models (VLMs) and find that even the latest ones are still not robust at this task. GPT-4V and LLaVA score 72{\%} and 42{\%} on our main VLM metric, although they may improve with more advanced prompting techniques. On the main ITM metric, contrastive models such as CLIP and SigLIP perform close to chance (at 12{\%} and 30{\%}, respectively), although the non-contrastive BLIP ITM model is stronger (87{\%}). We also find that finetuning on fewer than 2,000 examples yields significant performance gains on this out-of-distribution word-order understanding task. | [
"Burapacheep, Jirayu",
"Gaur, Ishan",
"Bhatia, Agam",
"Thrush, Tristan"
]
| {C}olor{S}wap: A Color and Word Order Dataset for Multimodal Evaluation | findings-acl.99 | Poster | 2402.04492v2 |
https://aclanthology.org/2024.findings-acl.100.bib | @inproceedings{zhang-etal-2024-revisiting-opro,
title = "Revisiting {OPRO}: The Limitations of Small-Scale {LLM}s as Optimizers",
author = "Zhang, Tuo and
Yuan, Jinyue and
Avestimehr, Salman",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.100",
pages = "1727--1735",
abstract = "Numerous recent works aim to enhance the efficacy of Large Language Models (LLMs) through strategic prompting. In particular, the Optimization by PROmpting (OPRO) approach provides state-of-the-art performance by leveraging LLMs as optimizers where the optimization task is to find instructions that maximize the task accuracy. In this paper, we revisit OPRO for automated prompting with relatively small-scale LLMs, such as LLaMa-2 family and Mistral 7B. Our investigation reveals that OPRO shows limited effectiveness in small-scale LLMs, with limited inference capabilities constraining optimization ability. We suggest future automatic prompting engineering to consider both model capabilities and computational costs. Additionally, for small-scale LLMs, we recommend direct instructions that clearly outline objectives and methodologies as robust prompt baselines, ensuring efficient and effective prompt engineering in ongoing research.",
}
| Numerous recent works aim to enhance the efficacy of Large Language Models (LLMs) through strategic prompting. In particular, the Optimization by PROmpting (OPRO) approach provides state-of-the-art performance by leveraging LLMs as optimizers where the optimization task is to find instructions that maximize the task accuracy. In this paper, we revisit OPRO for automated prompting with relatively small-scale LLMs, such as LLaMa-2 family and Mistral 7B. Our investigation reveals that OPRO shows limited effectiveness in small-scale LLMs, with limited inference capabilities constraining optimization ability. We suggest future automatic prompting engineering to consider both model capabilities and computational costs. Additionally, for small-scale LLMs, we recommend direct instructions that clearly outline objectives and methodologies as robust prompt baselines, ensuring efficient and effective prompt engineering in ongoing research. | [
"Zhang, Tuo",
"Yuan, Jinyue",
"Avestimehr, Salman"
]
| Revisiting {OPRO}: The Limitations of Small-Scale {LLM}s as Optimizers | findings-acl.100 | Poster | 2405.10276v2 |
https://aclanthology.org/2024.findings-acl.101.bib | @inproceedings{bajpai-hanawal-2024-ceebert,
title = "{C}ee{BERT}: Cross-Domain Inference in Early Exit {BERT}",
author = "Bajpai, Divya Jyoti and
Hanawal, Manjesh",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.101",
pages = "1736--1748",
abstract = "Pre-trained Language Models (PLMs), like BERT, with self-supervision objectives exhibit remarkable performance and generalization across various tasks. However, they suffer in inference latency due to their large size. To address this issue, side branches are attached at intermediate layers, enabling early inference of samples without requiring them to pass through all layers. However, the challenge is to decide which layer to infer and exit each sample so that the accuracy and latency are balanced. Moreover, the distribution of the samples to be inferred may differ from that used for training necessitating cross-domain adaptation. We propose an online learning algorithm named Cross-Domain Inference in Early Exit BERT (CeeBERT) that dynamically determines early exits of samples based on the level of confidence at each exit point. CeeBERT learns optimal thresholds from domain-specific confidence observed at intermediate layers on the fly, eliminating the need for labeled data. Experimental results on five distinct datasets with BERT and ALBERT models demonstrate CeeBERT{'}s ability to improve latency by reducing unnecessary computations with minimal drop in performance. By adapting to the threshold values, CeeBERT can speed up the BERT/ALBERT models by $2\times$ - $3.1\times$ with minimal drop in accuracy. The anonymized source code is available at https://github.com/Div290/CeeBERT.",
}
| Pre-trained Language Models (PLMs), like BERT, with self-supervision objectives exhibit remarkable performance and generalization across various tasks. However, they suffer in inference latency due to their large size. To address this issue, side branches are attached at intermediate layers, enabling early inference of samples without requiring them to pass through all layers. However, the challenge is to decide which layer to infer and exit each sample so that the accuracy and latency are balanced. Moreover, the distribution of the samples to be inferred may differ from that used for training necessitating cross-domain adaptation. We propose an online learning algorithm named Cross-Domain Inference in Early Exit BERT (CeeBERT) that dynamically determines early exits of samples based on the level of confidence at each exit point. CeeBERT learns optimal thresholds from domain-specific confidence observed at intermediate layers on the fly, eliminating the need for labeled data. Experimental results on five distinct datasets with BERT and ALBERT models demonstrate CeeBERT{'}s ability to improve latency by reducing unnecessary computations with minimal drop in performance. By adapting to the threshold values, CeeBERT can speed up the BERT/ALBERT models by $2\times$ - $3.1\times$ with minimal drop in accuracy. The anonymized source code is available at https://github.com/Div290/CeeBERT. | [
"Bajpai, Divya Jyoti",
"Hanawal, Manjesh"
]
| {C}ee{BERT}: Cross-Domain Inference in Early Exit {BERT} | findings-acl.101 | Poster | 2101.09755v1 |
https://aclanthology.org/2024.findings-acl.102.bib | @inproceedings{das-srihari-2024-uniwiz,
title = "{UNIWIZ}: A Unified Large Language Model Orchestrated Wizard for Safe Knowledge Grounded Conversations",
author = "Das, Souvik and
Srihari, Rohini",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.102",
pages = "1749--1762",
abstract = "Large Language Models (LLMs) have made significant progress in integrating safety and knowledge alignment. However, adversarial actors can manipulate these models into generating unsafe responses, and excessive safety alignment can lead to unintended hallucinations. To address these challenges, we introduce UniWiz, a novel 2-step data orchestration framework that unifies safety and knowledge data generation. We propose a {``}safety-priming{''} method to generate synthetic safety data and overcome safety bottlenecks. We also inject relevant knowledge into conversations by retrieving factual information from curated sources. UniWiz dataset consists of 17,638 quality-controlled conversations and 10,000 augmented preference data. Pretrained models fine-tuned on UniWiz show improvements across various metrics and outperform state-of-the-art instruction-tuned models trained on much larger datasets.",
}
| Large Language Models (LLMs) have made significant progress in integrating safety and knowledge alignment. However, adversarial actors can manipulate these models into generating unsafe responses, and excessive safety alignment can lead to unintended hallucinations. To address these challenges, we introduce UniWiz, a novel 2-step data orchestration framework that unifies safety and knowledge data generation. We propose a {``}safety-priming{''} method to generate synthetic safety data and overcome safety bottlenecks. We also inject relevant knowledge into conversations by retrieving factual information from curated sources. UniWiz dataset consists of 17,638 quality-controlled conversations and 10,000 augmented preference data. Pretrained models fine-tuned on UniWiz show improvements across various metrics and outperform state-of-the-art instruction-tuned models trained on much larger datasets. | [
"Das, Souvik",
"Srihari, Rohini"
]
| {UNIWIZ}: A Unified Large Language Model Orchestrated Wizard for Safe Knowledge Grounded Conversations | findings-acl.102 | Poster | 2002.07510v2 |
https://aclanthology.org/2024.findings-acl.103.bib | @inproceedings{thompson-etal-2024-shocking,
title = "A Shocking Amount of the Web is Machine Translated: Insights from Multi-Way Parallelism",
author = "Thompson, Brian and
Dhaliwal, Mehak and
Frisch, Peter and
Domhan, Tobias and
Federico, Marcello",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.103",
pages = "1763--1775",
abstract = "We show that content on the web is often translated into many languages, and the low quality of these multi-way translations indicates they were likely created using Machine Translation (MT). Multi-way parallel, machine generated content not only dominates the translations in lower resource languages; it also constitutes a large fraction of the total web content in those languages. We also find evidence of a selection bias in the type of content which is translated into many languages, consistent with low quality English content being translated en masse into many lower resource languages, via MT. Our work raises serious concerns about training models such as multilingual large language models on both monolingual and bilingual data scraped from the web.",
}
| We show that content on the web is often translated into many languages, and the low quality of these multi-way translations indicates they were likely created using Machine Translation (MT). Multi-way parallel, machine generated content not only dominates the translations in lower resource languages; it also constitutes a large fraction of the total web content in those languages. We also find evidence of a selection bias in the type of content which is translated into many languages, consistent with low quality English content being translated en masse into many lower resource languages, via MT. Our work raises serious concerns about training models such as multilingual large language models on both monolingual and bilingual data scraped from the web. | [
"Thompson, Brian",
"Dhaliwal, Mehak",
"Frisch, Peter",
"Domhan, Tobias",
"Federico, Marcello"
]
| A Shocking Amount of the Web is Machine Translated: Insights from Multi-Way Parallelism | findings-acl.103 | Poster | 2401.05749v2 |
https://aclanthology.org/2024.findings-acl.104.bib | @inproceedings{perin-etal-2024-rankmean,
title = "{R}ank{M}ean: Module-Level Importance Score for Merging Fine-tuned {LLM} Models",
author = "Perin, Gabriel and
Chen, Xuxi and
Liu, Shusen and
Kailkhura, Bhavya and
Wang, Zhangyang and
Gallagher, Brian",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.104",
pages = "1776--1782",
abstract = "Traditionally, developing new language models (LMs) capable of addressing multiple tasks involves fine-tuning pre-trained LMs using a wide collection of datasets, a process that often incurs significant computational expenses. Model merging emerges as a cost-effective alternative, allowing the integration of existing models fine-tuned on different tasks into a single model that performs well across all tasks, eliminating the need for additional training. In this paper, we propose RankMean, an algorithm for merging fine-tuned LMs without requiring any downstream data. RankMean determines merging coefficients based on the relative rankings of weight change magnitudes and applies these coefficients for module-wise integration of various fine-tuned models. Our experimental results demonstrate that RankMean outperforms existing baseline methods on multiple benchmarks. The code is available at https://github.com/VITA-Group/RankMean.",
}
| Traditionally, developing new language models (LMs) capable of addressing multiple tasks involves fine-tuning pre-trained LMs using a wide collection of datasets, a process that often incurs significant computational expenses. Model merging emerges as a cost-effective alternative, allowing the integration of existing models fine-tuned on different tasks into a single model that performs well across all tasks, eliminating the need for additional training. In this paper, we propose RankMean, an algorithm for merging fine-tuned LMs without requiring any downstream data. RankMean determines merging coefficients based on the relative rankings of weight change magnitudes and applies these coefficients for module-wise integration of various fine-tuned models. Our experimental results demonstrate that RankMean outperforms existing baseline methods on multiple benchmarks. The code is available at https://github.com/VITA-Group/RankMean. | [
"Perin, Gabriel",
"Chen, Xuxi",
"Liu, Shusen",
"Kailkhura, Bhavya",
"Wang, Zhangyang",
"Gallagher, Brian"
]
| {R}ank{M}ean: Module-Level Importance Score for Merging Fine-tuned {LLM} Models | findings-acl.104 | Poster | 2408.03092v1 |
https://aclanthology.org/2024.findings-acl.105.bib | @inproceedings{qiu-etal-2024-valor,
title = "{VALOR}-{EVAL}: Holistic Coverage and Faithfulness Evaluation of Large Vision-Language Models",
author = "Qiu, Haoyi and
Hu, Wenbo and
Dou, Zi-Yi and
Peng, Nanyun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.105",
pages = "1783--1805",
abstract = "Large Vision-Language Models (LVLMs) suffer from hallucination issues, wherein the models generate plausible-sounding but factually incorrect outputs, undermining their reliability. A comprehensive quantitative evaluation is necessary to identify and understand the extent of hallucinations in these models. However, existing benchmarks are often limited in scope, focusing mainly on object hallucinations. Furthermore, current evaluation methods struggle to effectively address the subtle semantic distinctions between model outputs and reference data, as well as the balance between hallucination and informativeness. To address these issues, we introduce a multi-dimensional benchmark covering objects, attributes, and relations, with challenging images selected based on associative biases. Moreover, we propose a large language model (LLM)-based two-stage evaluation framework that generalizes the popular CHAIR metric and incorporates both faithfulness and coverage into the evaluation. Experiments on 10 established LVLMs demonstrate that our evaluation metric is more comprehensive and better correlated with humans than existing work when evaluating on our challenging human-annotated benchmark dataset. Our work also highlights the critical balance between faithfulness and coverage of model outputs, and encourages future works to address hallucinations in LVLMs while keeping their outputs informative.",
}
| Large Vision-Language Models (LVLMs) suffer from hallucination issues, wherein the models generate plausible-sounding but factually incorrect outputs, undermining their reliability. A comprehensive quantitative evaluation is necessary to identify and understand the extent of hallucinations in these models. However, existing benchmarks are often limited in scope, focusing mainly on object hallucinations. Furthermore, current evaluation methods struggle to effectively address the subtle semantic distinctions between model outputs and reference data, as well as the balance between hallucination and informativeness. To address these issues, we introduce a multi-dimensional benchmark covering objects, attributes, and relations, with challenging images selected based on associative biases. Moreover, we propose a large language model (LLM)-based two-stage evaluation framework that generalizes the popular CHAIR metric and incorporates both faithfulness and coverage into the evaluation. Experiments on 10 established LVLMs demonstrate that our evaluation metric is more comprehensive and better correlated with humans than existing work when evaluating on our challenging human-annotated benchmark dataset. Our work also highlights the critical balance between faithfulness and coverage of model outputs, and encourages future works to address hallucinations in LVLMs while keeping their outputs informative. | [
"Qiu, Haoyi",
"Hu, Wenbo",
"Dou, Zi-Yi",
"Peng, Nanyun"
]
| {VALOR}-{EVAL}: Holistic Coverage and Faithfulness Evaluation of Large Vision-Language Models | findings-acl.105 | Poster | 2303.03278v1 |
https://aclanthology.org/2024.findings-acl.106.bib | @inproceedings{cheng-etal-2024-cyclical,
title = "Cyclical Contrastive Learning Based on Geodesic for Zero-shot Cross-lingual Spoken Language Understanding",
author = "Cheng, Xuxin and
Zhu, Zhihong and
Yang, Bang and
Zhuang, Xianwei and
Li, Hongxiang and
Zou, Yuexian",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.106",
pages = "1806--1816",
abstract = "Owing to the scarcity of labeled training data, Spoken Language Understanding (SLU) is still a challenging task in low-resource languages. Therefore, zero-shot cross-lingual SLU attracts more and more attention. Contrastive learning is widely applied to explicitly align representations of similar sentences across different languages. However, the vanilla contrastive learning method may face two problems in zero-shot cross-lingual SLU: (1) the consistency between different languages is neglected; (2) each utterance has two different kinds of SLU labels, i.e. slot and intent, the utterances with one different label are also pushed away without any discrimination, which limits the performance. In this paper, we propose Cyclical Contrastive Learning based on Geodesic (CCLG), which introduces cyclical contrastive learning to achieve the consistency between different languages and leverages geodesic to measure the similarity to construct the positive pairs and negative pairs. Experimental results demonstrate that our proposed framework achieves the new state-of-the-art performance on MultiATIS++ and MTOP datasets, and the model analysis further verifies that CCLG can effectively transfer knowledge between different languages.",
}
| Owing to the scarcity of labeled training data, Spoken Language Understanding (SLU) is still a challenging task in low-resource languages. Therefore, zero-shot cross-lingual SLU attracts more and more attention. Contrastive learning is widely applied to explicitly align representations of similar sentences across different languages. However, the vanilla contrastive learning method may face two problems in zero-shot cross-lingual SLU: (1) the consistency between different languages is neglected; (2) each utterance has two different kinds of SLU labels, i.e. slot and intent, the utterances with one different label are also pushed away without any discrimination, which limits the performance. In this paper, we propose Cyclical Contrastive Learning based on Geodesic (CCLG), which introduces cyclical contrastive learning to achieve the consistency between different languages and leverages geodesic to measure the similarity to construct the positive pairs and negative pairs. Experimental results demonstrate that our proposed framework achieves the new state-of-the-art performance on MultiATIS++ and MTOP datasets, and the model analysis further verifies that CCLG can effectively transfer knowledge between different languages. | [
"Cheng, Xuxin",
"Zhu, Zhihong",
"Yang, Bang",
"Zhuang, Xianwei",
"Li, Hongxiang",
"Zou, Yuexian"
]
| Cyclical Contrastive Learning Based on Geodesic for Zero-shot Cross-lingual Spoken Language Understanding | findings-acl.106 | Poster | 2311.11375v1 |
https://aclanthology.org/2024.findings-acl.107.bib | @inproceedings{liu-etal-2024-towards-safer,
title = "Towards Safer Large Language Models through Machine Unlearning",
author = "Liu, Zheyuan and
Dou, Guangyao and
Tan, Zhaoxuan and
Tian, Yijun and
Jiang, Meng",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.107",
pages = "1817--1829",
abstract = "The rapid advancement of Large Language Models (LLMs) has demonstrated their vast potential across various domains, attributed to their extensive pretraining knowledge and exceptional generalizability. However, LLMs often encounter challenges in generating harmful content when faced with problematic prompts. To address this problem, existing work attempted to implement a gradient ascent based approach to prevent LLMs from producing harmful output. While these methods can be effective, they frequently impact the model utility in responding to normal prompts. To address this gap, we introduce Selective Knowledge negation Unlearning (SKU), a novel unlearning framework for LLMs, designed to eliminate harmful knowledge while preserving utility on normal prompts. Specifically, SKU is consisted of two stages: harmful knowledge acquisition stage and knowledge negation stage. The first stage aims to identify and acquire harmful knowledge within the model, whereas the second is dedicated to remove this knowledge. SKU selectively isolates and removes harmful knowledge in model parameters, ensuring the model{'}s performance remains robust on normal prompts. Our experiments conducted across various LLM architectures demonstrate that SKU identifies a good balance point between removing harmful information and preserving utility.",
}
| The rapid advancement of Large Language Models (LLMs) has demonstrated their vast potential across various domains, attributed to their extensive pretraining knowledge and exceptional generalizability. However, LLMs often encounter challenges in generating harmful content when faced with problematic prompts. To address this problem, existing work attempted to implement a gradient ascent based approach to prevent LLMs from producing harmful output. While these methods can be effective, they frequently impact the model utility in responding to normal prompts. To address this gap, we introduce Selective Knowledge negation Unlearning (SKU), a novel unlearning framework for LLMs, designed to eliminate harmful knowledge while preserving utility on normal prompts. Specifically, SKU is consisted of two stages: harmful knowledge acquisition stage and knowledge negation stage. The first stage aims to identify and acquire harmful knowledge within the model, whereas the second is dedicated to remove this knowledge. SKU selectively isolates and removes harmful knowledge in model parameters, ensuring the model{'}s performance remains robust on normal prompts. Our experiments conducted across various LLM architectures demonstrate that SKU identifies a good balance point between removing harmful information and preserving utility. | [
"Liu, Zheyuan",
"Dou, Guangyao",
"Tan, Zhaoxuan",
"Tian, Yijun",
"Jiang, Meng"
]
| Towards Safer Large Language Models through Machine Unlearning | findings-acl.107 | Poster | 2402.10058v2 |
https://aclanthology.org/2024.findings-acl.108.bib | @inproceedings{jin-etal-2024-impact,
title = "The Impact of Reasoning Step Length on Large Language Models",
author = "Jin, Mingyu and
Yu, Qinkai and
Shu, Dong and
Zhao, Haiyan and
Hua, Wenyue and
Meng, Yanda and
Zhang, Yongfeng and
Du, Mengnan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.108",
pages = "1830--1842",
abstract = "Chain of Thought (CoT) is significant in improving the reasoning abilities of large language models (LLMs). However, the correlation between the effectiveness of CoT and the length of reasoning steps in prompts remains largely unknown. To shed light on this, we have conducted several empirical experiments to explore the relations. Specifically, we design experiments that expand and compress the rationale reasoning steps within CoT demonstrations, while keeping all other factors constant. We have the following key findings. First, the results indicate that lengthening the reasoning steps in prompts, even without adding new information into the prompt, considerably enhances LLMs{'} reasoning abilities across multiple datasets. Alternatively, shortening the reasoning steps, even while preserving the key information, significantly diminishes the reasoning abilities of models. This finding highlights the importance of the number of steps in CoT prompts and provides practical guidance to make better use of LLMs{'} potential in complex problem-solving scenarios. Second, we also investigated the relationship between the performance of CoT and the rationales used in demonstrations. Surprisingly, the result shows that even incorrect rationales can yield favorable outcomes if they maintain the requisite length of inference. Third, we observed that the advantages of increasing reasoning steps are task-dependent: simpler tasks require fewer steps, whereas complex tasks gain significantly from longer inference sequences.",
}
| Chain of Thought (CoT) is significant in improving the reasoning abilities of large language models (LLMs). However, the correlation between the effectiveness of CoT and the length of reasoning steps in prompts remains largely unknown. To shed light on this, we have conducted several empirical experiments to explore the relations. Specifically, we design experiments that expand and compress the rationale reasoning steps within CoT demonstrations, while keeping all other factors constant. We have the following key findings. First, the results indicate that lengthening the reasoning steps in prompts, even without adding new information into the prompt, considerably enhances LLMs{'} reasoning abilities across multiple datasets. Alternatively, shortening the reasoning steps, even while preserving the key information, significantly diminishes the reasoning abilities of models. This finding highlights the importance of the number of steps in CoT prompts and provides practical guidance to make better use of LLMs{'} potential in complex problem-solving scenarios. Second, we also investigated the relationship between the performance of CoT and the rationales used in demonstrations. Surprisingly, the result shows that even incorrect rationales can yield favorable outcomes if they maintain the requisite length of inference. Third, we observed that the advantages of increasing reasoning steps are task-dependent: simpler tasks require fewer steps, whereas complex tasks gain significantly from longer inference sequences. | [
"Jin, Mingyu",
"Yu, Qinkai",
"Shu, Dong",
"Zhao, Haiyan",
"Hua, Wenyue",
"Meng, Y",
"a",
"Zhang, Yongfeng",
"Du, Mengnan"
]
| The Impact of Reasoning Step Length on Large Language Models | findings-acl.108 | Poster | 2401.04925v4 |
https://aclanthology.org/2024.findings-acl.109.bib | @inproceedings{liu-etal-2024-towards-understanding,
title = "Towards Understanding Task-agnostic Debiasing Through the Lenses of Intrinsic Bias and Forgetfulness",
author = "Liu, Guangliang and
Afshari, Milad and
Zhang, Xitong and
Xue, Zhiyu and
Ghosh, Avrajit and
Bashyal, Bidhan and
Wang, Rongrong and
Johnson, Kristen",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.109",
pages = "1843--1856",
abstract = "While task-agnostic debiasing provides notable generalizability and reduced reliance on downstream data, its impact on language modeling ability and the risk of relearning social biases from downstream task-specific data remain as the two most significant challenges when debiasing Pretrained Language Models (PLMs). The impact on language modeling ability can be alleviated given a high-quality and long-contextualized debiasing corpus, but there remains a deficiency in understanding the specifics of relearning biases. We empirically ascertain that the effectiveness of task-agnostic debiasing hinges on the quantitative bias level of both the task-specific data used for downstream applications and the debiased model. We empirically show that the lower bound of the bias level of the downstream fine-tuned model can be approximated by the bias level of the debiased model, in most practical cases. To gain more in-depth understanding about how the parameters of PLMs change during fine-tuning due to the forgetting issue of PLMs, we propose a novel framework which can Propagate Socially-fair Debiasing to Downstream Fine-tuning, ProSocialTuning. Our proposed framework can push the fine-tuned model to approach the bias lower bound during downstream fine-tuning, indicating that the ineffectiveness of debiasing can be alleviated by overcoming the forgetting issue through regularizing successfully debiased attention heads based on the PLMs{'} bias levels from stages of pretraining and debiasing.",
}
| While task-agnostic debiasing provides notable generalizability and reduced reliance on downstream data, its impact on language modeling ability and the risk of relearning social biases from downstream task-specific data remain as the two most significant challenges when debiasing Pretrained Language Models (PLMs). The impact on language modeling ability can be alleviated given a high-quality and long-contextualized debiasing corpus, but there remains a deficiency in understanding the specifics of relearning biases. We empirically ascertain that the effectiveness of task-agnostic debiasing hinges on the quantitative bias level of both the task-specific data used for downstream applications and the debiased model. We empirically show that the lower bound of the bias level of the downstream fine-tuned model can be approximated by the bias level of the debiased model, in most practical cases. To gain more in-depth understanding about how the parameters of PLMs change during fine-tuning due to the forgetting issue of PLMs, we propose a novel framework which can Propagate Socially-fair Debiasing to Downstream Fine-tuning, ProSocialTuning. Our proposed framework can push the fine-tuned model to approach the bias lower bound during downstream fine-tuning, indicating that the ineffectiveness of debiasing can be alleviated by overcoming the forgetting issue through regularizing successfully debiased attention heads based on the PLMs{'} bias levels from stages of pretraining and debiasing. | [
"Liu, Guangliang",
"Afshari, Milad",
"Zhang, Xitong",
"Xue, Zhiyu",
"Ghosh, Avrajit",
"Bashyal, Bidhan",
"Wang, Rongrong",
"Johnson, Kristen"
]
| Towards Understanding Task-agnostic Debiasing Through the Lenses of Intrinsic Bias and Forgetfulness | findings-acl.109 | Poster | 2406.04146v1 |
https://aclanthology.org/2024.findings-acl.110.bib | @inproceedings{wang-etal-2024-skgsum,
title = "{SKGS}um: Structured Knowledge-Guided Document Summarization",
author = "Wang, Qiqi and
Wang, Ruofan and
Zhao, Kaiqi and
Amor, Robert and
Liu, Benjamin and
Liu, Jiamou and
Zheng, Xianda and
Huang, Zijian",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.110",
pages = "1857--1871",
abstract = "A summary structure is inherent to certain types of texts according to the Genre Theory of Linguistics. Such structures aid readers in efficiently locating information within summaries. However, most existing automatic summarization methods overlook the importance of summary structure, resulting in summaries that emphasize the most prominent information while omitting essential details from other sections. While a few summarizers recognize the importance of summary structure, they rely heavily on the predefined labels of summary structures in the source document and ground truth summaries. To address these shortcomings, we developed a Structured Knowledge-Guided Summarization (SKGSum) and its variant, SKGSum-W, which do not require structure labels. Instead, these methods rely on a set of automatically extracted summary points to generate summaries. We evaluate the proposed methods using three real-world datasets. The results indicate that our methods not only improve the quality of summaries, in terms of ROUGE and BERTScore, but also broaden the types of documents that can be effectively summarized.",
}
| A summary structure is inherent to certain types of texts according to the Genre Theory of Linguistics. Such structures aid readers in efficiently locating information within summaries. However, most existing automatic summarization methods overlook the importance of summary structure, resulting in summaries that emphasize the most prominent information while omitting essential details from other sections. While a few summarizers recognize the importance of summary structure, they rely heavily on the predefined labels of summary structures in the source document and ground truth summaries. To address these shortcomings, we developed a Structured Knowledge-Guided Summarization (SKGSum) and its variant, SKGSum-W, which do not require structure labels. Instead, these methods rely on a set of automatically extracted summary points to generate summaries. We evaluate the proposed methods using three real-world datasets. The results indicate that our methods not only improve the quality of summaries, in terms of ROUGE and BERTScore, but also broaden the types of documents that can be effectively summarized. | [
"Wang, Qiqi",
"Wang, Ruofan",
"Zhao, Kaiqi",
"Amor, Robert",
"Liu, Benjamin",
"Liu, Jiamou",
"Zheng, Xi",
"a",
"Huang, Zijian"
]
| {SKGS}um: Structured Knowledge-Guided Document Summarization | findings-acl.110 | Poster | 2203.10254v1 |
https://aclanthology.org/2024.findings-acl.111.bib | @inproceedings{zhou-etal-2024-chinese,
title = "{C}hinese Spoken Named Entity Recognition in Real-world Scenarios: Dataset and Approaches",
author = "Zhou, Shilin and
Li, Zhenghua and
Gong, Chen and
Zhang, Lei and
Hong, Yu and
Zhang, Min",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.111",
pages = "1872--1884",
abstract = "Spoken Named Entity Recognition (NER) aims to extract entities from speech. The extracted entities can help voice assistants better understand user{'}s questions and instructions. However, current Chinese Spoken NER datasets are laboratory-controlled data that are collected by reading existing texts in quiet environments, rather than natural spoken data, and the texts used for reading are also limited in topics. These limitations obstruct the development of Spoken NER in more natural and common real-world scenarios. To address this gap, we introduce a real-world Chinese Spoken NER dataset (RWCS-NER), encompassing open-domain daily conversations and task-oriented intelligent cockpit instructions. We compare several mainstream pipeline approaches on RWCS-NER. The results indicate that the current methods, affected by Automatic Speech Recognition (ASR) errors, do not perform satisfactorily in real settings. Aiming to enhance Spoken NER in real-world scenarios, we propose two approaches: self-training-asr and mapping then distilling (MDistilling). Experiments show that both approaches can achieve significant improvements, particularly MDistilling. Even compared with GPT4.0, MDistilling still reaches better results. We believe that our work will advance the field of Spoken NER in real-world settings.",
}
| Spoken Named Entity Recognition (NER) aims to extract entities from speech. The extracted entities can help voice assistants better understand user{'}s questions and instructions. However, current Chinese Spoken NER datasets are laboratory-controlled data that are collected by reading existing texts in quiet environments, rather than natural spoken data, and the texts used for reading are also limited in topics. These limitations obstruct the development of Spoken NER in more natural and common real-world scenarios. To address this gap, we introduce a real-world Chinese Spoken NER dataset (RWCS-NER), encompassing open-domain daily conversations and task-oriented intelligent cockpit instructions. We compare several mainstream pipeline approaches on RWCS-NER. The results indicate that the current methods, affected by Automatic Speech Recognition (ASR) errors, do not perform satisfactorily in real settings. Aiming to enhance Spoken NER in real-world scenarios, we propose two approaches: self-training-asr and mapping then distilling (MDistilling). Experiments show that both approaches can achieve significant improvements, particularly MDistilling. Even compared with GPT4.0, MDistilling still reaches better results. We believe that our work will advance the field of Spoken NER in real-world settings. | [
"Zhou, Shilin",
"Li, Zhenghua",
"Gong, Chen",
"Zhang, Lei",
"Hong, Yu",
"Zhang, Min"
]
| {C}hinese Spoken Named Entity Recognition in Real-world Scenarios: Dataset and Approaches | findings-acl.111 | Poster | 2202.08533v1 |
https://aclanthology.org/2024.findings-acl.112.bib | @inproceedings{kim-etal-2024-debate,
title = "{DEBATE}: Devil{'}s Advocate-Based Assessment and Text Evaluation",
author = "Kim, Alex and
Kim, Keonwoo and
Yoon, Sangwon",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.112",
pages = "1885--1897",
abstract = "As natural language generation (NLG) models have become prevalent, systematically assessing the quality of machine-generated texts has become increasingly important. Recent studies introduce LLM-based evaluators that operate as reference-free metrics, demonstrating their capability to adeptly handle novel tasks. However, these models generally rely on a single-agent approach, which, we argue, introduces an inherent limit to their performance. This is because there exist biases in LLM agent{'}s responses, including preferences for certain text structure or content. In this work, we propose DEBATE, an NLG evaluation framework based on multi-agent scoring system augmented with a concept of Devil{'}s Advocate. Within the framework, one agent is instructed to criticize other agents{'} arguments, potentially resolving the bias in LLM agent{'}s answers. DEBATE substantially outperforms the previous state-of-the-art methods in two meta-evaluation benchmarks in NLG evaluation, SummEval and TopicalChat. We also show that the extensiveness of debates among agents and the persona of an agent can influence the performance of evaluators.",
}
| As natural language generation (NLG) models have become prevalent, systematically assessing the quality of machine-generated texts has become increasingly important. Recent studies introduce LLM-based evaluators that operate as reference-free metrics, demonstrating their capability to adeptly handle novel tasks. However, these models generally rely on a single-agent approach, which, we argue, introduces an inherent limit to their performance. This is because there exist biases in LLM agent{'}s responses, including preferences for certain text structure or content. In this work, we propose DEBATE, an NLG evaluation framework based on multi-agent scoring system augmented with a concept of Devil{'}s Advocate. Within the framework, one agent is instructed to criticize other agents{'} arguments, potentially resolving the bias in LLM agent{'}s answers. DEBATE substantially outperforms the previous state-of-the-art methods in two meta-evaluation benchmarks in NLG evaluation, SummEval and TopicalChat. We also show that the extensiveness of debates among agents and the persona of an agent can influence the performance of evaluators. | [
"Kim, Alex",
"Kim, Keonwoo",
"Yoon, Sangwon"
]
| {DEBATE}: Devil{'}s Advocate-Based Assessment and Text Evaluation | findings-acl.112 | Poster | 1810.06639v4 |
https://aclanthology.org/2024.findings-acl.113.bib | @inproceedings{yang-etal-2024-large,
title = "Can Large Multimodal Models Uncover Deep Semantics Behind Images?",
author = "Yang, Yixin and
Li, Zheng and
Dong, Qingxiu and
Xia, Heming and
Sui, Zhifang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.113",
pages = "1898--1912",
abstract = "Understanding the deep semantics of images is essential in the era dominated by social media. However, current research works primarily on the superficial description of images, revealing a notable deficiency in the systematic investigation of the inherent deep semantics. In this work, we introduce DEEPEVAL, a comprehensive benchmark to assess Large Multimodal Models{'} (LMMs) capacities of visual deep semantics. DEEPEVAL includes human-annotated dataset and three progressive subtasks: fine-grained description selection, in-depth title matching, and deep semantics understanding. Utilizing DEEPEVAL, we evaluate 9 open-source LMMs and GPT-4V(ision). Our evaluation demonstrates a substantial gap between the deep semantic comprehension capabilities of existing LMMs and humans. For example, GPT-4V is 30{\%} behind humans in understanding deep semantics, even though it achieves human-comparable performance in image description. Further analysis reveals that LMM performance on DEEPEVAL varies according to the specific facets of deep semantics explored, indicating the fundamental challenges remaining in developing LMMs.",
}
| Understanding the deep semantics of images is essential in the era dominated by social media. However, current research works primarily on the superficial description of images, revealing a notable deficiency in the systematic investigation of the inherent deep semantics. In this work, we introduce DEEPEVAL, a comprehensive benchmark to assess Large Multimodal Models{'} (LMMs) capacities of visual deep semantics. DEEPEVAL includes human-annotated dataset and three progressive subtasks: fine-grained description selection, in-depth title matching, and deep semantics understanding. Utilizing DEEPEVAL, we evaluate 9 open-source LMMs and GPT-4V(ision). Our evaluation demonstrates a substantial gap between the deep semantic comprehension capabilities of existing LMMs and humans. For example, GPT-4V is 30{\%} behind humans in understanding deep semantics, even though it achieves human-comparable performance in image description. Further analysis reveals that LMM performance on DEEPEVAL varies according to the specific facets of deep semantics explored, indicating the fundamental challenges remaining in developing LMMs. | [
"Yang, Yixin",
"Li, Zheng",
"Dong, Qingxiu",
"Xia, Heming",
"Sui, Zhifang"
]
| Can Large Multimodal Models Uncover Deep Semantics Behind Images? | findings-acl.113 | Poster | 2408.03845v1 |
https://aclanthology.org/2024.findings-acl.114.bib | @inproceedings{gao-etal-2024-harvesting,
title = "Harvesting Events from Multiple Sources: Towards a Cross-Document Event Extraction Paradigm",
author = "Gao, Qiang and
Meng, Zixiang and
Li, Bobo and
Zhou, Jun and
Li, Fei and
Teng, Chong and
Ji, Donghong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.114",
pages = "1913--1927",
abstract = "Document-level event extraction aims to extract structured event information from unstructured text. However, a single document often contains limited event information and the roles of different event arguments may be biased due to the influence of the information source.This paper addresses the limitations of traditional document-level event extraction by proposing the task of cross-document event extraction (CDEE) to integrate event information from multiple documents and provide a comprehensive perspective on events. We construct a novel cross-document event extraction dataset, namely CLES, which contains 20,059 documents and 37,688 mention-level events, where over 70{\%} of them are cross-document. To address the task, we propose a CDEE pipeline that includes 5 steps, namely event extraction, coreference resolution, entity normalization, role normalization and entity-role resolution. Our CDEE pipeline achieves about 72{\%} F1 in end-to-end cross-document event extraction, suggesting the challenge of this task and setting up a benchmark for future research. Our work builds a new line of information extraction research and will attract new research attention.",
}
| Document-level event extraction aims to extract structured event information from unstructured text. However, a single document often contains limited event information and the roles of different event arguments may be biased due to the influence of the information source.This paper addresses the limitations of traditional document-level event extraction by proposing the task of cross-document event extraction (CDEE) to integrate event information from multiple documents and provide a comprehensive perspective on events. We construct a novel cross-document event extraction dataset, namely CLES, which contains 20,059 documents and 37,688 mention-level events, where over 70{\%} of them are cross-document. To address the task, we propose a CDEE pipeline that includes 5 steps, namely event extraction, coreference resolution, entity normalization, role normalization and entity-role resolution. Our CDEE pipeline achieves about 72{\%} F1 in end-to-end cross-document event extraction, suggesting the challenge of this task and setting up a benchmark for future research. Our work builds a new line of information extraction research and will attract new research attention. | [
"Gao, Qiang",
"Meng, Zixiang",
"Li, Bobo",
"Zhou, Jun",
"Li, Fei",
"Teng, Chong",
"Ji, Donghong"
]
| Harvesting Events from Multiple Sources: Towards a Cross-Document Event Extraction Paradigm | findings-acl.114 | Poster | 2406.16021v1 |
https://aclanthology.org/2024.findings-acl.115.bib | @inproceedings{hwang-etal-2024-graph,
title = "A Graph per Persona: Reasoning about Subjective Natural Language Descriptions",
author = "Hwang, EunJeong and
Shwartz, Vered and
Gutfreund, Dan and
Thost, Veronika",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.115",
pages = "1928--1942",
abstract = "Reasoning about subjective natural language descriptions, such as opinions and preferences, is a challenging topic that largely remains unsolved to date. In particular, state-of-the-art large language models (LLMs) perform disappointingly in this task, show strong biases, and do not meet the interpretability requirements often needed in these kinds of applications. We propose a novel approach for reasoning about subjective knowledge that integrates potential and implicit meanings and explicitly models the relational nature of the information. We apply supervised graph learning, offer explanations for the model{'}s reasoning, and show that our model performs well across all 15 topics of OpinionQA, outperforming several prominent LLMs. Our detailed analysis further shows its unique advantages and the complementary nature it offers in comparison to LLMs.",
}
| Reasoning about subjective natural language descriptions, such as opinions and preferences, is a challenging topic that largely remains unsolved to date. In particular, state-of-the-art large language models (LLMs) perform disappointingly in this task, show strong biases, and do not meet the interpretability requirements often needed in these kinds of applications. We propose a novel approach for reasoning about subjective knowledge that integrates potential and implicit meanings and explicitly models the relational nature of the information. We apply supervised graph learning, offer explanations for the model{'}s reasoning, and show that our model performs well across all 15 topics of OpinionQA, outperforming several prominent LLMs. Our detailed analysis further shows its unique advantages and the complementary nature it offers in comparison to LLMs. | [
"Hwang, EunJeong",
"Shwartz, Vered",
"Gutfreund, Dan",
"Thost, Veronika"
]
| A Graph per Persona: Reasoning about Subjective Natural Language Descriptions | findings-acl.115 | Poster | 1508.03843v1 |
https://aclanthology.org/2024.findings-acl.116.bib | @inproceedings{fang-etal-2024-moltc,
title = "{M}ol{TC}: Towards Molecular Relational Modeling In Language Models",
author = "Fang, Junfeng and
Zhang, Shuai and
Wu, Chang and
Yang, Zhengyi and
Liu, Zhiyuan and
Li, Sihang and
Wang, Kun and
Du, Wenjie and
Wang, Xiang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.116",
pages = "1943--1958",
abstract = "Molecular Relational Learning (MRL), aiming to understand interactions between molecular pairs, plays a pivotal role in advancing biochemical research. Recently, the adoption of large language models (LLMs), known for their vast knowledge repositories and advanced logical inference capabilities, has emerged as a promising way for efficient and effective MRL. Despite their potential, these methods predominantly rely on textual data, thus not fully harnessing the wealth of structural information inherent in molecular graphs. Moreover, the absence of a unified framework exacerbates the issue of insufficient data exploitation, as it hinders the sharing of interaction mechanism learned across various datasets. To address these challenges, this work proposes a novel LLM-based multi-modal framework for molecular interaction modeling following Chain-of-Thought (CoT) theory, termed MolTC, which effectively integrate graphical information of two molecules in pair. To train this integrated framework efficiently, we introduce a *multi-hierarchical CoT theory* to refine its training paradigm, and conduct a comprehensive *Molecular Interactive Instructions* dataset for the development of biochemical LLMs involving MRL.Our experiments,conducted across various datasets involving over 4,000,000 molecular pairs, exhibit the superiority of our method over current GNN and LLM-based baselines. Code is available at https://github.com/MangoKiller/MolTC.",
}
| Molecular Relational Learning (MRL), aiming to understand interactions between molecular pairs, plays a pivotal role in advancing biochemical research. Recently, the adoption of large language models (LLMs), known for their vast knowledge repositories and advanced logical inference capabilities, has emerged as a promising way for efficient and effective MRL. Despite their potential, these methods predominantly rely on textual data, thus not fully harnessing the wealth of structural information inherent in molecular graphs. Moreover, the absence of a unified framework exacerbates the issue of insufficient data exploitation, as it hinders the sharing of interaction mechanism learned across various datasets. To address these challenges, this work proposes a novel LLM-based multi-modal framework for molecular interaction modeling following Chain-of-Thought (CoT) theory, termed MolTC, which effectively integrate graphical information of two molecules in pair. To train this integrated framework efficiently, we introduce a *multi-hierarchical CoT theory* to refine its training paradigm, and conduct a comprehensive *Molecular Interactive Instructions* dataset for the development of biochemical LLMs involving MRL.Our experiments,conducted across various datasets involving over 4,000,000 molecular pairs, exhibit the superiority of our method over current GNN and LLM-based baselines. Code is available at https://github.com/MangoKiller/MolTC. | [
"Fang, Junfeng",
"Zhang, Shuai",
"Wu, Chang",
"Yang, Zhengyi",
"Liu, Zhiyuan",
"Li, Sihang",
"Wang, Kun",
"Du, Wenjie",
"Wang, Xiang"
]
| {M}ol{TC}: Towards Molecular Relational Modeling In Language Models | findings-acl.116 | Poster | 2311.16208v1 |
https://aclanthology.org/2024.findings-acl.117.bib | @inproceedings{wu-etal-2024-kpeval,
title = "{KPE}val: Towards Fine-Grained Semantic-Based Keyphrase Evaluation",
author = "Wu, Di and
Yin, Da and
Chang, Kai-Wei",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.117",
pages = "1959--1981",
abstract = "Despite the significant advancements in keyphrase extraction and keyphrase generation methods, the predominant approach for evaluation mainly relies on exact matching with human references. This scheme fails to recognize systems that generate keyphrases semantically equivalent to the references or diverse keyphrases that carry practical utility. To better assess the capability of keyphrase systems, we propose KPEval, a comprehensive evaluation framework consisting of four critical aspects: reference agreement, faithfulness, diversity, and utility. For each aspect, we design semantic-based metrics to reflect the evaluation objectives. Meta-evaluation studies demonstrate that our evaluation strategy correlates better with human preferences compared to a range of previously proposed metrics. Using KPEval, we re-evaluate 23 keyphrase systems and discover that (1) established model comparison results have blind-spots especially when considering reference-free evaluation; (2) large language models are underestimated by prior evaluation works; and (3) there is no single best model that can excel in all the aspects.",
}
| Despite the significant advancements in keyphrase extraction and keyphrase generation methods, the predominant approach for evaluation mainly relies on exact matching with human references. This scheme fails to recognize systems that generate keyphrases semantically equivalent to the references or diverse keyphrases that carry practical utility. To better assess the capability of keyphrase systems, we propose KPEval, a comprehensive evaluation framework consisting of four critical aspects: reference agreement, faithfulness, diversity, and utility. For each aspect, we design semantic-based metrics to reflect the evaluation objectives. Meta-evaluation studies demonstrate that our evaluation strategy correlates better with human preferences compared to a range of previously proposed metrics. Using KPEval, we re-evaluate 23 keyphrase systems and discover that (1) established model comparison results have blind-spots especially when considering reference-free evaluation; (2) large language models are underestimated by prior evaluation works; and (3) there is no single best model that can excel in all the aspects. | [
"Wu, Di",
"Yin, Da",
"Chang, Kai-Wei"
]
| {KPE}val: Towards Fine-Grained Semantic-Based Keyphrase Evaluation | findings-acl.117 | Poster | 9902007v1 |
https://aclanthology.org/2024.findings-acl.118.bib | @inproceedings{li-etal-2024-learning,
title = "Learning Low-dimensional Multi-domain Knowledge Graph Embedding via Dual Archimedean Spirals",
author = "Li, Jiang and
Su, Xiangdong and
Zhang, Fujun and
Gao, Guanglai",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.118",
pages = "1982--1994",
abstract = "Knowledge graph embedding (KGE) is extensively employed for link prediction by representing entities and relations as low-dimensional vectors. In real-world scenarios, knowledge graphs (KGs) usually encompass diverse domains, which poses challenges to KG representations. However, existing KGE methods rarely make domain constraints on the embedding distribution of multi-domain KGs, leading to the embedding overlapping of different domains and performance degradation of link prediction. To address this challenge, we propose \textbf{Du}al \textbf{A}rchimedean \textbf{S}piral Knowledge Graph \textbf{E}mbedding (DuASE), a low-dimensional KGE model for multi-domain KGs. DuASE is inspired by our discovery that relation types can distinguish entities from different domains. Specifically, DuASE encodes entities with the same relation on the same Archimedean spiral, allowing it to differentiate the entities from different domains. To avoid embedding overlapping across domains, DuASE further makes the head and the tail spirals in the same triplet cluster to their respective domain space by a regularization function. Thus, DuASE can better capture the domain information and the dependencies between entities when modeling the multi-domain KGs, leading to improved KG representations. We validate the effectiveness of DuASE on the novel multi-domain dataset ($n$-MDKG) introduced in this study and three other benchmark datasets.",
}
| Knowledge graph embedding (KGE) is extensively employed for link prediction by representing entities and relations as low-dimensional vectors. In real-world scenarios, knowledge graphs (KGs) usually encompass diverse domains, which poses challenges to KG representations. However, existing KGE methods rarely make domain constraints on the embedding distribution of multi-domain KGs, leading to the embedding overlapping of different domains and performance degradation of link prediction. To address this challenge, we propose \textbf{Du}al \textbf{A}rchimedean \textbf{S}piral Knowledge Graph \textbf{E}mbedding (DuASE), a low-dimensional KGE model for multi-domain KGs. DuASE is inspired by our discovery that relation types can distinguish entities from different domains. Specifically, DuASE encodes entities with the same relation on the same Archimedean spiral, allowing it to differentiate the entities from different domains. To avoid embedding overlapping across domains, DuASE further makes the head and the tail spirals in the same triplet cluster to their respective domain space by a regularization function. Thus, DuASE can better capture the domain information and the dependencies between entities when modeling the multi-domain KGs, leading to improved KG representations. We validate the effectiveness of DuASE on the novel multi-domain dataset ($n$-MDKG) introduced in this study and three other benchmark datasets. | [
"Li, Jiang",
"Su, Xiangdong",
"Zhang, Fujun",
"Gao, Guanglai"
]
| Learning Low-dimensional Multi-domain Knowledge Graph Embedding via Dual Archimedean Spirals | findings-acl.118 | Poster | 1909.03794v1 |
https://aclanthology.org/2024.findings-acl.119.bib | @inproceedings{wang-etal-2024-lora,
title = "{L}o{RA} Meets Dropout under a Unified Framework",
author = "Wang, Sheng and
Chen, Liheng and
Jiang, Jiyue and
Xue, Boyang and
Kong, Lingpeng and
Wu, Chuan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.119",
pages = "1995--2008",
abstract = "With the remarkable capabilities, large language models (LLMs) have emergedas essential elements in numerous NLP applications, while parameter-efficientfinetuning, especially LoRA, has gained popularity as a lightweight approachfor model customization. Meanwhile, various dropout methods, initially designedfor full finetuning with all the parameters updated, alleviates overfittingassociated with excessive parameter redundancy. Hence, a possible contradictionarises from negligible trainable parameters of LoRA and the effectiveness ofprevious dropout methods, which has been largely overlooked. To fill this gap,we first confirm that parameter-efficient LoRA is also overfitting-prone. Wethen revisit transformer-specific dropout methods, and establish theirequivalence and distinctions mathematically and empirically. Building upon thiscomparative analysis, we introduce a unified framework for a comprehensiveinvestigation, which instantiates these methods based on dropping position,structural pattern and compensation measure. Through this framework, we revealthe new preferences and performance comparisons of them when involved withlimited trainable parameters. This framework also allows us to amalgamate themost favorable aspects into a novel dropout method named HiddenKey. Extensiveexperiments verify the remarkable superiority and sufficiency of HiddenKeyacross multiple models and tasks, which highlights it as the preferred approachfor high-performance and parameter-efficient finetuning of LLMs.",
}
| With the remarkable capabilities, large language models (LLMs) have emergedas essential elements in numerous NLP applications, while parameter-efficientfinetuning, especially LoRA, has gained popularity as a lightweight approachfor model customization. Meanwhile, various dropout methods, initially designedfor full finetuning with all the parameters updated, alleviates overfittingassociated with excessive parameter redundancy. Hence, a possible contradictionarises from negligible trainable parameters of LoRA and the effectiveness ofprevious dropout methods, which has been largely overlooked. To fill this gap,we first confirm that parameter-efficient LoRA is also overfitting-prone. Wethen revisit transformer-specific dropout methods, and establish theirequivalence and distinctions mathematically and empirically. Building upon thiscomparative analysis, we introduce a unified framework for a comprehensiveinvestigation, which instantiates these methods based on dropping position,structural pattern and compensation measure. Through this framework, we revealthe new preferences and performance comparisons of them when involved withlimited trainable parameters. This framework also allows us to amalgamate themost favorable aspects into a novel dropout method named HiddenKey. Extensiveexperiments verify the remarkable superiority and sufficiency of HiddenKeyacross multiple models and tasks, which highlights it as the preferred approachfor high-performance and parameter-efficient finetuning of LLMs. | [
"Wang, Sheng",
"Chen, Liheng",
"Jiang, Jiyue",
"Xue, Boyang",
"Kong, Lingpeng",
"Wu, Chuan"
]
| {L}o{RA} Meets Dropout under a Unified Framework | findings-acl.119 | Poster | 2403.00812v2 |
https://aclanthology.org/2024.findings-acl.120.bib | @inproceedings{mao-etal-2024-enhancing,
title = "Enhancing Text-to-{SQL} Parsing through Question Rewriting and Execution-Guided Refinement",
author = "Mao, Wenxin and
Wang, Ruiqi and
Guo, Jiyu and
Zeng, Jichuan and
Gao, Cuiyun and
Han, Peiyi and
Liu, Chuanyi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.120",
pages = "2009--2024",
abstract = "Large Language Model (LLM)-based approach has become the mainstream for Text-to-SQL task and achieves remarkable performance. In this paper, we augment the existing prompt engineering methods by exploiting the database content and execution feedback. Specifically, we introduce DART-SQL, which comprises two key components: (1) Question Rewriting: DART-SQL rewrites natural language questions by leveraging database content information to eliminate ambiguity. (2) Execution-Guided Refinement: DART-SQL incorporates database content information and utilizes the execution results of the generated SQL to iteratively refine the SQL. We apply this framework to the two LLM-based approaches (DAIL-SQL and C3) and test it on four widely used benchmarks (Spider-dev, Spider-test, Realistic and DK). Experiments show that our framework for DAIL-SQL and C3 achieves an average improvement of 12.41{\%} and 5.38{\%}, respectively, in terms of execution accuracy(EX) metric.",
}
| Large Language Model (LLM)-based approach has become the mainstream for Text-to-SQL task and achieves remarkable performance. In this paper, we augment the existing prompt engineering methods by exploiting the database content and execution feedback. Specifically, we introduce DART-SQL, which comprises two key components: (1) Question Rewriting: DART-SQL rewrites natural language questions by leveraging database content information to eliminate ambiguity. (2) Execution-Guided Refinement: DART-SQL incorporates database content information and utilizes the execution results of the generated SQL to iteratively refine the SQL. We apply this framework to the two LLM-based approaches (DAIL-SQL and C3) and test it on four widely used benchmarks (Spider-dev, Spider-test, Realistic and DK). Experiments show that our framework for DAIL-SQL and C3 achieves an average improvement of 12.41{\%} and 5.38{\%}, respectively, in terms of execution accuracy(EX) metric. | [
"Mao, Wenxin",
"Wang, Ruiqi",
"Guo, Jiyu",
"Zeng, Jichuan",
"Gao, Cuiyun",
"Han, Peiyi",
"Liu, Chuanyi"
]
| Enhancing Text-to-{SQL} Parsing through Question Rewriting and Execution-Guided Refinement | findings-acl.120 | Poster | 2406.10991v1 |
https://aclanthology.org/2024.findings-acl.121.bib | @inproceedings{zhang-etal-2024-knowledge-alignment,
title = "The Knowledge Alignment Problem: Bridging Human and External Knowledge for Large Language Models",
author = "Zhang, Shuo and
Pan, Liangming and
Zhao, Junzhou and
Wang, William Yang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.121",
pages = "2025--2038",
abstract = "Large language models often necessitate grounding on external knowledge to generate faithful and reliable answers. Yet even with the correct groundings in the reference, they can ignore them and rely on wrong groundings or their inherent biases to hallucinate when users, being largely unaware of the specifics of the stored information, pose questions that might not directly correlate with the retrieved groundings. In this work, we formulate this knowledge alignment problem and introduce MixAlign, a framework that interacts with both the human user and the knowledge base to obtain and integrate clarifications on how the user question relates to the stored information. MixAlign employs a language model to achieve automatic knowledge alignment and, if necessary, further enhances this alignment through human user clarifications. Experimental results highlight the crucial role of knowledge alignment in boosting model performance and mitigating hallucination, with improvements noted up to 22.2{\%} and 27.1{\%} respectively. We also demonstrate the effectiveness of MixAlign in improving knowledge alignment by producing high-quality, user-centered clarifications.",
}
| Large language models often necessitate grounding on external knowledge to generate faithful and reliable answers. Yet even with the correct groundings in the reference, they can ignore them and rely on wrong groundings or their inherent biases to hallucinate when users, being largely unaware of the specifics of the stored information, pose questions that might not directly correlate with the retrieved groundings. In this work, we formulate this knowledge alignment problem and introduce MixAlign, a framework that interacts with both the human user and the knowledge base to obtain and integrate clarifications on how the user question relates to the stored information. MixAlign employs a language model to achieve automatic knowledge alignment and, if necessary, further enhances this alignment through human user clarifications. Experimental results highlight the crucial role of knowledge alignment in boosting model performance and mitigating hallucination, with improvements noted up to 22.2{\%} and 27.1{\%} respectively. We also demonstrate the effectiveness of MixAlign in improving knowledge alignment by producing high-quality, user-centered clarifications. | [
"Zhang, Shuo",
"Pan, Liangming",
"Zhao, Junzhou",
"Wang, William Yang"
]
| The Knowledge Alignment Problem: Bridging Human and External Knowledge for Large Language Models | findings-acl.121 | Poster | 2305.13669v3 |
https://aclanthology.org/2024.findings-acl.122.bib | @inproceedings{luo-etal-2024-chatkbqa,
title = "{C}hat{KBQA}: A Generate-then-Retrieve Framework for Knowledge Base Question Answering with Fine-tuned Large Language Models",
author = "Luo, Haoran and
E, Haihong and
Tang, Zichen and
Peng, Shiyao and
Guo, Yikai and
Zhang, Wentai and
Ma, Chenghao and
Dong, Guanting and
Song, Meina and
Lin, Wei and
Zhu, Yifan and
Luu, Anh Tuan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.122",
pages = "2039--2056",
abstract = "Knowledge Base Question Answering (KBQA) aims to answer natural language questions over large-scale knowledge bases (KBs), which can be summarized into two crucial steps: knowledge retrieval and semantic parsing. However, three core challenges remain: inefficient knowledge retrieval, mistakes of retrieval adversely impacting semantic parsing, and the complexity of previous KBQA methods. To tackle these challenges, we introduce ChatKBQA, a novel and simple generate-then-retrieve KBQA framework, which proposes first generating the logical form with fine-tuned LLMs, then retrieving and replacing entities and relations with an unsupervised retrieval method, to improve both generation and retrieval more directly. Experimental results show that ChatKBQA achieves new state-of-the-art performance on standard KBQA datasets, WebQSP, and CWQ. This work can also be regarded as a new paradigm for combining LLMs with knowledge graphs (KGs) for interpretable and knowledge-required question answering.",
}
| Knowledge Base Question Answering (KBQA) aims to answer natural language questions over large-scale knowledge bases (KBs), which can be summarized into two crucial steps: knowledge retrieval and semantic parsing. However, three core challenges remain: inefficient knowledge retrieval, mistakes of retrieval adversely impacting semantic parsing, and the complexity of previous KBQA methods. To tackle these challenges, we introduce ChatKBQA, a novel and simple generate-then-retrieve KBQA framework, which proposes first generating the logical form with fine-tuned LLMs, then retrieving and replacing entities and relations with an unsupervised retrieval method, to improve both generation and retrieval more directly. Experimental results show that ChatKBQA achieves new state-of-the-art performance on standard KBQA datasets, WebQSP, and CWQ. This work can also be regarded as a new paradigm for combining LLMs with knowledge graphs (KGs) for interpretable and knowledge-required question answering. | [
"Luo, Haoran",
"E, Haihong",
"Tang, Zichen",
"Peng, Shiyao",
"Guo, Yikai",
"Zhang, Wentai",
"Ma, Chenghao",
"Dong, Guanting",
"Song, Meina",
"Lin, Wei",
"Zhu, Yifan",
"Luu, Anh Tuan"
]
| {C}hat{KBQA}: A Generate-then-Retrieve Framework for Knowledge Base Question Answering with Fine-tuned Large Language Models | findings-acl.122 | Poster | 2402.02503v1 |
https://aclanthology.org/2024.findings-acl.123.bib | @inproceedings{wang-etal-2024-achilles,
title = "Achilles-Bench: A Challenging Benchmark for Low-Resource Evaluation",
author = "Wang, Yudong and
Ma, Chang and
Dong, Qingxiu and
Sui, Zhifang and
Kong, Lingpeng and
Xu, Jingjing",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.123",
pages = "2057--2080",
abstract = "With promising yet saturated results in high-resource settings, low-resource datasets have gradually become crucial benchmarks (e.g., BigBench Hard, superGLUE) for evaluating the learning ability of advanced neural networks. In this work, we find that there exists a set of {``}hard examples{''} in low-resource settings that challenge neural networks but are not well evaluated, which causes over-estimated performance. We first give a theoretical analysis on which factors bring the difficulty of low-resource learning. It then motivates us to propose a challenging benchmark Achilles-Bench to better evaluate the learning ability, which covers 11 datasets, including 8 natural language process (NLP) datasets and 3 computer vision (CV) datasets. Experiments on a wide range of models show that neural networks, even pre-trained language models, have sharp performance drops on our benchmark, demonstrating the effectiveness of evaluating the weaknesses of neural networks. On NLP tasks, we surprisingly find that despite better results on traditional low-resource benchmarks, pre-trained networks, does not show performance improvements on our benchmarks. there is still a large robustness gap between existing models and human-level performance, highlighting the need for robust low-resource learning models.",
}
| With promising yet saturated results in high-resource settings, low-resource datasets have gradually become crucial benchmarks (e.g., BigBench Hard, superGLUE) for evaluating the learning ability of advanced neural networks. In this work, we find that there exists a set of {``}hard examples{''} in low-resource settings that challenge neural networks but are not well evaluated, which causes over-estimated performance. We first give a theoretical analysis on which factors bring the difficulty of low-resource learning. It then motivates us to propose a challenging benchmark Achilles-Bench to better evaluate the learning ability, which covers 11 datasets, including 8 natural language process (NLP) datasets and 3 computer vision (CV) datasets. Experiments on a wide range of models show that neural networks, even pre-trained language models, have sharp performance drops on our benchmark, demonstrating the effectiveness of evaluating the weaknesses of neural networks. On NLP tasks, we surprisingly find that despite better results on traditional low-resource benchmarks, pre-trained networks, does not show performance improvements on our benchmarks. there is still a large robustness gap between existing models and human-level performance, highlighting the need for robust low-resource learning models. | [
"Wang, Yudong",
"Ma, Chang",
"Dong, Qingxiu",
"Sui, Zhifang",
"Kong, Lingpeng",
"Xu, Jingjing"
]
| Achilles-Bench: A Challenging Benchmark for Low-Resource Evaluation | findings-acl.123 | Poster | 2402.19472v1 |
https://aclanthology.org/2024.findings-acl.124.bib | @inproceedings{wang-etal-2024-intervenor,
title = "{INTERVENOR}: Prompting the Coding Ability of Large Language Models with the Interactive Chain of Repair",
author = "Wang, Hanbin and
Liu, Zhenghao and
Wang, Shuo and
Cui, Ganqu and
Ding, Ning and
Liu, Zhiyuan and
Yu, Ge",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.124",
pages = "2081--2107",
abstract = "This paper introduces INTERVENOR (INTERactiVE chaiN Of Repair), a system designed to emulate the interactive code repair processes observed in humans, encompassing both code diagnosis and code repair. INTERVENOR prompts Large Language Models (LLMs) to play distinct roles during the code repair process, functioning as both a Code Learner and a Code Teacher. Specifically, the Code Learner is tasked with adhering to instructions to generate or repair code, while the Code Teacher is responsible for crafting a Chain-of-Repair (CoR) to serve as guidance for the Code Learner. During generating the CoR, the Code Teacher needs to check the generated codes from Code Learner and reassess how to address code bugs based on error feedback received from compilers. Experimental results demonstrate that INTERVENOR surpasses baseline models, exhibiting improvements of approximately 18{\%} and 4.3{\%} over GPT-3.5 in code generation and code translation tasks, respectively. Our further analyses show that CoR is effective to illuminate the reasons behind bugs and outline solution plans in natural language. With the feedback of code compilers, INTERVENOR can accurately identify syntax errors and assertion errors and provide precise instructions to repair codes. All data and codes are available at [https://github.com/NEUIR/INTERVENOR](https://github.com/NEUIR/INTERVENOR).",
}
| This paper introduces INTERVENOR (INTERactiVE chaiN Of Repair), a system designed to emulate the interactive code repair processes observed in humans, encompassing both code diagnosis and code repair. INTERVENOR prompts Large Language Models (LLMs) to play distinct roles during the code repair process, functioning as both a Code Learner and a Code Teacher. Specifically, the Code Learner is tasked with adhering to instructions to generate or repair code, while the Code Teacher is responsible for crafting a Chain-of-Repair (CoR) to serve as guidance for the Code Learner. During generating the CoR, the Code Teacher needs to check the generated codes from Code Learner and reassess how to address code bugs based on error feedback received from compilers. Experimental results demonstrate that INTERVENOR surpasses baseline models, exhibiting improvements of approximately 18{\%} and 4.3{\%} over GPT-3.5 in code generation and code translation tasks, respectively. Our further analyses show that CoR is effective to illuminate the reasons behind bugs and outline solution plans in natural language. With the feedback of code compilers, INTERVENOR can accurately identify syntax errors and assertion errors and provide precise instructions to repair codes. All data and codes are available at [https://github.com/NEUIR/INTERVENOR](https://github.com/NEUIR/INTERVENOR). | [
"Wang, Hanbin",
"Liu, Zhenghao",
"Wang, Shuo",
"Cui, Ganqu",
"Ding, Ning",
"Liu, Zhiyuan",
"Yu, Ge"
]
| {INTERVENOR}: Prompting the Coding Ability of Large Language Models with the Interactive Chain of Repair | findings-acl.124 | Poster | 2311.09868v5 |
https://aclanthology.org/2024.findings-acl.125.bib | @inproceedings{chen-etal-2024-socialbench,
title = "{S}ocial{B}ench: Sociality Evaluation of Role-Playing Conversational Agents",
author = "Chen, Hongzhan and
Chen, Hehong and
Yan, Ming and
Xu, Wenshen and
Xing, Gao and
Shen, Weizhou and
Quan, Xiaojun and
Li, Chenliang and
Zhang, Ji and
Huang, Fei",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.125",
pages = "2108--2126",
abstract = "Large language models (LLMs) have advanced the development of various AI conversational agents, including role-playing agents that mimic diverse characters and human behaviors. While prior research has predominantly focused on enhancing the conversational capability, role-specific knowledge and style of these agents, there has been a noticeable gap in assessing their social intelligence. In this paper, we introduce SocialBench, the first benchmark designed to systematically evaluate the sociality of role-playing agents at both individual and group levels of social interactions. SocialBench is constructed from various sources and covers a wide range of 500 characters and over 6,000 question prompts and 30,800 multi-turn role-playing utterances. We conduct comprehensive evaluations on this benchmark using mainstream LLMs. We find that agents excelling in individual level does not imply their proficiency in group level. Experimental results on SocialBench confirm its significance as a testbed for assessing the social interaction of role-playing agents. The benchmark is publicly accessible at https://github.com/X-PLUG/RoleInteract.",
}
| Large language models (LLMs) have advanced the development of various AI conversational agents, including role-playing agents that mimic diverse characters and human behaviors. While prior research has predominantly focused on enhancing the conversational capability, role-specific knowledge and style of these agents, there has been a noticeable gap in assessing their social intelligence. In this paper, we introduce SocialBench, the first benchmark designed to systematically evaluate the sociality of role-playing agents at both individual and group levels of social interactions. SocialBench is constructed from various sources and covers a wide range of 500 characters and over 6,000 question prompts and 30,800 multi-turn role-playing utterances. We conduct comprehensive evaluations on this benchmark using mainstream LLMs. We find that agents excelling in individual level does not imply their proficiency in group level. Experimental results on SocialBench confirm its significance as a testbed for assessing the social interaction of role-playing agents. The benchmark is publicly accessible at https://github.com/X-PLUG/RoleInteract. | [
"Chen, Hongzhan",
"Chen, Hehong",
"Yan, Ming",
"Xu, Wenshen",
"Xing, Gao",
"Shen, Weizhou",
"Quan, Xiaojun",
"Li, Chenliang",
"Zhang, Ji",
"Huang, Fei"
]
| {S}ocial{B}ench: Sociality Evaluation of Role-Playing Conversational Agents | findings-acl.125 | Poster | 2406.00627v1 |
https://aclanthology.org/2024.findings-acl.126.bib | @inproceedings{ma-etal-2024-model,
title = "From Model-centered to Human-Centered: Revision Distance as a Metric for Text Evaluation in {LLM}s-based Applications",
author = "Ma, Yongqiang and
Qing, Lizhi and
Liu, Jiawei and
Kang, Yangyang and
Zhang, Yue and
Lu, Wei and
Liu, Xiaozhong and
Cheng, Qikai",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.126",
pages = "2127--2137",
abstract = "Evaluating large language models (LLMs) is fundamental, particularly in the context of practical applications. Conventional evaluation methods, typically designed primarily for LLM development, yield numerical scores that ignore the user experience. Therefore, our study shifts the focus from model-centered to human-centered evaluation in the context of AI-powered writing assistance applications. Our proposed metric, termed {``}Revision Distance,{''} utilizes LLMs to suggest revision edits that mimic the human writing process. It is determined by counting the revision edits generated by LLMs. Benefiting from the generated revision edit details, our metric can provide a self-explained text evaluation result in a human-understandable manner beyond the context-independent score. Our results show that for the easy-writing task, {``}Revision Distance{''} is consistent with established metrics (ROUGE, Bert-score, and GPT-score), but offers more insightful, detailed feedback and better distinguishes between texts. Moreover, in the context of challenging academic writing tasks, our metric still delivers reliable evaluations where other metrics tend to struggle. Furthermore, our metric also holds significant potential for scenarios lacking reference texts.",
}
| Evaluating large language models (LLMs) is fundamental, particularly in the context of practical applications. Conventional evaluation methods, typically designed primarily for LLM development, yield numerical scores that ignore the user experience. Therefore, our study shifts the focus from model-centered to human-centered evaluation in the context of AI-powered writing assistance applications. Our proposed metric, termed {``}Revision Distance,{''} utilizes LLMs to suggest revision edits that mimic the human writing process. It is determined by counting the revision edits generated by LLMs. Benefiting from the generated revision edit details, our metric can provide a self-explained text evaluation result in a human-understandable manner beyond the context-independent score. Our results show that for the easy-writing task, {``}Revision Distance{''} is consistent with established metrics (ROUGE, Bert-score, and GPT-score), but offers more insightful, detailed feedback and better distinguishes between texts. Moreover, in the context of challenging academic writing tasks, our metric still delivers reliable evaluations where other metrics tend to struggle. Furthermore, our metric also holds significant potential for scenarios lacking reference texts. | [
"Ma, Yongqiang",
"Qing, Lizhi",
"Liu, Jiawei",
"Kang, Yangyang",
"Zhang, Yue",
"Lu, Wei",
"Liu, Xiaozhong",
"Cheng, Qikai"
]
| From Model-centered to Human-Centered: Revision Distance as a Metric for Text Evaluation in {LLM}s-based Applications | findings-acl.126 | Poster | 2310.14829v1 |
https://aclanthology.org/2024.findings-acl.127.bib | @inproceedings{guo-etal-2024-context,
title = "Context-Aware Tracking and Dynamic Introduction for Incomplete Utterance Rewriting in Extended Multi-Turn Dialogues",
author = "Guo, Xinnan and
Zhu, Qian and
Shi, Qiuhui and
Lin, Xuan and
Wang, Liubin and
DaqianLi, DaqianLi and
Chen, Yongrui",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.127",
pages = "2138--2148",
abstract = "Incomplete utterance rewriting (IUR) aims to reconstruct the utterance with omitted information and pronouns to be standalone and complete based on the context. The existing works predominantly focus on simple ellipsis and coreference problems in brief multi-turn dialogues. But in actual scenarios: 1) the context of the dialogues frequently comprises multiple similar candidates for ellipsis and coreference resolution, pouring to confuse. 2) the number of turns tends to be more extensive, while the content with various topics also grows more complex. This paper proposes a novel method called CaT to address these issues. In particular, we first devise a tacker model, distilled from GPT4-turbo, to adopt Context Tracking that dynamically updates a list of key phrases turn by turn, as accurate candidates for ellipsis and coreference resolution. Second, we further present the Dynamic Context Introduction mechanism to filter irrelevant preceding contexts that are not relied on by any element within the key phrase list to condense extended dialogues. Comprehensive experiments indicate that our solution provides a significant improvement over the existing baselines, and achieves state-of-the-art on three benchmarks.",
}
| Incomplete utterance rewriting (IUR) aims to reconstruct the utterance with omitted information and pronouns to be standalone and complete based on the context. The existing works predominantly focus on simple ellipsis and coreference problems in brief multi-turn dialogues. But in actual scenarios: 1) the context of the dialogues frequently comprises multiple similar candidates for ellipsis and coreference resolution, pouring to confuse. 2) the number of turns tends to be more extensive, while the content with various topics also grows more complex. This paper proposes a novel method called CaT to address these issues. In particular, we first devise a tacker model, distilled from GPT4-turbo, to adopt Context Tracking that dynamically updates a list of key phrases turn by turn, as accurate candidates for ellipsis and coreference resolution. Second, we further present the Dynamic Context Introduction mechanism to filter irrelevant preceding contexts that are not relied on by any element within the key phrase list to condense extended dialogues. Comprehensive experiments indicate that our solution provides a significant improvement over the existing baselines, and achieves state-of-the-art on three benchmarks. | [
"Guo, Xinnan",
"Zhu, Qian",
"Shi, Qiuhui",
"Lin, Xuan",
"Wang, Liubin",
"DaqianLi, DaqianLi",
"Chen, Yongrui"
]
| Context-Aware Tracking and Dynamic Introduction for Incomplete Utterance Rewriting in Extended Multi-Turn Dialogues | findings-acl.127 | Poster | 2310.20479v1 |
https://aclanthology.org/2024.findings-acl.128.bib | @inproceedings{chen-etal-2024-emotionqueen,
title = "{E}motion{Q}ueen: A Benchmark for Evaluating Empathy of Large Language Models",
author = "Chen, Yuyan and
Yan, Songzhou and
Liu, Sijia and
Li, Yueze and
Xiao, Yanghua",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.128",
pages = "2149--2176",
abstract = "Emotional intelligence in large language models (LLMs) is of great importance in Natural Language Processing. However, the previous research mainly focus on basic sentiment analysis tasks, such as emotion recognition, which is not enough to evaluate LLMs{'} overall emotional intelligence. Therefore, this paper presents a novel framework named EmotionQueen for evaluating the emotional intelligence of LLMs. The framework includes four distinctive tasks: Key Event Recognition, Mixed Event Recognition, Implicit Emotional Recognition, and Intention Recognition. LLMs are requested to recognize important event or implicit emotions and generate empathetic response.We also design two metrics to evaluate LLMs{'} capabilities in recognition and response for emotion-related statements. Experiments yield significant conclusions about LLMs{'} capabilities and limitations in emotion intelligence.",
}
| Emotional intelligence in large language models (LLMs) is of great importance in Natural Language Processing. However, the previous research mainly focus on basic sentiment analysis tasks, such as emotion recognition, which is not enough to evaluate LLMs{'} overall emotional intelligence. Therefore, this paper presents a novel framework named EmotionQueen for evaluating the emotional intelligence of LLMs. The framework includes four distinctive tasks: Key Event Recognition, Mixed Event Recognition, Implicit Emotional Recognition, and Intention Recognition. LLMs are requested to recognize important event or implicit emotions and generate empathetic response.We also design two metrics to evaluate LLMs{'} capabilities in recognition and response for emotion-related statements. Experiments yield significant conclusions about LLMs{'} capabilities and limitations in emotion intelligence. | [
"Chen, Yuyan",
"Yan, Songzhou",
"Liu, Sijia",
"Li, Yueze",
"Xiao, Yanghua"
]
| {E}motion{Q}ueen: A Benchmark for Evaluating Empathy of Large Language Models | findings-acl.128 | Poster | 2407.18538v1 |
https://aclanthology.org/2024.findings-acl.129.bib | @inproceedings{pan-etal-2024-plum,
title = "Plum: Prompt Learning using Metaheuristics",
author = "Pan, Rui and
Xing, Shuo and
Diao, Shizhe and
Sun, Wenhe and
Liu, Xiang and
Shum, KaShun and
Zhang, Jipeng and
Pi, Renjie and
Zhang, Tong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.129",
pages = "2177--2197",
abstract = "Since the emergence of large language models, prompt learning has become a popular method for optimizing and customizing these models. Special prompts, such as Chain-of-Thought, have even revealed previously unknown reasoning capabilities within these models. However, the progress of discovering effective prompts has been slow, driving a desire for general prompt optimization methods. Unfortunately, few existing prompt learning methods satisfy the criteria of being truly {``}general{''}, i.e., automatic, discrete, black-box, gradient-free, and interpretable all at once. In this paper, we introduce metaheuristics, a branch of discrete non-convex optimization methods with over 100 options, as a promising approach to prompt learning. Within our paradigm, we test six typical methods: hill climbing, simulated annealing, genetic algorithms with/without crossover, tabu search, and harmony search, demonstrating their effectiveness in white-box and black-box prompt learning. Furthermore, we show that these methods can be used to discover more human-understandable prompts that were previously unknown in both reasoning and image generation tasks, opening the door to a cornucopia of possibilities in prompt optimization.",
}
| Since the emergence of large language models, prompt learning has become a popular method for optimizing and customizing these models. Special prompts, such as Chain-of-Thought, have even revealed previously unknown reasoning capabilities within these models. However, the progress of discovering effective prompts has been slow, driving a desire for general prompt optimization methods. Unfortunately, few existing prompt learning methods satisfy the criteria of being truly {``}general{''}, i.e., automatic, discrete, black-box, gradient-free, and interpretable all at once. In this paper, we introduce metaheuristics, a branch of discrete non-convex optimization methods with over 100 options, as a promising approach to prompt learning. Within our paradigm, we test six typical methods: hill climbing, simulated annealing, genetic algorithms with/without crossover, tabu search, and harmony search, demonstrating their effectiveness in white-box and black-box prompt learning. Furthermore, we show that these methods can be used to discover more human-understandable prompts that were previously unknown in both reasoning and image generation tasks, opening the door to a cornucopia of possibilities in prompt optimization. | [
"Pan, Rui",
"Xing, Shuo",
"Diao, Shizhe",
"Sun, Wenhe",
"Liu, Xiang",
"Shum, KaShun",
"Zhang, Jipeng",
"Pi, Renjie",
"Zhang, Tong"
]
| Plum: Prompt Learning using Metaheuristics | findings-acl.129 | Poster | 2407.20777v1 |
https://aclanthology.org/2024.findings-acl.130.bib | @inproceedings{chen-etal-2024-hotvcom,
title = "{HOTVCOM}: Generating Buzzworthy Comments for Videos",
author = "Chen, Yuyan and
Yan, Songzhou and
Guo, Qingpei and
Jia, Jiyuan and
Li, Zhixu and
Xiao, Yanghua",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.130",
pages = "2198--2224",
abstract = "In the era of social media video platforms, popular {``}hot-comments{''} play a crucial role in attracting user impressions of short-form videos, making them vital for marketing and branding purpose. However, existing research predominantly focuses on generating descriptive comments or {``}danmaku{''} in English, offering immediate reactions to specific video moments. Addressing this gap, our study introduces HOTVCOM, the largest Chinese video hot-comment dataset, comprising 94k diverse videos and 137 million comments. We also present the ComHeat framework, which synergistically integrates visual, auditory, and textual data to generate influential hot-comments on the Chinese video dataset. Empirical evaluations highlight the effectiveness of our framework, demonstrating its excellence on both the newly constructed and existing datasets.",
}
| In the era of social media video platforms, popular {``}hot-comments{''} play a crucial role in attracting user impressions of short-form videos, making them vital for marketing and branding purpose. However, existing research predominantly focuses on generating descriptive comments or {``}danmaku{''} in English, offering immediate reactions to specific video moments. Addressing this gap, our study introduces HOTVCOM, the largest Chinese video hot-comment dataset, comprising 94k diverse videos and 137 million comments. We also present the ComHeat framework, which synergistically integrates visual, auditory, and textual data to generate influential hot-comments on the Chinese video dataset. Empirical evaluations highlight the effectiveness of our framework, demonstrating its excellence on both the newly constructed and existing datasets. | [
"Chen, Yuyan",
"Yan, Songzhou",
"Guo, Qingpei",
"Jia, Jiyuan",
"Li, Zhixu",
"Xiao, Yanghua"
]
| {HOTVCOM}: Generating Buzzworthy Comments for Videos | findings-acl.130 | Poster | 1808.04091v1 |
https://aclanthology.org/2024.findings-acl.131.bib | @inproceedings{chen-etal-2024-large,
title = "Do Large Language Models have Problem-Solving Capability under Incomplete Information Scenarios?",
author = "Chen, Yuyan and
Li, Yueze and
Yan, Songzhou and
Liu, Sijia and
Liang, Jiaqing and
Xiao, Yanghua",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.131",
pages = "2225--2238",
abstract = "The evaluation of the problem-solving capability under incomplete information scenarios of Large Language Models (LLMs) is increasingly important, encompassing capabilities such as questioning, knowledge search, error detection, and path planning. Current research mainly focus on LLMs{'} problem-solving capability such as {``}Twenty Questions{''}.However, these kinds of games do not require recognizing misleading cues which are necessary in the incomplete information scenario.Moreover, the existing game such as {``}Who is undercover{''} are highly subjective, making it challenging for evaluation.Therefore, in this paper, we introduce a novel game named BrainKing based on the {``}Who is undercover{''} and {``}Twenty Questions{''} for evaluating LLM capabilities under incomplete information scenarios. It requires LLMs to identify target entities with limited yes-or-no questions and potential misleading answers. By setting up easy, medium, and hard difficulty modes, we comprehensively assess the performance of LLMs across various aspects. Our results reveal the capabilities and limitations of LLMs in BrainKing, providing significant insights of LLM problem-solving levels.",
}
| The evaluation of the problem-solving capability under incomplete information scenarios of Large Language Models (LLMs) is increasingly important, encompassing capabilities such as questioning, knowledge search, error detection, and path planning. Current research mainly focus on LLMs{'} problem-solving capability such as {``}Twenty Questions{''}.However, these kinds of games do not require recognizing misleading cues which are necessary in the incomplete information scenario.Moreover, the existing game such as {``}Who is undercover{''} are highly subjective, making it challenging for evaluation.Therefore, in this paper, we introduce a novel game named BrainKing based on the {``}Who is undercover{''} and {``}Twenty Questions{''} for evaluating LLM capabilities under incomplete information scenarios. It requires LLMs to identify target entities with limited yes-or-no questions and potential misleading answers. By setting up easy, medium, and hard difficulty modes, we comprehensively assess the performance of LLMs across various aspects. Our results reveal the capabilities and limitations of LLMs in BrainKing, providing significant insights of LLM problem-solving levels. | [
"Chen, Yuyan",
"Li, Yueze",
"Yan, Songzhou",
"Liu, Sijia",
"Liang, Jiaqing",
"Xiao, Yanghua"
]
| Do Large Language Models have Problem-Solving Capability under Incomplete Information Scenarios? | findings-acl.131 | Poster | 1907.05562v2 |
https://aclanthology.org/2024.findings-acl.132.bib | @inproceedings{stacey-rei-2024-distilling,
title = "Distilling Robustness into Natural Language Inference Models with Domain-Targeted Augmentation",
author = "Stacey, Joe and
Rei, Marek",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.132",
pages = "2239--2258",
abstract = "Knowledge distillation optimises a smaller student model to behave similarly to a larger teacher model, retaining some of the performance benefits. While this method can improve results on in-distribution examples, it does not necessarily generalise to out-of-distribution (OOD) settings. We investigate two complementary methods for improving the robustness of the resulting student models on OOD domains. The first approach augments the distillation with generated unlabeled examples that match the target distribution. The second method upsamples data points among the training set that are similar to the target distribution. When applied on the task of natural language inference (NLI), our experiments on MNLI show that distillation with these modifications outperforms previous robustness solutions. We also find that these methods improve performance on OOD domains even beyond the target domain.",
}
| Knowledge distillation optimises a smaller student model to behave similarly to a larger teacher model, retaining some of the performance benefits. While this method can improve results on in-distribution examples, it does not necessarily generalise to out-of-distribution (OOD) settings. We investigate two complementary methods for improving the robustness of the resulting student models on OOD domains. The first approach augments the distillation with generated unlabeled examples that match the target distribution. The second method upsamples data points among the training set that are similar to the target distribution. When applied on the task of natural language inference (NLI), our experiments on MNLI show that distillation with these modifications outperforms previous robustness solutions. We also find that these methods improve performance on OOD domains even beyond the target domain. | [
"Stacey, Joe",
"Rei, Marek"
]
| Distilling Robustness into Natural Language Inference Models with Domain-Targeted Augmentation | findings-acl.132 | Poster | 2305.13067v3 |
https://aclanthology.org/2024.findings-acl.133.bib | @inproceedings{paz-argaman-etal-2024-unknown,
title = "Into the Unknown: Generating Geospatial Descriptions for New Environments",
author = "Paz-Argaman, Tzuf and
Palowitch, John and
Kulkarni, Sayali and
Tsarfaty, Reut and
Baldridge, Jason",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.133",
pages = "2259--2273",
abstract = "Similar to vision-and-language navigation (VLN) tasks that focus on bridging the gap between vision and language for embodied navigation, the new Rendezvous (RVS) task requires reasoning over allocentric spatial relationships using non-sequential navigation instructions and maps. However, performance substantially drops in new environments with no training data.Using opensource descriptions paired with coordinates (e.g., Wikipedia) provides training data but suffers from limited spatially-oriented text resulting in low geolocation resolution. We propose a large-scale augmentation method for generating high-quality synthetic data for new environments using readily available geospatial data. Our method constructs a grounded knowledge-graph, capturing entity relationships. Sampled entities and relations ({``}shop north of school{''}) generate navigation instructions via (i) generating numerous templates using context-free grammar (CFG) to embed specific entities and relations; (ii) feeding the entities and relation into a large language model (LLM) for instruction generation. A comprehensive evaluation on RVS, showed that our approach improves the 100-meter accuracy by 45.83{\%} on unseen environments. Furthermore, we demonstrate that models trained with CFG-based augmentation achieve superior performance compared with those trained with LLM-based augmentation, both in unseen and seen environments. These findings suggest that the potential advantages of explicitly structuring spatial information for text-based geospatial reasoning in previously unknown, can unlock data-scarce scenarios.",
}
| Similar to vision-and-language navigation (VLN) tasks that focus on bridging the gap between vision and language for embodied navigation, the new Rendezvous (RVS) task requires reasoning over allocentric spatial relationships using non-sequential navigation instructions and maps. However, performance substantially drops in new environments with no training data.Using opensource descriptions paired with coordinates (e.g., Wikipedia) provides training data but suffers from limited spatially-oriented text resulting in low geolocation resolution. We propose a large-scale augmentation method for generating high-quality synthetic data for new environments using readily available geospatial data. Our method constructs a grounded knowledge-graph, capturing entity relationships. Sampled entities and relations ({``}shop north of school{''}) generate navigation instructions via (i) generating numerous templates using context-free grammar (CFG) to embed specific entities and relations; (ii) feeding the entities and relation into a large language model (LLM) for instruction generation. A comprehensive evaluation on RVS, showed that our approach improves the 100-meter accuracy by 45.83{\%} on unseen environments. Furthermore, we demonstrate that models trained with CFG-based augmentation achieve superior performance compared with those trained with LLM-based augmentation, both in unseen and seen environments. These findings suggest that the potential advantages of explicitly structuring spatial information for text-based geospatial reasoning in previously unknown, can unlock data-scarce scenarios. | [
"Paz-Argaman, Tzuf",
"Palowitch, John",
"Kulkarni, Sayali",
"Tsarfaty, Reut",
"Baldridge, Jason"
]
| Into the Unknown: Generating Geospatial Descriptions for New Environments | findings-acl.133 | Poster | 2405.07664v1 |
https://aclanthology.org/2024.findings-acl.134.bib | @inproceedings{goldman-etal-2024-unpacking,
title = "Unpacking Tokenization: Evaluating Text Compression and its Correlation with Model Performance",
author = "Goldman, Omer and
Caciularu, Avi and
Eyal, Matan and
Cao, Kris and
Szpektor, Idan and
Tsarfaty, Reut",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.134",
pages = "2274--2286",
abstract = "Despite it being the cornerstone of BPE, the most common tokenization algorithm, the importance of compression in the tokenization process is still unclear. In this paper, we argue for the theoretical importance of compression, that can be viewed as 0-gram language modeling where equal probability is assigned to all tokens. We also demonstrate the empirical importance of compression for downstream success of pre-trained language models. We control the compression ability of several BPE tokenizers by varying the amount of documents available during their training: from 1 million documents to a character-based tokenizer equivalent to no training data at all. We then pre-train English language models based on those tokenizers and fine-tune them over several tasks. We show that there is a correlation between tokenizers{'} compression and models{'} downstream performance, suggesting that compression is a reliable intrinsic indicator of tokenization quality. These correlations are more pronounced for generation tasks (over classification) or for smaller models (over large ones). We replicated a representative part of our experiments on Turkish and found similar results, confirming that our results hold for languages with typological characteristics dissimilar to English. We conclude that building better compressing tokenizers is a fruitful avenue for further research and for improving overall model performance.",
}
| Despite it being the cornerstone of BPE, the most common tokenization algorithm, the importance of compression in the tokenization process is still unclear. In this paper, we argue for the theoretical importance of compression, that can be viewed as 0-gram language modeling where equal probability is assigned to all tokens. We also demonstrate the empirical importance of compression for downstream success of pre-trained language models. We control the compression ability of several BPE tokenizers by varying the amount of documents available during their training: from 1 million documents to a character-based tokenizer equivalent to no training data at all. We then pre-train English language models based on those tokenizers and fine-tune them over several tasks. We show that there is a correlation between tokenizers{'} compression and models{'} downstream performance, suggesting that compression is a reliable intrinsic indicator of tokenization quality. These correlations are more pronounced for generation tasks (over classification) or for smaller models (over large ones). We replicated a representative part of our experiments on Turkish and found similar results, confirming that our results hold for languages with typological characteristics dissimilar to English. We conclude that building better compressing tokenizers is a fruitful avenue for further research and for improving overall model performance. | [
"Goldman, Omer",
"Caciularu, Avi",
"Eyal, Matan",
"Cao, Kris",
"Szpektor, Idan",
"Tsarfaty, Reut"
]
| Unpacking Tokenization: Evaluating Text Compression and its Correlation with Model Performance | findings-acl.134 | Poster | 2403.06265v2 |
https://aclanthology.org/2024.findings-acl.135.bib | @inproceedings{lee-etal-2024-length,
title = "Length-aware Byte Pair Encoding for Mitigating Over-segmentation in {K}orean Machine Translation",
author = "Lee, Jungseob and
Moon, Hyeonseok and
Lee, Seungjun and
Park, Chanjun and
Eo, Sugyeong and
Ko, Hyunwoong and
Seo, Jaehyung and
Lee, Seungyoon and
Lim, Heuiseok",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.135",
pages = "2287--2303",
abstract = "Byte Pair Encoding is an effective approach in machine translation across several languages. However, our analysis indicates that BPE is prone to over-segmentation in the morphologically rich language, Korean, which can erode word semantics and lead to semantic confusion during training. This semantic confusion, stemming from over-segmentation, ultimately contributes to a degradation of overall translation quality. To address this issue, we introduce Length-aware Subword Vocabulary Construction (LeVoC), a novel approach strategically incorporating longer words into the vocabulary. By utilizing an external monolingual Korean corpus, LeVoC extracts and integrates long words, effectively preserving morphological information and reducing semantic confusion. Our experiments demonstrate that LeVoC not only significantly outperforms BPE, but also can be applied to and surpass current state-of-the-art morpheme-aware subword tokenization methods. We provide evidence that the difficulty in translating sentences with long words in Korean is associated with morphological compositionality, and LeVoC{'}s ability to reduce semantic confusion during training leads to improved translation quality.",
}
| Byte Pair Encoding is an effective approach in machine translation across several languages. However, our analysis indicates that BPE is prone to over-segmentation in the morphologically rich language, Korean, which can erode word semantics and lead to semantic confusion during training. This semantic confusion, stemming from over-segmentation, ultimately contributes to a degradation of overall translation quality. To address this issue, we introduce Length-aware Subword Vocabulary Construction (LeVoC), a novel approach strategically incorporating longer words into the vocabulary. By utilizing an external monolingual Korean corpus, LeVoC extracts and integrates long words, effectively preserving morphological information and reducing semantic confusion. Our experiments demonstrate that LeVoC not only significantly outperforms BPE, but also can be applied to and surpass current state-of-the-art morpheme-aware subword tokenization methods. We provide evidence that the difficulty in translating sentences with long words in Korean is associated with morphological compositionality, and LeVoC{'}s ability to reduce semantic confusion during training leads to improved translation quality. | [
"Lee, Jungseob",
"Moon, Hyeonseok",
"Lee, Seungjun",
"Park, Chanjun",
"Eo, Sugyeong",
"Ko, Hyunwoong",
"Seo, Jaehyung",
"Lee, Seungyoon",
"Lim, Heuiseok"
]
| Length-aware Byte Pair Encoding for Mitigating Over-segmentation in {K}orean Machine Translation | findings-acl.135 | Poster | 2010.02534v1 |
https://aclanthology.org/2024.findings-acl.136.bib | @inproceedings{shaham-etal-2024-multilingual,
title = "Multilingual Instruction Tuning With Just a Pinch of Multilinguality",
author = "Shaham, Uri and
Herzig, Jonathan and
Aharoni, Roee and
Szpektor, Idan and
Tsarfaty, Reut and
Eyal, Matan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.136",
pages = "2304--2317",
abstract = "As instruction-tuned large language models (LLMs) gain global adoption, their ability to follow instructions in multiple languages becomes increasingly crucial. In this work, we investigate how multilinguality during instruction tuning of a multilingual LLM affects instruction-following across languages from the pre-training corpus. We first show that many languages transfer some instruction-following capabilities to other languages from even monolingual tuning. Furthermore, we find that only 40 multilingual examples integrated in an English tuning set substantially improve multilingual instruction-following, both in seen and unseen languages during tuning. In general, we observe that models tuned on multilingual mixtures exhibit comparable or superior performance in multiple languages compared to monolingually tuned models, despite training on 10x fewer examples in those languages. Finally, we find that diversifying the instruction tuning set with even just 2-4 languages significantly improves cross-lingual generalization. Our results suggest that building massively multilingual instruction-tuned models can be done with only a very small set of multilingual instruction-responses.",
}
| As instruction-tuned large language models (LLMs) gain global adoption, their ability to follow instructions in multiple languages becomes increasingly crucial. In this work, we investigate how multilinguality during instruction tuning of a multilingual LLM affects instruction-following across languages from the pre-training corpus. We first show that many languages transfer some instruction-following capabilities to other languages from even monolingual tuning. Furthermore, we find that only 40 multilingual examples integrated in an English tuning set substantially improve multilingual instruction-following, both in seen and unseen languages during tuning. In general, we observe that models tuned on multilingual mixtures exhibit comparable or superior performance in multiple languages compared to monolingually tuned models, despite training on 10x fewer examples in those languages. Finally, we find that diversifying the instruction tuning set with even just 2-4 languages significantly improves cross-lingual generalization. Our results suggest that building massively multilingual instruction-tuned models can be done with only a very small set of multilingual instruction-responses. | [
"Shaham, Uri",
"Herzig, Jonathan",
"Aharoni, Roee",
"Szpektor, Idan",
"Tsarfaty, Reut",
"Eyal, Matan"
]
| Multilingual Instruction Tuning With Just a Pinch of Multilinguality | findings-acl.136 | Poster | 2401.01854v4 |
https://aclanthology.org/2024.findings-acl.137.bib | @inproceedings{chen-etal-2024-m3,
title = "{M}3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation",
author = "Chen, Jianlyu and
Xiao, Shitao and
Zhang, Peitian and
Luo, Kun and
Lian, Defu and
Liu, Zheng",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.137",
pages = "2318--2335",
abstract = "In this paper, we introduce a new embedding model called M3-Embedding, which is distinguished for its versatility in Multi-Linguality, Multi-Functionality, and Multi-Granularity. It provides a uniform support for the semantic retrieval of more than 100 working languages. It can simultaneously accomplish the three common retrieval functionalities: dense retrieval, multi-vector retrieval, and sparse retrieval. Besides, it is also capable of processing inputs of different granularities, spanning from short sentences to long documents of up to 8,192 tokens. The effective training of M3-Embedding presents a series of technical contributions. Notably, we propose a novel self-knowledge distillation approach, where the relevance scores from different retrieval functionalities can be integrated as the teacher signal to enhance the training quality. We also optimize the batching strategy, which enables a large batch size and high training throughput to improve the discriminativeness of embeddings. M3-Embedding exhibits a superior performance in our experiment, leading to new state-of-the-art results on multilingual, cross-lingual, and long-document retrieval benchmarks.",
}
| In this paper, we introduce a new embedding model called M3-Embedding, which is distinguished for its versatility in Multi-Linguality, Multi-Functionality, and Multi-Granularity. It provides a uniform support for the semantic retrieval of more than 100 working languages. It can simultaneously accomplish the three common retrieval functionalities: dense retrieval, multi-vector retrieval, and sparse retrieval. Besides, it is also capable of processing inputs of different granularities, spanning from short sentences to long documents of up to 8,192 tokens. The effective training of M3-Embedding presents a series of technical contributions. Notably, we propose a novel self-knowledge distillation approach, where the relevance scores from different retrieval functionalities can be integrated as the teacher signal to enhance the training quality. We also optimize the batching strategy, which enables a large batch size and high training throughput to improve the discriminativeness of embeddings. M3-Embedding exhibits a superior performance in our experiment, leading to new state-of-the-art results on multilingual, cross-lingual, and long-document retrieval benchmarks. | [
"Chen, Jianlyu",
"Xiao, Shitao",
"Zhang, Peitian",
"Luo, Kun",
"Lian, Defu",
"Liu, Zheng"
]
| {M}3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation | findings-acl.137 | Poster | 2312.12659v1 |
https://aclanthology.org/2024.findings-acl.138.bib | @inproceedings{bi-etal-2024-iterative,
title = "Iterative Refinement of Project-Level Code Context for Precise Code Generation with Compiler Feedback",
author = "Bi, Zhangqian and
Wan, Yao and
Wang, Zheng and
Zhang, Hongyu and
Guan, Batu and
Lu, Fangxin and
Zhang, Zili and
Sui, Yulei and
Jin, Hai and
Shi, Xuanhua",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.138",
pages = "2336--2353",
abstract = "Large Language Models (LLMs) have shown remarkable progress in automated code generation. Yet, LLM-generated code may contain errors in API usage, class, data structure, or missing project-specific information. As much of this project-specific context cannot fit into the prompts of LLMs, we must find ways to allow the model to explore the project-level code context. We present CoCoGen, a new code generation approach that uses compiler feedback to improve the LLM-generated code. CoCoGen first leverages static analysis to identify mismatches between the generated code and the project{'}s context. It then iteratively aligns and fixes the identified errors using information extracted from the code repository. We integrate CoCoGen with two representative LLMs, i.e., GPT-3.5-Turbo and Code Llama (13B), and apply it to Python code generation. Experimental results show that CoCoGen significantly improves the vanilla LLMs by over 80{\%} in generating code dependent on the project context and consistently outperforms the existing retrieval-based code generation baselines.",
}
| Large Language Models (LLMs) have shown remarkable progress in automated code generation. Yet, LLM-generated code may contain errors in API usage, class, data structure, or missing project-specific information. As much of this project-specific context cannot fit into the prompts of LLMs, we must find ways to allow the model to explore the project-level code context. We present CoCoGen, a new code generation approach that uses compiler feedback to improve the LLM-generated code. CoCoGen first leverages static analysis to identify mismatches between the generated code and the project{'}s context. It then iteratively aligns and fixes the identified errors using information extracted from the code repository. We integrate CoCoGen with two representative LLMs, i.e., GPT-3.5-Turbo and Code Llama (13B), and apply it to Python code generation. Experimental results show that CoCoGen significantly improves the vanilla LLMs by over 80{\%} in generating code dependent on the project context and consistently outperforms the existing retrieval-based code generation baselines. | [
"Bi, Zhangqian",
"Wan, Yao",
"Wang, Zheng",
"Zhang, Hongyu",
"Guan, Batu",
"Lu, Fangxin",
"Zhang, Zili",
"Sui, Yulei",
"Jin, Hai",
"Shi, Xuanhua"
]
| Iterative Refinement of Project-Level Code Context for Precise Code Generation with Compiler Feedback | findings-acl.138 | Poster | 2403.16792v3 |
https://aclanthology.org/2024.findings-acl.139.bib | @inproceedings{deng-etal-2024-element,
title = "An Element is Worth a Thousand Words: Enhancing Legal Case Retrieval by Incorporating Legal Elements",
author = "Deng, Chenlong and
Dou, Zhicheng and
Zhou, Yujia and
Zhang, Peitian and
Mao, Kelong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.139",
pages = "2354--2365",
abstract = "Legal case retrieval plays an important role in promoting judicial justice and fairness. One of its greatest challenges is that the definition of relevance goes far beyond the common semantic relevance as in ad-hoc retrieval. In this paper, we reveal that the legal elements, which typically comprise key facts in a specialized legal context, can largely improve the relevance matching of legal case retrieval. To facilitate the use of legal elements, we construct a Chinese legal element dataset called LeCaRD-Elem based on the widely-used LeCaRD dataset, through a two-stage semi-automatic method with a minimized reliance on human labor. Meanwhile, we introduce two new models to enhance legal search using legal elements. The first, Elem4LCR-E, is a two-stage model that explicitly predicts legal elements from texts and then leverages them for improved ranking. Recognizing the potential benefits of more seamless integration, we further propose an end-to-end model called Elem4LCR-I, which internalizes the legal element knowledge into its model parameters using a tailored teacher-student training framework. Extensive experiments underscore the significant value of legal elements and demonstrate the superiority of our two proposed models in enhancing legal search over existing methods.",
}
| Legal case retrieval plays an important role in promoting judicial justice and fairness. One of its greatest challenges is that the definition of relevance goes far beyond the common semantic relevance as in ad-hoc retrieval. In this paper, we reveal that the legal elements, which typically comprise key facts in a specialized legal context, can largely improve the relevance matching of legal case retrieval. To facilitate the use of legal elements, we construct a Chinese legal element dataset called LeCaRD-Elem based on the widely-used LeCaRD dataset, through a two-stage semi-automatic method with a minimized reliance on human labor. Meanwhile, we introduce two new models to enhance legal search using legal elements. The first, Elem4LCR-E, is a two-stage model that explicitly predicts legal elements from texts and then leverages them for improved ranking. Recognizing the potential benefits of more seamless integration, we further propose an end-to-end model called Elem4LCR-I, which internalizes the legal element knowledge into its model parameters using a tailored teacher-student training framework. Extensive experiments underscore the significant value of legal elements and demonstrate the superiority of our two proposed models in enhancing legal search over existing methods. | [
"Deng, Chenlong",
"Dou, Zhicheng",
"Zhou, Yujia",
"Zhang, Peitian",
"Mao, Kelong"
]
| An Element is Worth a Thousand Words: Enhancing Legal Case Retrieval by Incorporating Legal Elements | findings-acl.139 | Poster | 2304.11370v1 |
https://aclanthology.org/2024.findings-acl.140.bib | @inproceedings{zhang-etal-2024-somelvlm,
title = "{S}o{M}e{LVLM}: A Large Vision Language Model for Social Media Processing",
author = "Zhang, Xinnong and
Kuang, Haoyu and
Mou, Xinyi and
Lyu, Hanjia and
Wu, Kun and
Chen, Siming and
Luo, Jiebo and
Huang, Xuanjing and
Wei, Zhongyu",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.140",
pages = "2366--2389",
abstract = "The growth of social media, characterized by its multimodal nature, has led to the emergence of diverse phenomena and challenges, which calls for an effective approach to uniformly solve automated tasks. The powerful Large Vision Language Models make it possible to handle a variety of tasks simultaneously, but even with carefully designed prompting methods, the general domain models often fall short in aligning with the unique speaking style and context of social media tasks. In this paper, we introduce a Large Vision Language Model for Social Media Processing (SoMeLVLM), which is a cognitive framework equipped with five key capabilities including knowledge {\&} comprehension, application, analysis, evaluation, and creation. SoMeLVLM is designed to understand and generate realistic social media behavior. We have developed a 654k multimodal social media instruction-tuning dataset to support our cognitive framework and fine-tune our model. Our experiments demonstrate that SoMeLVLM achieves state-of-the-art performance in multiple social media tasks. Further analysis shows its significant advantages over baselines in terms of cognitive abilities.",
}
| The growth of social media, characterized by its multimodal nature, has led to the emergence of diverse phenomena and challenges, which calls for an effective approach to uniformly solve automated tasks. The powerful Large Vision Language Models make it possible to handle a variety of tasks simultaneously, but even with carefully designed prompting methods, the general domain models often fall short in aligning with the unique speaking style and context of social media tasks. In this paper, we introduce a Large Vision Language Model for Social Media Processing (SoMeLVLM), which is a cognitive framework equipped with five key capabilities including knowledge {\&} comprehension, application, analysis, evaluation, and creation. SoMeLVLM is designed to understand and generate realistic social media behavior. We have developed a 654k multimodal social media instruction-tuning dataset to support our cognitive framework and fine-tune our model. Our experiments demonstrate that SoMeLVLM achieves state-of-the-art performance in multiple social media tasks. Further analysis shows its significant advantages over baselines in terms of cognitive abilities. | [
"Zhang, Xinnong",
"Kuang, Haoyu",
"Mou, Xinyi",
"Lyu, Hanjia",
"Wu, Kun",
"Chen, Siming",
"Luo, Jiebo",
"Huang, Xuanjing",
"Wei, Zhongyu"
]
| {S}o{M}e{LVLM}: A Large Vision Language Model for Social Media Processing | findings-acl.140 | Poster | 2402.13022v1 |
https://aclanthology.org/2024.findings-acl.141.bib | @inproceedings{seo-etal-2024-kocommongen,
title = "{K}o{C}ommon{GEN} v2: A Benchmark for Navigating {K}orean Commonsense Reasoning Challenges in Large Language Models",
author = "Seo, Jaehyung and
Lee, Jaewook and
Park, Chanjun and
Hong, SeongTae and
Lee, Seungjun and
Lim, Heuiseok",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.141",
pages = "2390--2415",
abstract = "The evolution of large language models (LLMs) has culminated in a multitask model paradigm where prompts drive the generation of user-specific outputs. However, this advancement has revealed a critical challenge: LLMs frequently produce outputs against socially acceptable commonsense standards in various scenarios. To address this gap in commonsense reasoning, we present KoCommonGEN v2, a fine-grained benchmark dataset focused on Korean commonsense reasoning. This dataset, enriched with human annotations, comprises multiple-choice questions across seven error categories. These categories include commonsense memorization, numerical commonsense, toxic speech, and more, which are vulnerable to undermining the reliability of LLMs{'} commonsense reasoning capabilities. The empirical results present that LLMs struggle with Korean commonsense reasoning. With human accuracy benchmarked at approximately 85{\%}, GPT-4{'}s performance lags at about 74{\%}, and other LLMs demonstrate an average accuracy of around 42{\%}. Our findings emphasize the need for targeted improvements in Korean commonsense reasoning within LLMs, paving the way for more socially and contextually sensitive AI models.",
}
| The evolution of large language models (LLMs) has culminated in a multitask model paradigm where prompts drive the generation of user-specific outputs. However, this advancement has revealed a critical challenge: LLMs frequently produce outputs against socially acceptable commonsense standards in various scenarios. To address this gap in commonsense reasoning, we present KoCommonGEN v2, a fine-grained benchmark dataset focused on Korean commonsense reasoning. This dataset, enriched with human annotations, comprises multiple-choice questions across seven error categories. These categories include commonsense memorization, numerical commonsense, toxic speech, and more, which are vulnerable to undermining the reliability of LLMs{'} commonsense reasoning capabilities. The empirical results present that LLMs struggle with Korean commonsense reasoning. With human accuracy benchmarked at approximately 85{\%}, GPT-4{'}s performance lags at about 74{\%}, and other LLMs demonstrate an average accuracy of around 42{\%}. Our findings emphasize the need for targeted improvements in Korean commonsense reasoning within LLMs, paving the way for more socially and contextually sensitive AI models. | [
"Seo, Jaehyung",
"Lee, Jaewook",
"Park, Chanjun",
"Hong, SeongTae",
"Lee, Seungjun",
"Lim, Heuiseok"
]
| {K}o{C}ommon{GEN} v2: A Benchmark for Navigating {K}orean Commonsense Reasoning Challenges in Large Language Models | findings-acl.141 | Poster | 2312.17661v1 |
https://aclanthology.org/2024.findings-acl.142.bib | @inproceedings{dhurandhar-etal-2024-neuroprune,
title = "{N}euro{P}rune: A Neuro-inspired Topological Sparse Training Algorithm for Large Language Models",
author = "Dhurandhar, Amit and
Pedapati, Tejaswini and
Luss, Ronny and
Dan, Soham and
Lozano, Aurelie and
Das, Payel and
Kollias, Georgios",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.142",
pages = "2416--2430",
abstract = "Transformer-based Language Models have become ubiquitous in Natural Language Processing (NLP) due to their impressive performance on various tasks. However, expensive training as well as inference remains a significant impediment to their widespread applicability. While enforcing sparsity at various levels of the model architecture has found promise in addressing scaling and efficiency issues, there remains a disconnect between how sparsity affects network topology. Inspired by brain neuronal networks, we explore sparsity approaches through the lens of network topology. Specifically, we exploit mechanisms seen in biological networks, such as preferential attachment and redundant synapse pruning, and show that principled, model-agnostic sparsity approaches are performant and efficient across diverse NLP tasks, spanning both classification (such as natural language inference) and generation (summarization, machine translation), despite our sole objective not being optimizing performance. NeuroPrune is competitive with (or sometimes superior to) baselines on performance and can be up to 10x faster in terms of training time for a given level of sparsity, simultaneously exhibiting measurable improvements in inference time in many cases.",
}
| Transformer-based Language Models have become ubiquitous in Natural Language Processing (NLP) due to their impressive performance on various tasks. However, expensive training as well as inference remains a significant impediment to their widespread applicability. While enforcing sparsity at various levels of the model architecture has found promise in addressing scaling and efficiency issues, there remains a disconnect between how sparsity affects network topology. Inspired by brain neuronal networks, we explore sparsity approaches through the lens of network topology. Specifically, we exploit mechanisms seen in biological networks, such as preferential attachment and redundant synapse pruning, and show that principled, model-agnostic sparsity approaches are performant and efficient across diverse NLP tasks, spanning both classification (such as natural language inference) and generation (summarization, machine translation), despite our sole objective not being optimizing performance. NeuroPrune is competitive with (or sometimes superior to) baselines on performance and can be up to 10x faster in terms of training time for a given level of sparsity, simultaneously exhibiting measurable improvements in inference time in many cases. | [
"Dhur",
"har, Amit",
"Pedapati, Tejaswini",
"Luss, Ronny",
"Dan, Soham",
"Lozano, Aurelie",
"Das, Payel",
"Kollias, Georgios"
]
| {N}euro{P}rune: A Neuro-inspired Topological Sparse Training Algorithm for Large Language Models | findings-acl.142 | Poster | 2111.05754v1 |
https://aclanthology.org/2024.findings-acl.143.bib | @inproceedings{dhurandhar-etal-2024-ranking,
title = "Ranking Large Language Models without Ground Truth",
author = "Dhurandhar, Amit and
Nair, Rahul and
Singh, Moninder and
Daly, Elizabeth and
Natesan Ramamurthy, Karthikeyan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.143",
pages = "2431--2452",
abstract = "Evaluation and ranking of large language models (LLMs) has become an important problem with the proliferation of these models and their impact. Evaluation methods either require human responses which are expensive to acquire or use pairs of LLMs to evaluate each other which can be unreliable. In this paper, we provide a novel perspective where, given a dataset of prompts (viz. questions, instructions, etc.) and a set of LLMs, we rank them without access to any ground truth or reference responses. Inspired by real life where both an expert and a knowledgeable person can identify a novice our main idea is to consider triplets of models, where each one of them evaluates the other two, correctly identifying the worst model in the triplet with high probability. We also analyze our idea and provide sufficient conditions for it to succeed. Applying this idea repeatedly we propose two methods to rank LLMs. In experiments on different generative tasks (summarization, multiple-choice, and dialog), our methods reliably recover true rankings without reference data. This points to a viable low-resource mechanism for practical use.",
}
| Evaluation and ranking of large language models (LLMs) has become an important problem with the proliferation of these models and their impact. Evaluation methods either require human responses which are expensive to acquire or use pairs of LLMs to evaluate each other which can be unreliable. In this paper, we provide a novel perspective where, given a dataset of prompts (viz. questions, instructions, etc.) and a set of LLMs, we rank them without access to any ground truth or reference responses. Inspired by real life where both an expert and a knowledgeable person can identify a novice our main idea is to consider triplets of models, where each one of them evaluates the other two, correctly identifying the worst model in the triplet with high probability. We also analyze our idea and provide sufficient conditions for it to succeed. Applying this idea repeatedly we propose two methods to rank LLMs. In experiments on different generative tasks (summarization, multiple-choice, and dialog), our methods reliably recover true rankings without reference data. This points to a viable low-resource mechanism for practical use. | [
"Dhur",
"har, Amit",
"Nair, Rahul",
"Singh, Moninder",
"Daly, Elizabeth",
"Natesan Ramamurthy, Karthikeyan"
]
| Ranking Large Language Models without Ground Truth | findings-acl.143 | Poster | 2402.14860v4 |
https://aclanthology.org/2024.findings-acl.144.bib | @inproceedings{dou-etal-2024-integrating,
title = "Integrating Physician Diagnostic Logic into Large Language Models: Preference Learning from Process Feedback",
author = "Dou, Chengfeng and
Zhang, Ying and
Jin, Zhi and
Jiao, Wenpin and
Zhao, Haiyan and
Zhao, Yongqiang and
Tao, Zhengwei",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.144",
pages = "2453--2473",
abstract = "The utilization of large language models for medical dialogue generation has attracted considerable attention due to its potential to enhance response richness and coherence. While previous studies have made strides in optimizing model performance, there is a pressing need to bolster the model{'}s capacity for diagnostic logic to ensure patient safety. In response to this need, we propose an approach termed preference learning from process feedback (PLPF), which involves integrating the doctor{'}s diagnostic logic into LLMs. PLPF encompasses three key components: rule modeling, preference data generation, and preference alignment. These components collectively serve to train the model to adhere to the diagnostic process. Our experimental results, utilizing Standardized Patient Testing, demonstrate that PLPF enhances the diagnostic accuracy of the baseline model in medical conversations by 17.6{\%}, surpassing the performance of traditional approaches. Moreover, PLPF exhibits effectiveness in both multi-round and single-round dialogue tasks, thereby highlighting its potential in improving medical dialogue generation. Our dataset is available at https://github.com/Chengfeng-Dou/SpTesting.",
}
| The utilization of large language models for medical dialogue generation has attracted considerable attention due to its potential to enhance response richness and coherence. While previous studies have made strides in optimizing model performance, there is a pressing need to bolster the model{'}s capacity for diagnostic logic to ensure patient safety. In response to this need, we propose an approach termed preference learning from process feedback (PLPF), which involves integrating the doctor{'}s diagnostic logic into LLMs. PLPF encompasses three key components: rule modeling, preference data generation, and preference alignment. These components collectively serve to train the model to adhere to the diagnostic process. Our experimental results, utilizing Standardized Patient Testing, demonstrate that PLPF enhances the diagnostic accuracy of the baseline model in medical conversations by 17.6{\%}, surpassing the performance of traditional approaches. Moreover, PLPF exhibits effectiveness in both multi-round and single-round dialogue tasks, thereby highlighting its potential in improving medical dialogue generation. Our dataset is available at https://github.com/Chengfeng-Dou/SpTesting. | [
"Dou, Chengfeng",
"Zhang, Ying",
"Jin, Zhi",
"Jiao, Wenpin",
"Zhao, Haiyan",
"Zhao, Yongqiang",
"Tao, Zhengwei"
]
| Integrating Physician Diagnostic Logic into Large Language Models: Preference Learning from Process Feedback | findings-acl.144 | Poster | 2401.05695v2 |
https://aclanthology.org/2024.findings-acl.145.bib | @inproceedings{xiao-etal-2024-lm,
title = "{LM}-Cocktail: Resilient Tuning of Language Models via Model Merging",
author = "Xiao, Shitao and
Liu, Zheng and
Zhang, Peitian and
Xing, Xingrun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.145",
pages = "2474--2488",
abstract = "The pre-trained language models are continually fine-tuned to better support downstream applications. However, this operation may result in significant performance degeneration on general tasks beyond the targeted domain. To overcome this problem, we propose LM-Cocktail which enables the fine-tuned model to stay resilient in general perspectives. Our method is conducted in the form of model merging, where the fine-tuned language model is merged with the pre-trained base model or the peer models from other domains through weighted average. Despite simplicity, LM-Cocktail is surprisingly effective: the resulted model is able to achieve a strong empirical performance in the whole scope of general tasks while preserving a superior capacity in its targeted domain.",
}
| The pre-trained language models are continually fine-tuned to better support downstream applications. However, this operation may result in significant performance degeneration on general tasks beyond the targeted domain. To overcome this problem, we propose LM-Cocktail which enables the fine-tuned model to stay resilient in general perspectives. Our method is conducted in the form of model merging, where the fine-tuned language model is merged with the pre-trained base model or the peer models from other domains through weighted average. Despite simplicity, LM-Cocktail is surprisingly effective: the resulted model is able to achieve a strong empirical performance in the whole scope of general tasks while preserving a superior capacity in its targeted domain. | [
"Xiao, Shitao",
"Liu, Zheng",
"Zhang, Peitian",
"Xing, Xingrun"
]
| {LM}-Cocktail: Resilient Tuning of Language Models via Model Merging | findings-acl.145 | Poster | 2311.13534v4 |
https://aclanthology.org/2024.findings-acl.146.bib | @inproceedings{miao-etal-2024-episodic,
title = "Episodic Memory Retrieval from {LLM}s: A Neuromorphic Mechanism to Generate Commonsense Counterfactuals for Relation Extraction",
author = "Miao, Xin and
Li, Yongqi and
Zhou, Shen and
Qian, Tieyun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.146",
pages = "2489--2511",
abstract = "Large language models (LLMs) have achieved satisfactory performance in counterfactual generation. However, confined by the stochastic generation process of LLMs, there often are misalignments between LLMs and humans which hinder LLMs from handling complex tasks like relation extraction. As a result, LLMs may generate commonsense-violated counterfactuals like {`}eggs were produced by a box{'}. To bridge this gap, we propose to mimick the episodic memory retrieval, the working mechanism of human hippocampus, to align LLMs{'} generation process with that of humans. In this way, LLMs can derive experience from their extensive memory, which keeps in line with the way humans gain commonsense. We then implement two central functions in the hippocampus, i.e., pattern separation and pattern completion, to retrieve the episodic memory from LLMs and generate commonsense counterfactuals for relation extraction. Experimental results demonstrate the improvements of our framework over existing methods in terms of the quality of counterfactuals.",
}
| Large language models (LLMs) have achieved satisfactory performance in counterfactual generation. However, confined by the stochastic generation process of LLMs, there often are misalignments between LLMs and humans which hinder LLMs from handling complex tasks like relation extraction. As a result, LLMs may generate commonsense-violated counterfactuals like {`}eggs were produced by a box{'}. To bridge this gap, we propose to mimick the episodic memory retrieval, the working mechanism of human hippocampus, to align LLMs{'} generation process with that of humans. In this way, LLMs can derive experience from their extensive memory, which keeps in line with the way humans gain commonsense. We then implement two central functions in the hippocampus, i.e., pattern separation and pattern completion, to retrieve the episodic memory from LLMs and generate commonsense counterfactuals for relation extraction. Experimental results demonstrate the improvements of our framework over existing methods in terms of the quality of counterfactuals. | [
"Miao, Xin",
"Li, Yongqi",
"Zhou, Shen",
"Qian, Tieyun"
]
| Episodic Memory Retrieval from {LLM}s: A Neuromorphic Mechanism to Generate Commonsense Counterfactuals for Relation Extraction | findings-acl.146 | Poster | 2306.02388v1 |
https://aclanthology.org/2024.findings-acl.147.bib | @inproceedings{ousidhoum-etal-2024-semrel2024,
title = "{S}em{R}el2024: A Collection of Semantic Textual Relatedness Datasets for 13 Languages",
author = "Ousidhoum, Nedjma and
Muhammad, Shamsuddeen and
Abdalla, Mohamed and
Abdulmumin, Idris and
Ahmad, Ibrahim and
Ahuja, Sanchit and
Aji, Alham and
Araujo, Vladimir and
Ayele, Abinew and
Baswani, Pavan and
Beloucif, Meriem and
Biemann, Chris and
Bourhim, Sofia and
Kock, Christine and
Dekebo, Genet and
Hourrane, Oumaima and
Kanumolu, Gopichand and
Madasu, Lokesh and
Rutunda, Samuel and
Shrivastava, Manish and
Solorio, Thamar and
Surange, Nirmal and
Tilaye, Hailegnaw and
Vishnubhotla, Krishnapriya and
Winata, Genta and
Yimam, Seid and
Mohammad, Saif",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.147",
pages = "2512--2530",
abstract = "Exploring and quantifying semantic relatedness is central to representing language and holds significant implications across various NLP tasks. While earlier NLP research primarily focused on semantic similarity, often within the English language context, we instead investigate the broader phenomenon of semantic relatedness. In this paper, we present \textit{SemRel}, a new semantic relatedness dataset collection annotated by native speakers across 13 languages: \textit{Afrikaans, Algerian Arabic, Amharic, English, Hausa, Hindi, Indonesian, Kinyarwanda, Marathi, Moroccan Arabic, Modern Standard Arabic, Spanish,} and \textit{Telugu}. These languages originate from five distinct language families and are predominantly spoken in Africa and Asia {--} regions characterised by a relatively limited availability of NLP resources. Each instance in the SemRel datasets is a sentence pair associated with a score that represents the degree of semantic textual relatedness between the two sentences. The scores are obtained using a comparative annotation framework. We describe the data collection and annotation processes, challenges when building the datasets, baseline experiments, and their impact and utility in NLP.",
}
| Exploring and quantifying semantic relatedness is central to representing language and holds significant implications across various NLP tasks. While earlier NLP research primarily focused on semantic similarity, often within the English language context, we instead investigate the broader phenomenon of semantic relatedness. In this paper, we present \textit{SemRel}, a new semantic relatedness dataset collection annotated by native speakers across 13 languages: \textit{Afrikaans, Algerian Arabic, Amharic, English, Hausa, Hindi, Indonesian, Kinyarwanda, Marathi, Moroccan Arabic, Modern Standard Arabic, Spanish,} and \textit{Telugu}. These languages originate from five distinct language families and are predominantly spoken in Africa and Asia {--} regions characterised by a relatively limited availability of NLP resources. Each instance in the SemRel datasets is a sentence pair associated with a score that represents the degree of semantic textual relatedness between the two sentences. The scores are obtained using a comparative annotation framework. We describe the data collection and annotation processes, challenges when building the datasets, baseline experiments, and their impact and utility in NLP. | [
"Ousidhoum, Nedjma",
"Muhammad, Shamsuddeen",
"Abdalla, Mohamed",
"Abdulmumin, Idris",
"Ahmad, Ibrahim",
"Ahuja, Sanchit",
"Aji, Alham",
"Araujo, Vladimir",
"Ayele, Abinew",
"Baswani, Pavan",
"Beloucif, Meriem",
"Biemann, Chris",
"Bourhim, Sofia",
"Kock, Christine",
"Dekebo, Genet",
"Hourrane, Oumaima",
"Kanumolu, Gopich",
"",
"Madasu, Lokesh",
"Rutunda, Samuel",
"Shrivastava, Manish",
"Solorio, Thamar",
"Surange, Nirmal",
"Tilaye, Hailegnaw",
"Vishnubhotla, Krishnapriya",
"Winata, Genta",
"Yimam, Seid",
"Mohammad, Saif"
]
| {S}em{R}el2024: A Collection of Semantic Textual Relatedness Datasets for 13 Languages | findings-acl.147 | Poster | 2110.04845v4 |
https://aclanthology.org/2024.findings-acl.148.bib | @inproceedings{yang-quan-2024-alirector,
title = "Alirector: Alignment-Enhanced {C}hinese Grammatical Error Corrector",
author = "Yang, Haihui and
Quan, Xiaojun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.148",
pages = "2531--2546",
abstract = "Chinese grammatical error correction (CGEC) faces serious overcorrection challenges when employing autoregressive generative models such as sequence-to-sequence (Seq2Seq) models and decoder-only large language models (LLMs). While previous methods aim to address overcorrection in Seq2Seq models, they are difficult to adapt to decoder-only LLMs. In this paper, we propose an alignment-enhanced corrector for the overcorrection problem that applies to both Seq2Seq models and decoder-only LLMs. Our method first trains a correction model to generate an initial correction of the source sentence. Then, we combine the source sentence with the initial correction and feed it through an alignment model for another round of correction, aiming to enforce the alignment model to focus on potential overcorrection. Moreover, to enhance the model{'}s ability to identify nuances, we further explore the reverse alignment of the source sentence and the initial correction. Finally, we transfer the alignment knowledge from two alignment models to the correction model, instructing it on how to avoid overcorrection. Experimental results on three CGEC datasets demonstrate the effectiveness of our approach in alleviating overcorrection and improving overall performance. Our code has been made publicly available.",
}
| Chinese grammatical error correction (CGEC) faces serious overcorrection challenges when employing autoregressive generative models such as sequence-to-sequence (Seq2Seq) models and decoder-only large language models (LLMs). While previous methods aim to address overcorrection in Seq2Seq models, they are difficult to adapt to decoder-only LLMs. In this paper, we propose an alignment-enhanced corrector for the overcorrection problem that applies to both Seq2Seq models and decoder-only LLMs. Our method first trains a correction model to generate an initial correction of the source sentence. Then, we combine the source sentence with the initial correction and feed it through an alignment model for another round of correction, aiming to enforce the alignment model to focus on potential overcorrection. Moreover, to enhance the model{'}s ability to identify nuances, we further explore the reverse alignment of the source sentence and the initial correction. Finally, we transfer the alignment knowledge from two alignment models to the correction model, instructing it on how to avoid overcorrection. Experimental results on three CGEC datasets demonstrate the effectiveness of our approach in alleviating overcorrection and improving overall performance. Our code has been made publicly available. | [
"Yang, Haihui",
"Quan, Xiaojun"
]
| Alirector: Alignment-Enhanced {C}hinese Grammatical Error Corrector | findings-acl.148 | Poster | 2402.04601v2 |
https://aclanthology.org/2024.findings-acl.149.bib | @inproceedings{alikasifoglu-etal-2024-vispool,
title = "{VISP}ool: Enhancing Transformer Encoders with Vector Visibility Graph Neural Networks",
author = "Alika{\c{s}}ifo{\u{g}}lu, Tuna and
Aras, Arda and
Koc, Aykut",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.149",
pages = "2547--2556",
abstract = "The emergence of transformers has revolutionized natural language processing (NLP), as evidenced in various NLP tasks. While graph neural networks (GNNs) show recent promise in NLP, they are not standalone replacements for transformers. Rather, recent research explores combining transformers and GNNs. Existing GNN-based approaches rely on static graph construction methods requiring excessive text processing, and most of them are not scalable with the increasing document and word counts. We address these limitations by proposing a novel dynamic graph construction method for text documents based on vector visibility graphs (VVGs) generated from transformer output. Then, we introduce visibility pooler (VISPool), a scalable model architecture that seamlessly integrates VVG convolutional networks into transformer pipelines. We evaluate the proposed model on the General Language Understanding Evaluation (GLUE) benchmark datasets. VISPool outperforms the baselines with less trainable parameters, demonstrating the viability of the visibility-based graph construction method for enhancing transformers with GNNs.",
}
| The emergence of transformers has revolutionized natural language processing (NLP), as evidenced in various NLP tasks. While graph neural networks (GNNs) show recent promise in NLP, they are not standalone replacements for transformers. Rather, recent research explores combining transformers and GNNs. Existing GNN-based approaches rely on static graph construction methods requiring excessive text processing, and most of them are not scalable with the increasing document and word counts. We address these limitations by proposing a novel dynamic graph construction method for text documents based on vector visibility graphs (VVGs) generated from transformer output. Then, we introduce visibility pooler (VISPool), a scalable model architecture that seamlessly integrates VVG convolutional networks into transformer pipelines. We evaluate the proposed model on the General Language Understanding Evaluation (GLUE) benchmark datasets. VISPool outperforms the baselines with less trainable parameters, demonstrating the viability of the visibility-based graph construction method for enhancing transformers with GNNs. | [
"Alika{\\c{s}}ifo{\\u{g}}lu, Tuna",
"Aras, Arda",
"Koc, Aykut"
]
| {VISP}ool: Enhancing Transformer Encoders with Vector Visibility Graph Neural Networks | findings-acl.149 | Poster | 2006.03996v2 |
https://aclanthology.org/2024.findings-acl.150.bib | @inproceedings{vishnubhotla-etal-2024-emotion,
title = "The Emotion Dynamics of Literary Novels",
author = "Vishnubhotla, Krishnapriya and
Hammond, Adam and
Hirst, Graeme and
Mohammad, Saif",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.150",
pages = "2557--2574",
abstract = "Stories are rich in the emotions they exhibit in their narratives and evoke in the readers. The emotional journeys of the various characters within a story are central to their appeal. Computational analysis of the emotions of novels, however, has rarely examined the variation in the emotional trajectories of the different characters within them, instead considering the entire novel to represent a single story arc. In this work, we use character dialogue to distinguish between the emotion arcs of the narration and the various characters. We analyze the emotion arcs of the various characters in a dataset of English literary novels using the framework of Utterance Emotion Dynamics. Our findings show that the narration and the dialogue largely express disparate emotions through the course of a novel, and that the commonalities or differences in the emotional arcs of stories are more accurately captured by those associated with individual characters.",
}
| Stories are rich in the emotions they exhibit in their narratives and evoke in the readers. The emotional journeys of the various characters within a story are central to their appeal. Computational analysis of the emotions of novels, however, has rarely examined the variation in the emotional trajectories of the different characters within them, instead considering the entire novel to represent a single story arc. In this work, we use character dialogue to distinguish between the emotion arcs of the narration and the various characters. We analyze the emotion arcs of the various characters in a dataset of English literary novels using the framework of Utterance Emotion Dynamics. Our findings show that the narration and the dialogue largely express disparate emotions through the course of a novel, and that the commonalities or differences in the emotional arcs of stories are more accurately captured by those associated with individual characters. | [
"Vishnubhotla, Krishnapriya",
"Hammond, Adam",
"Hirst, Graeme",
"Mohammad, Saif"
]
| The Emotion Dynamics of Literary Novels | findings-acl.150 | Poster | 2403.02474v1 |
https://aclanthology.org/2024.findings-acl.151.bib | @inproceedings{yao-barbosa-2024-accurate,
title = "Accurate and Nuanced Open-{QA} Evaluation Through Textual Entailment",
author = "Yao, Peiran and
Barbosa, Denilson",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.151",
pages = "2575--2587",
abstract = "Open-domain question answering (Open-QA) is a common task for evaluating large language models (LLMs). However, current Open-QA evaluations are criticized for the ambiguity in questions and the lack of semantic understanding in evaluators. Complex evaluators, powered by foundation models or LLMs and pertaining to semantic equivalence, still deviate from human judgments by a large margin. We propose to study the entailment relations of answers to identify more informative and more general system answers, offering a much closer evaluation to human judgment on both NaturalQuestions and TriviaQA while being learning-free. The entailment-based evaluation we propose allows the assignment of bonus or partial marks by quantifying the inference gap between answers, enabling a nuanced ranking of answer correctness that has higher AUC than current methods.",
}
| Open-domain question answering (Open-QA) is a common task for evaluating large language models (LLMs). However, current Open-QA evaluations are criticized for the ambiguity in questions and the lack of semantic understanding in evaluators. Complex evaluators, powered by foundation models or LLMs and pertaining to semantic equivalence, still deviate from human judgments by a large margin. We propose to study the entailment relations of answers to identify more informative and more general system answers, offering a much closer evaluation to human judgment on both NaturalQuestions and TriviaQA while being learning-free. The entailment-based evaluation we propose allows the assignment of bonus or partial marks by quantifying the inference gap between answers, enabling a nuanced ranking of answer correctness that has higher AUC than current methods. | [
"Yao, Peiran",
"Barbosa, Denilson"
]
| Accurate and Nuanced Open-{QA} Evaluation Through Textual Entailment | findings-acl.151 | Poster | 2405.16702v1 |
https://aclanthology.org/2024.findings-acl.152.bib | @inproceedings{dimakis-etal-2024-dictionary,
title = "Dictionary-Aided Translation for Handling Multi-Word Expressions in Low-Resource Languages",
author = "Dimakis, Antonios and
Markantonatou, Stella and
Anastasopoulos, Antonios",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.152",
pages = "2588--2595",
abstract = "Multi-word expressions (MWEs) present unique challenges in natural language processing (NLP), particularly within the context of translation systems, due to their inherent scarcity, non-compositional nature, and other distinct lexical and morphosyntactic characteristics, issues that are exacerbated in low-resource settings.In this study, we elucidate and attempt to address these challenges by leveraging a substantial corpus of human-annotated Greek MWEs. To address the complexity of translating such phrases, we propose a novel method leveraging an available out-of-context lexicon.We assess the translation capabilities of current state-of-the-art systems on this task, employing both automated metrics and human evaluators.We find that by using our method when applicable, the performance of current systems can be significantly improved, however these models are still unable to produce translations comparable to those of a human speaker.",
}
| Multi-word expressions (MWEs) present unique challenges in natural language processing (NLP), particularly within the context of translation systems, due to their inherent scarcity, non-compositional nature, and other distinct lexical and morphosyntactic characteristics, issues that are exacerbated in low-resource settings.In this study, we elucidate and attempt to address these challenges by leveraging a substantial corpus of human-annotated Greek MWEs. To address the complexity of translating such phrases, we propose a novel method leveraging an available out-of-context lexicon.We assess the translation capabilities of current state-of-the-art systems on this task, employing both automated metrics and human evaluators.We find that by using our method when applicable, the performance of current systems can be significantly improved, however these models are still unable to produce translations comparable to those of a human speaker. | [
"Dimakis, Antonios",
"Markantonatou, Stella",
"Anastasopoulos, Antonios"
]
| Dictionary-Aided Translation for Handling Multi-Word Expressions in Low-Resource Languages | findings-acl.152 | Poster | 1809.01431v2 |
https://aclanthology.org/2024.findings-acl.153.bib | @inproceedings{li-etal-2024-lans,
title = "{LANS}: A Layout-Aware Neural Solver for Plane Geometry Problem",
author = "Li, Zhong-Zhi and
Zhang, Ming-Liang and
Yin, Fei and
Liu, Cheng-Lin",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.153",
pages = "2596--2608",
abstract = "Geometry problem solving (GPS) is a challenging mathematical reasoning task requiring multi-modal understanding, fusion, and reasoning. Existing neural solvers take GPS as a vision-language task but are short in the representation of geometry diagrams that carry rich and complex layout information. In this paper, we propose a layout-aware neural solver named LANS, integrated with two new modules: multimodal layout-aware pre-trained language module (MLA-PLM) and layout-aware fusion attention (LA-FA). MLA-PLM adopts structural-semantic pre-training (SSP) to implement global relationship modeling, and point-match pre-training (PMP) to achieve alignment between visual points and textual points. LA-FA employs a layout-aware attention mask to realize point-guided cross-modal fusion for further boosting layout awareness of LANS. Extensive experiments on datasets Geometry3K and PGPS9K validate the effectiveness of the layout-aware modules and superior problem-solving performance of our LANS solver, over existing symbolic and neural solvers. We have made our code and data publicly available.",
}
| Geometry problem solving (GPS) is a challenging mathematical reasoning task requiring multi-modal understanding, fusion, and reasoning. Existing neural solvers take GPS as a vision-language task but are short in the representation of geometry diagrams that carry rich and complex layout information. In this paper, we propose a layout-aware neural solver named LANS, integrated with two new modules: multimodal layout-aware pre-trained language module (MLA-PLM) and layout-aware fusion attention (LA-FA). MLA-PLM adopts structural-semantic pre-training (SSP) to implement global relationship modeling, and point-match pre-training (PMP) to achieve alignment between visual points and textual points. LA-FA employs a layout-aware attention mask to realize point-guided cross-modal fusion for further boosting layout awareness of LANS. Extensive experiments on datasets Geometry3K and PGPS9K validate the effectiveness of the layout-aware modules and superior problem-solving performance of our LANS solver, over existing symbolic and neural solvers. We have made our code and data publicly available. | [
"Li, Zhong-Zhi",
"Zhang, Ming-Liang",
"Yin, Fei",
"Liu, Cheng-Lin"
]
| {LANS}: A Layout-Aware Neural Solver for Plane Geometry Problem | findings-acl.153 | Poster | 2311.16476v2 |
https://aclanthology.org/2024.findings-acl.154.bib | @inproceedings{ding-etal-2024-knowledge,
title = "Knowledge Crosswords: Geometric Knowledge Reasoning with Large Language Models",
author = "Ding, Wenxuan and
Feng, Shangbin and
Liu, Yuhan and
Tan, Zhaoxuan and
Balachandran, Vidhisha and
He, Tianxing and
Tsvetkov, Yulia",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.154",
pages = "2609--2636",
abstract = "We propose Knowledge Crosswords, a geometric knowledge reasoning benchmark consisting of incomplete knowledge networks bounded by structured factual constraints, where LLMs are tasked with inferring the missing facts to meet all constraints. The novel setting of geometric knowledge reasoning necessitates new LM abilities beyond existing atomic/linear multi-hop QA, such as backtracking, verifying facts and constraints, reasoning with uncertainty, and more. Knowledge Crosswords contains 2,101 individual problems, covering diverse knowledge domains, and is further divided into three difficulty levels. We conduct extensive experiments to evaluate existing LLMs and approaches on Knowledge Crosswords. Results demonstrate that baseline approaches struggle with larger knowledge networks and semantically-equivalent entity distractors. In light of their limitations, we propose two new approaches, Staged Prompting and Verify-All, to augment LLMs{'} abilities for error-aware backtracking and constraint verification. Our Verify-All significantly outperforms prior methods and is more robust towards problems in the hard subset. Further analysis shows that geometric knowledge reasoning poses new challenges to LLMs{'} knowledge abilities, particularly in robustness towards varying option orders, complex structural constraints in knowledge networks, {``}none of the above{''} scenarios, and more.",
}
| We propose Knowledge Crosswords, a geometric knowledge reasoning benchmark consisting of incomplete knowledge networks bounded by structured factual constraints, where LLMs are tasked with inferring the missing facts to meet all constraints. The novel setting of geometric knowledge reasoning necessitates new LM abilities beyond existing atomic/linear multi-hop QA, such as backtracking, verifying facts and constraints, reasoning with uncertainty, and more. Knowledge Crosswords contains 2,101 individual problems, covering diverse knowledge domains, and is further divided into three difficulty levels. We conduct extensive experiments to evaluate existing LLMs and approaches on Knowledge Crosswords. Results demonstrate that baseline approaches struggle with larger knowledge networks and semantically-equivalent entity distractors. In light of their limitations, we propose two new approaches, Staged Prompting and Verify-All, to augment LLMs{'} abilities for error-aware backtracking and constraint verification. Our Verify-All significantly outperforms prior methods and is more robust towards problems in the hard subset. Further analysis shows that geometric knowledge reasoning poses new challenges to LLMs{'} knowledge abilities, particularly in robustness towards varying option orders, complex structural constraints in knowledge networks, {``}none of the above{''} scenarios, and more. | [
"Ding, Wenxuan",
"Feng, Shangbin",
"Liu, Yuhan",
"Tan, Zhaoxuan",
"Balach",
"ran, Vidhisha",
"He, Tianxing",
"Tsvetkov, Yulia"
]
| Knowledge Crosswords: Geometric Knowledge Reasoning with Large Language Models | findings-acl.154 | Poster | 2406.16655v1 |
https://aclanthology.org/2024.findings-acl.155.bib | @inproceedings{wan-etal-2024-dell,
title = "{DELL}: Generating Reactions and Explanations for {LLM}-Based Misinformation Detection",
author = "Wan, Herun and
Feng, Shangbin and
Tan, Zhaoxuan and
Wang, Heng and
Tsvetkov, Yulia and
Luo, Minnan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.155",
pages = "2637--2667",
abstract = "Large language models are limited by challenges in factuality and hallucinations to be directly employed off-the-shelf for judging the veracity of news articles, where factual accuracy is paramount. In this work, we propose DELL that identifies three key stages in misinformation detection where LLMs could be incorporated as part of the pipeline: 1) LLMs could generate news reactions to represent diverse perspectives and simulate user-news interaction networks; 2) LLMs could generate explanations for proxy tasks (e.g., sentiment, stance) to enrich the contexts of news articles and produce experts specializing in various aspects of news understanding; 3) LLMs could merge task-specific experts and provide an overall prediction by incorporating the predictions and confidence scores of varying experts. Extensive experiments on seven datasets with three LLMs demonstrate that DELL outperforms state-of-the-art baselines by up to 16.8{\%} in macro f1-score. Further analysis reveals that the generated reactions and explanations are greatly helpful in misinformation detection, while our proposed LLM-guided expert merging helps produce better-calibrated predictions.",
}
| Large language models are limited by challenges in factuality and hallucinations to be directly employed off-the-shelf for judging the veracity of news articles, where factual accuracy is paramount. In this work, we propose DELL that identifies three key stages in misinformation detection where LLMs could be incorporated as part of the pipeline: 1) LLMs could generate news reactions to represent diverse perspectives and simulate user-news interaction networks; 2) LLMs could generate explanations for proxy tasks (e.g., sentiment, stance) to enrich the contexts of news articles and produce experts specializing in various aspects of news understanding; 3) LLMs could merge task-specific experts and provide an overall prediction by incorporating the predictions and confidence scores of varying experts. Extensive experiments on seven datasets with three LLMs demonstrate that DELL outperforms state-of-the-art baselines by up to 16.8{\%} in macro f1-score. Further analysis reveals that the generated reactions and explanations are greatly helpful in misinformation detection, while our proposed LLM-guided expert merging helps produce better-calibrated predictions. | [
"Wan, Herun",
"Feng, Shangbin",
"Tan, Zhaoxuan",
"Wang, Heng",
"Tsvetkov, Yulia",
"Luo, Minnan"
]
| {DELL}: Generating Reactions and Explanations for {LLM}-Based Misinformation Detection | findings-acl.155 | Poster | 2405.08254v1 |
https://aclanthology.org/2024.findings-acl.156.bib | @inproceedings{shen-etal-2024-language,
title = "The Language Barrier: Dissecting Safety Challenges of {LLM}s in Multilingual Contexts",
author = "Shen, Lingfeng and
Tan, Weiting and
Chen, Sihao and
Chen, Yunmo and
Zhang, Jingyu and
Xu, Haoran and
Zheng, Boyuan and
Koehn, Philipp and
Khashabi, Daniel",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.156",
pages = "2668--2680",
abstract = "As the influence of large language models (LLMs) spans across global communities, their safety challenges in multilingual settings become paramount for alignment research. This paper examines the variations in safety challenges faced by LLMs across different languages and discusses approaches to alleviating such concerns. By comparing how state-of-the-art LLMs respond to the same set of malicious prompts written in higher- vs. lower-resource languages,we observe that (1) LLMs tend to generate unsafe responses much more often when a malicious prompt is written in a lower-resource language, and (2) LLMs tend to generate more irrelevant responses to malicious prompts in lower-resource languages. To understand where the discrepancy can be attributed, we study the effect of instruction tuning with reinforcement learning from human feedback (RLHF) or supervised finetuning (SFT) on the HH-RLHF dataset. Surprisingly, while training with high-resource languages improves model alignment, training in lower-resource languages yields minimal improvement. This suggests that the bottleneck of cross-lingual alignment is rooted in the pretraining stage. Our findings highlight the challenges in cross-lingual LLM safety, and we hope they inform future research in this direction.",
}
| As the influence of large language models (LLMs) spans across global communities, their safety challenges in multilingual settings become paramount for alignment research. This paper examines the variations in safety challenges faced by LLMs across different languages and discusses approaches to alleviating such concerns. By comparing how state-of-the-art LLMs respond to the same set of malicious prompts written in higher- vs. lower-resource languages,we observe that (1) LLMs tend to generate unsafe responses much more often when a malicious prompt is written in a lower-resource language, and (2) LLMs tend to generate more irrelevant responses to malicious prompts in lower-resource languages. To understand where the discrepancy can be attributed, we study the effect of instruction tuning with reinforcement learning from human feedback (RLHF) or supervised finetuning (SFT) on the HH-RLHF dataset. Surprisingly, while training with high-resource languages improves model alignment, training in lower-resource languages yields minimal improvement. This suggests that the bottleneck of cross-lingual alignment is rooted in the pretraining stage. Our findings highlight the challenges in cross-lingual LLM safety, and we hope they inform future research in this direction. | [
"Shen, Lingfeng",
"Tan, Weiting",
"Chen, Sihao",
"Chen, Yunmo",
"Zhang, Jingyu",
"Xu, Haoran",
"Zheng, Boyuan",
"Koehn, Philipp",
"Khashabi, Daniel"
]
| The Language Barrier: Dissecting Safety Challenges of {LLM}s in Multilingual Contexts | findings-acl.156 | Poster | 2407.07342v1 |
https://aclanthology.org/2024.findings-acl.157.bib | @inproceedings{kang-etal-2024-self,
title = "Self-Specialization: Uncovering Latent Expertise within Large Language Models",
author = "Kang, Junmo and
Luo, Hongyin and
Zhu, Yada and
Hansen, Jacob and
Glass, James and
Cox, David and
Ritter, Alan and
Feris, Rogerio and
Karlinsky, Leonid",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.157",
pages = "2681--2706",
abstract = "Recent works have demonstrated the effectiveness of self-alignment in which a large language model is aligned to follow general instructions using instructional data generated from the model itself starting from a handful of human-written seeds. Instead of general alignment, in this work, we focus on self-alignment for expert domain specialization (e.g., biomedicine, finance). As a preliminary, we quantitively show the marginal effect that generic instruction-following training has on downstream expert domains{'} performance. To remedy this, we propose self-specialization - allowing for effective model specialization while achieving cross-task generalization by leveraging only a few labeled seeds. Self-specialization offers a data- and parameter-efficient way of {``}carving out{''} an expert model out of a generalist pre-trained LLM. Exploring a variety of popular open large models as a base for specialization, our experimental results in both biomedical and financial domains show that our self-specialized models outperform their base models by a large margin, and even larger models that are generally instruction-tuned or that have been adapted to the target domain by other means.",
}
| Recent works have demonstrated the effectiveness of self-alignment in which a large language model is aligned to follow general instructions using instructional data generated from the model itself starting from a handful of human-written seeds. Instead of general alignment, in this work, we focus on self-alignment for expert domain specialization (e.g., biomedicine, finance). As a preliminary, we quantitively show the marginal effect that generic instruction-following training has on downstream expert domains{'} performance. To remedy this, we propose self-specialization - allowing for effective model specialization while achieving cross-task generalization by leveraging only a few labeled seeds. Self-specialization offers a data- and parameter-efficient way of {``}carving out{''} an expert model out of a generalist pre-trained LLM. Exploring a variety of popular open large models as a base for specialization, our experimental results in both biomedical and financial domains show that our self-specialized models outperform their base models by a large margin, and even larger models that are generally instruction-tuned or that have been adapted to the target domain by other means. | [
"Kang, Junmo",
"Luo, Hongyin",
"Zhu, Yada",
"Hansen, Jacob",
"Glass, James",
"Cox, David",
"Ritter, Alan",
"Feris, Rogerio",
"Karlinsky, Leonid"
]
| Self-Specialization: Uncovering Latent Expertise within Large Language Models | findings-acl.157 | Poster | 1705.02518v1 |
https://aclanthology.org/2024.findings-acl.158.bib | @inproceedings{xu-etal-2024-fuse,
title = "{FUSE}: Measure-Theoretic Compact Fuzzy Set Representation for Taxonomy Expansion",
author = "Xu, Fred and
Jiang, Song and
Huang, Zijie and
Luo, Xiao and
Zhang, Shichang and
Chen, Yuanzhou and
Sun, Yizhou",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.158",
pages = "2707--2720",
abstract = "Taxonomy Expansion, which relies on modeling concepts and concept relations, can be formulated as a set representation learning task. The generalization of set, fuzzy set, incorporates uncertainty and measures the information within a semantic concept, making it suitable for concept modeling. Existing works usually model sets as vectors or geometric objects such as boxes, which are not closed under set operations. In this work, we propose a sound and efficient formulation of set representation learning based on its volume approximation as a fuzzy set. The resulting embedding framework, \textit{Fuzzy Set Embedding}, satisfies all set operations and compactly approximates the underlying fuzzy set, hence preserving information while being efficient to learn, relying on minimum neural architecture. We empirically demonstrate the power of FUSE on the task of taxonomy expansion, where FUSE achieves remarkable improvements up to 23{\%} compared with existing baselines. Our work marks the first attempt to understand and efficiently compute the embeddings of fuzzy sets.",
}
| Taxonomy Expansion, which relies on modeling concepts and concept relations, can be formulated as a set representation learning task. The generalization of set, fuzzy set, incorporates uncertainty and measures the information within a semantic concept, making it suitable for concept modeling. Existing works usually model sets as vectors or geometric objects such as boxes, which are not closed under set operations. In this work, we propose a sound and efficient formulation of set representation learning based on its volume approximation as a fuzzy set. The resulting embedding framework, \textit{Fuzzy Set Embedding}, satisfies all set operations and compactly approximates the underlying fuzzy set, hence preserving information while being efficient to learn, relying on minimum neural architecture. We empirically demonstrate the power of FUSE on the task of taxonomy expansion, where FUSE achieves remarkable improvements up to 23{\%} compared with existing baselines. Our work marks the first attempt to understand and efficiently compute the embeddings of fuzzy sets. | [
"Xu, Fred",
"Jiang, Song",
"Huang, Zijie",
"Luo, Xiao",
"Zhang, Shichang",
"Chen, Yuanzhou",
"Sun, Yizhou"
]
| {FUSE}: Measure-Theoretic Compact Fuzzy Set Representation for Taxonomy Expansion | findings-acl.158 | Poster | 1304.1827v1 |
https://aclanthology.org/2024.findings-acl.159.bib | @inproceedings{servantez-etal-2024-chain,
title = "Chain of Logic: Rule-Based Reasoning with Large Language Models",
author = "Servantez, Sergio and
Barrow, Joe and
Hammond, Kristian and
Jain, Rajiv",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.159",
pages = "2721--2733",
abstract = "Rule-based reasoning, a fundamental type of legal reasoning, enables us to draw conclusions by accurately applying a rule to a set of facts. We explore causal language models as rule-based reasoners, specifically with respect to compositional rules - rules consisting of multiple elements which form a complex logical expression. Reasoning about compositional rules is challenging because it requires multiple reasoning steps, and attending to the logical relationships between elements. We introduce a new prompting method, Chain of Logic, which elicits rule-based reasoning through decomposition (solving elements as independent threads of logic), and recomposition (recombining these sub-answers to resolve the underlying logical expression). This method was inspired by the IRAC (Issue, Rule, Application, Conclusion) framework, a sequential reasoning approach used by lawyers. We evaluate chain of logic across eight rule-based reasoning tasks involving three distinct compositional rules from the LegalBench benchmark and demonstrate it consistently outperforms other prompting methods, including chain of thought and self-ask, using open-source and commercial language models.",
}
| Rule-based reasoning, a fundamental type of legal reasoning, enables us to draw conclusions by accurately applying a rule to a set of facts. We explore causal language models as rule-based reasoners, specifically with respect to compositional rules - rules consisting of multiple elements which form a complex logical expression. Reasoning about compositional rules is challenging because it requires multiple reasoning steps, and attending to the logical relationships between elements. We introduce a new prompting method, Chain of Logic, which elicits rule-based reasoning through decomposition (solving elements as independent threads of logic), and recomposition (recombining these sub-answers to resolve the underlying logical expression). This method was inspired by the IRAC (Issue, Rule, Application, Conclusion) framework, a sequential reasoning approach used by lawyers. We evaluate chain of logic across eight rule-based reasoning tasks involving three distinct compositional rules from the LegalBench benchmark and demonstrate it consistently outperforms other prompting methods, including chain of thought and self-ask, using open-source and commercial language models. | [
"Servantez, Sergio",
"Barrow, Joe",
"Hammond, Kristian",
"Jain, Rajiv"
]
| Chain of Logic: Rule-Based Reasoning with Large Language Models | findings-acl.159 | Poster | 2403.19167v1 |
https://aclanthology.org/2024.findings-acl.160.bib | @inproceedings{chiang-lee-2024-merging,
title = "Merging Facts, Crafting Fallacies: Evaluating the Contradictory Nature of Aggregated Factual Claims in Long-Form Generations",
author = "Chiang, Cheng-Han and
Lee, Hung-yi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.160",
pages = "2734--2751",
abstract = "Long-form generations from large language models (LLMs) contain a mix of factual and non-factual claims, making evaluating factuality difficult.Prior works evaluate the factuality of a long paragraph by decomposing it into multiple facts, verifying those facts independently, and aggregating the results.Such methods assume that combining factual claims forms a factual paragraph.The above assumption can be violated: we show that strong open-source models like Llama-chat can generate paragraphs that contain verifiable facts, but the facts are combined into a non-factual paragraph due to entity ambiguity.We further reveal that existing factuality metrics, including FActScore and citation recall, cannot properly evaluate these non-factual paragraphs and overestimate their factuality.To address this, we introduce an enhanced metric, **D-FActScore**, specifically designed for content with ambiguous entities.We evaluate the D-FActScores of people biographies generated by retrieval-augmented LLMs.We show that D-FActScore can better assess the factuality of paragraphs with entity ambiguity than FActScore.We also find that four widely used open-source LLMs tend to mix information of distinct entities to form non-factual paragraphs, making their D-FActScore much lower than FActScore by over 10{\%}.",
}
| Long-form generations from large language models (LLMs) contain a mix of factual and non-factual claims, making evaluating factuality difficult.Prior works evaluate the factuality of a long paragraph by decomposing it into multiple facts, verifying those facts independently, and aggregating the results.Such methods assume that combining factual claims forms a factual paragraph.The above assumption can be violated: we show that strong open-source models like Llama-chat can generate paragraphs that contain verifiable facts, but the facts are combined into a non-factual paragraph due to entity ambiguity.We further reveal that existing factuality metrics, including FActScore and citation recall, cannot properly evaluate these non-factual paragraphs and overestimate their factuality.To address this, we introduce an enhanced metric, **D-FActScore**, specifically designed for content with ambiguous entities.We evaluate the D-FActScores of people biographies generated by retrieval-augmented LLMs.We show that D-FActScore can better assess the factuality of paragraphs with entity ambiguity than FActScore.We also find that four widely used open-source LLMs tend to mix information of distinct entities to form non-factual paragraphs, making their D-FActScore much lower than FActScore by over 10{\%}. | [
"Chiang, Cheng-Han",
"Lee, Hung-yi"
]
| Merging Facts, Crafting Fallacies: Evaluating the Contradictory Nature of Aggregated Factual Claims in Long-Form Generations | findings-acl.160 | Poster | 2402.05629v4 |
https://aclanthology.org/2024.findings-acl.161.bib | @inproceedings{merrill-etal-2024-learn,
title = "Can You Learn Semantics Through Next-Word Prediction? The Case of Entailment",
author = "Merrill, William and
Wu, Zhaofeng and
Naka, Norihito and
Kim, Yoon and
Linzen, Tal",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.161",
pages = "2752--2773",
abstract = "Do LMs infer the semantics of text from co-occurrence patterns in their training data? Merrill et al. (2022) argue that, in theory, sentence co-occurrence probabilities predicted by an optimal LM should reflect the entailment relationship of the constituent sentences, but it is unclear whether probabilities predicted by neural LMs encode entailment in this way because of strong assumptions made by Merrill et al. (namely, that humans always avoid redundancy). In this work, we investigate whether their theory can be used to decode entailment relations from neural LMs. We find that a test similar to theirs can decode entailment relations between natural sentences, well above random chance, though not perfectly, across many datasets and LMs. This suggests LMs implicitly model aspects of semantics to predict semantic effects on sentence co-occurrence patterns. However, we find the test that predicts entailment in practice works in the opposite direction to the theoretical test. We thus revisit the assumptions underlying the original test, finding its derivation did not adequately account for redundancy in human-written text. We argue that better accounting for redundancy related to *explanations* might derive the observed flipped test and, more generally, improve computational models of speakers in linguistics.",
}
| Do LMs infer the semantics of text from co-occurrence patterns in their training data? Merrill et al. (2022) argue that, in theory, sentence co-occurrence probabilities predicted by an optimal LM should reflect the entailment relationship of the constituent sentences, but it is unclear whether probabilities predicted by neural LMs encode entailment in this way because of strong assumptions made by Merrill et al. (namely, that humans always avoid redundancy). In this work, we investigate whether their theory can be used to decode entailment relations from neural LMs. We find that a test similar to theirs can decode entailment relations between natural sentences, well above random chance, though not perfectly, across many datasets and LMs. This suggests LMs implicitly model aspects of semantics to predict semantic effects on sentence co-occurrence patterns. However, we find the test that predicts entailment in practice works in the opposite direction to the theoretical test. We thus revisit the assumptions underlying the original test, finding its derivation did not adequately account for redundancy in human-written text. We argue that better accounting for redundancy related to *explanations* might derive the observed flipped test and, more generally, improve computational models of speakers in linguistics. | [
"Merrill, William",
"Wu, Zhaofeng",
"Naka, Norihito",
"Kim, Yoon",
"Linzen, Tal"
]
| Can You Learn Semantics Through Next-Word Prediction? The Case of Entailment | findings-acl.161 | Poster | 2402.13956v3 |
https://aclanthology.org/2024.findings-acl.162.bib | @inproceedings{ma-etal-2024-simulated,
title = "Simulated Misinformation Susceptibility ({SMISTS}): Enhancing Misinformation Research with Large Language Model Simulations",
author = "Ma, Weicheng and
Deng, Chunyuan and
Moossavi, Aram and
Wang, Lili and
Vosoughi, Soroush and
Yang, Diyi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.162",
pages = "2774--2788",
abstract = "Psychological inoculation, a strategy designed to build resistance against persuasive misinformation, has shown efficacy in curbing its spread and mitigating its adverse effects at early stages. Despite its effectiveness, the design and optimization of these inoculations typically demand substantial human and financial resources, primarily due to the need for repeated experimental trials. To address these challenges, this paper introduces Simulated Misinformation Susceptibility Tests (SMISTs), leveraging Large Language Models (LLMs) to simulate participant responses in misinformation studies. SMIST employs a life experience-driven simulation methodology, which accounts for various aspects of participants{'} backgrounds, to mitigate common issues of caricatures and stereotypes in LLM simulations and enhance response diversity. Our extensive experimentation demonstrates that SMIST, utilizing GPT-4 as the backend model, yields results that align closely with those obtained from human-subject studies in misinformation susceptibility. This alignment suggests that LLMs can effectively serve as proxies in evaluating the impact of psychological inoculations. Moreover, SMIST offers the critical benefit of being applicable to emerging or anticipated misinformation scenarios without exposing human participants to potentially harmful content. This characteristic of SMIST not only preserves participant safety but also expands the scope of misinformation research to include more sensitive or speculative topics.",
}
| Psychological inoculation, a strategy designed to build resistance against persuasive misinformation, has shown efficacy in curbing its spread and mitigating its adverse effects at early stages. Despite its effectiveness, the design and optimization of these inoculations typically demand substantial human and financial resources, primarily due to the need for repeated experimental trials. To address these challenges, this paper introduces Simulated Misinformation Susceptibility Tests (SMISTs), leveraging Large Language Models (LLMs) to simulate participant responses in misinformation studies. SMIST employs a life experience-driven simulation methodology, which accounts for various aspects of participants{'} backgrounds, to mitigate common issues of caricatures and stereotypes in LLM simulations and enhance response diversity. Our extensive experimentation demonstrates that SMIST, utilizing GPT-4 as the backend model, yields results that align closely with those obtained from human-subject studies in misinformation susceptibility. This alignment suggests that LLMs can effectively serve as proxies in evaluating the impact of psychological inoculations. Moreover, SMIST offers the critical benefit of being applicable to emerging or anticipated misinformation scenarios without exposing human participants to potentially harmful content. This characteristic of SMIST not only preserves participant safety but also expands the scope of misinformation research to include more sensitive or speculative topics. | [
"Ma, Weicheng",
"Deng, Chunyuan",
"Moossavi, Aram",
"Wang, Lili",
"Vosoughi, Soroush",
"Yang, Diyi"
]
| Simulated Misinformation Susceptibility ({SMISTS}): Enhancing Misinformation Research with Large Language Model Simulations | findings-acl.162 | Poster | null |
https://aclanthology.org/2024.findings-acl.163.bib | @inproceedings{li-etal-2024-social,
title = "Social Intelligence Data Infrastructure: Structuring the Present and Navigating the Future",
author = "Li, Minzhi and
Shi, Weiyan and
Ziems, Caleb and
Yang, Diyi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.163",
pages = "2789--2805",
abstract = "As Natural Language Processing (NLP) systems become increasingly integrated into human social life, these technologies will need to increasingly rely on social intelligence. Although there are many valuable datasets that benchmark isolated dimensions of social intelligence, there does not yet exist any body of work to join these threads into a cohesive subfield in which researchers can quickly identify research gaps and future directions. Towards this goal, we build a Social AI Data Infrastructure, which consists of a comprehensive social AI taxonomy and a data library of 480 NLP datasets. Our infrastructure allows us to analyze existing dataset efforts, and also evaluate language models{'} performance in different social intelligence aspects. Our analyses demonstrate its utility in enabling a thorough understanding of current data landscape and providing a holistic perspective on potential directions for future dataset development. We show there is a need for multifaceted datasets, increased diversity in language and culture, more long-tailed social situations, and more interactive data in future social intelligence data efforts.",
}
| As Natural Language Processing (NLP) systems become increasingly integrated into human social life, these technologies will need to increasingly rely on social intelligence. Although there are many valuable datasets that benchmark isolated dimensions of social intelligence, there does not yet exist any body of work to join these threads into a cohesive subfield in which researchers can quickly identify research gaps and future directions. Towards this goal, we build a Social AI Data Infrastructure, which consists of a comprehensive social AI taxonomy and a data library of 480 NLP datasets. Our infrastructure allows us to analyze existing dataset efforts, and also evaluate language models{'} performance in different social intelligence aspects. Our analyses demonstrate its utility in enabling a thorough understanding of current data landscape and providing a holistic perspective on potential directions for future dataset development. We show there is a need for multifaceted datasets, increased diversity in language and culture, more long-tailed social situations, and more interactive data in future social intelligence data efforts. | [
"Li, Minzhi",
"Shi, Weiyan",
"Ziems, Caleb",
"Yang, Diyi"
]
| Social Intelligence Data Infrastructure: Structuring the Present and Navigating the Future | findings-acl.163 | Poster | 2309.12568v1 |
https://aclanthology.org/2024.findings-acl.164.bib | @inproceedings{zhang-etal-2024-selective,
title = "Selective Prefix Tuning for Pre-trained Language Models",
author = "Zhang, Hongyi and
Li, Zuchao and
Wang, Ping and
Zhao, Hai",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.164",
pages = "2806--2813",
abstract = "The prevalent approach for optimizing pre-trained language models in downstream tasks is fine-tuning. However, it is both time-consuming and memory-inefficient. In response, a more efficient method called Prefix Tuning, which insert learnable vectors into each Transformer layers, has been proposed and proven effective. Recent investigations reveal that prefix tokens carry context-specific information, prompting the hypothesis that enhancing their specialization can improve model performance. To address this, we propose Selective Prefix Tuning (SPT), integrating a selective mechanism inspired by selective self-attention. Additionally, we introduce Selective Loss (SL) to encourage diversity in prefix tokens. Extensive experiments validate the effectiveness of SPT in sentence and token classification tasks. We contribute insight into understanding the role of prefix in model adaptation.",
}
| The prevalent approach for optimizing pre-trained language models in downstream tasks is fine-tuning. However, it is both time-consuming and memory-inefficient. In response, a more efficient method called Prefix Tuning, which insert learnable vectors into each Transformer layers, has been proposed and proven effective. Recent investigations reveal that prefix tokens carry context-specific information, prompting the hypothesis that enhancing their specialization can improve model performance. To address this, we propose Selective Prefix Tuning (SPT), integrating a selective mechanism inspired by selective self-attention. Additionally, we introduce Selective Loss (SL) to encourage diversity in prefix tokens. Extensive experiments validate the effectiveness of SPT in sentence and token classification tasks. We contribute insight into understanding the role of prefix in model adaptation. | [
"Zhang, Hongyi",
"Li, Zuchao",
"Wang, Ping",
"Zhao, Hai"
]
| Selective Prefix Tuning for Pre-trained Language Models | findings-acl.164 | Poster | 2203.10378v1 |
https://aclanthology.org/2024.findings-acl.165.bib | @inproceedings{guo-vosoughi-2024-modabs,
title = "{MODABS}: Multi-Objective Learning for Dynamic Aspect-Based Summarization",
author = "Guo, Xiaobo and
Vosoughi, Soroush",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.165",
pages = "2814--2827",
abstract = "The rapid proliferation of online content necessitates effective summarization methods, among which dynamic aspect-based summarization stands out. Unlike its traditional counterpart, which assumes a fixed set of known aspects, this approach adapts to the varied aspects of the input text. We introduce a novel multi-objective learning framework employing a Longformer-Encoder-Decoder for this task. The framework optimizes aspect number prediction, minimizes disparity between generated and reference summaries for each aspect, and maximizes dissimilarity across aspect-specific summaries. Extensive experiments show our method significantly outperforms baselines on three diverse datasets, largely due to the effective alignment of generated and reference aspect counts without sacrificing single-aspect summarization quality.",
}
| The rapid proliferation of online content necessitates effective summarization methods, among which dynamic aspect-based summarization stands out. Unlike its traditional counterpart, which assumes a fixed set of known aspects, this approach adapts to the varied aspects of the input text. We introduce a novel multi-objective learning framework employing a Longformer-Encoder-Decoder for this task. The framework optimizes aspect number prediction, minimizes disparity between generated and reference summaries for each aspect, and maximizes dissimilarity across aspect-specific summaries. Extensive experiments show our method significantly outperforms baselines on three diverse datasets, largely due to the effective alignment of generated and reference aspect counts without sacrificing single-aspect summarization quality. | [
"Guo, Xiaobo",
"Vosoughi, Soroush"
]
| {MODABS}: Multi-Objective Learning for Dynamic Aspect-Based Summarization | findings-acl.165 | Poster | 2406.03479v2 |
https://aclanthology.org/2024.findings-acl.166.bib | @inproceedings{zhou-bhat-2024-non,
title = "Non-compositional Expression Generation and its Continual Learning",
author = "Zhou, Jianing and
Bhat, Suma",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.166",
pages = "2828--2839",
abstract = "Non-compositional expressions are an integral part of natural language and their meanings cannot be directly derived from the meanings of their component words. Recent work has shown how their processing remains a challenge for pre-trained language models. Here we consider the fact that prior knowledge of their component words is inadequate to infer their meaning as a whole and that these expressions constitute a long-tailed process in language (based on their occurrence in corpora and their coming into use as an idiomatic expression in a continual manner). Against this backdrop, this paper studies the ability of recent pre-trained language models to generate non-compositional expressions in English and their continual learning. Formulating this as a mask infilling task termed as CLoNE, the study uncovers the combined challenges of non-compositionality and their continual learning. Using a set of three diverse idiomatic expression datasets repurposed for this task, we benchmark different large pre-trained language models and different continual learning methods on the task of non-compositional expression generation. Our experiments on the CLoNE task show that large pre-trained language models are limited in their ability to generate non-compositional expressions and available continual learning methods are inadequate for our proposed CLoNE task which calls for more effective methods for continual learning of non-compositionality. Our datasets and code will be released publicly upon acceptance.",
}
| Non-compositional expressions are an integral part of natural language and their meanings cannot be directly derived from the meanings of their component words. Recent work has shown how their processing remains a challenge for pre-trained language models. Here we consider the fact that prior knowledge of their component words is inadequate to infer their meaning as a whole and that these expressions constitute a long-tailed process in language (based on their occurrence in corpora and their coming into use as an idiomatic expression in a continual manner). Against this backdrop, this paper studies the ability of recent pre-trained language models to generate non-compositional expressions in English and their continual learning. Formulating this as a mask infilling task termed as CLoNE, the study uncovers the combined challenges of non-compositionality and their continual learning. Using a set of three diverse idiomatic expression datasets repurposed for this task, we benchmark different large pre-trained language models and different continual learning methods on the task of non-compositional expression generation. Our experiments on the CLoNE task show that large pre-trained language models are limited in their ability to generate non-compositional expressions and available continual learning methods are inadequate for our proposed CLoNE task which calls for more effective methods for continual learning of non-compositionality. Our datasets and code will be released publicly upon acceptance. | [
"Zhou, Jianing",
"Bhat, Suma"
]
| Non-compositional Expression Generation and its Continual Learning | findings-acl.166 | Poster | 1709.03842v2 |
https://aclanthology.org/2024.findings-acl.167.bib | @inproceedings{shi-etal-2024-medical,
title = "Medical Dialogue System: A Survey of Categories, Methods, Evaluation and Challenges",
author = "Shi, Xiaoming and
Liu, Zeming and
Du, Li and
Wang, Yuxuan and
Wang, Hongru and
Guo, Yuhang and
Ruan, Tong and
Xu, Jie and
Zhang, Xiaofan and
Zhang, Shaoting",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.167",
pages = "2840--2861",
abstract = "This paper surveys and organizes research works of medical dialog systems, which is an important yet challenging task. Although these systems have been surveyed in the medical community from an application perspective, a systematic review from a rigorous technical perspective has to date remained noticeably absent. As a result, an overview of the categories, methods, evaluation of medical dialogue systems remain limited and underspecified, hindering the further improvement of this area. To fill this gap, we investigate an initial pool of 325 papers from well-known computer science, natural language processing conferences and journals, and make an overview. Recently, large language models have shown strong model capacity on downstream tasks, which also reshape medical dialog systems{'} foundation.Despite the alluring practical application value, current medical dialogue systems still suffer from problems. To this end, this paper lists grand challenges of medical dialog systems, especially of large language models.",
}
| This paper surveys and organizes research works of medical dialog systems, which is an important yet challenging task. Although these systems have been surveyed in the medical community from an application perspective, a systematic review from a rigorous technical perspective has to date remained noticeably absent. As a result, an overview of the categories, methods, evaluation of medical dialogue systems remain limited and underspecified, hindering the further improvement of this area. To fill this gap, we investigate an initial pool of 325 papers from well-known computer science, natural language processing conferences and journals, and make an overview. Recently, large language models have shown strong model capacity on downstream tasks, which also reshape medical dialog systems{'} foundation.Despite the alluring practical application value, current medical dialogue systems still suffer from problems. To this end, this paper lists grand challenges of medical dialog systems, especially of large language models. | [
"Shi, Xiaoming",
"Liu, Zeming",
"Du, Li",
"Wang, Yuxuan",
"Wang, Hongru",
"Guo, Yuhang",
"Ruan, Tong",
"Xu, Jie",
"Zhang, Xiaofan",
"Zhang, Shaoting"
]
| Medical Dialogue System: A Survey of Categories, Methods, Evaluation and Challenges | findings-acl.167 | Poster | 2405.10630v1 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.