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.868.bib
|
@inproceedings{liang-etal-2024-debatrix,
title = "Debatrix: Multi-dimensional Debate Judge with Iterative Chronological Analysis Based on {LLM}",
author = "Liang, Jingcong and
Ye, Rong and
Han, Meng and
Lai, Ruofei and
Zhang, Xinyu 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.868",
pages = "14575--14595",
abstract = "How can we construct an automated debate judge to evaluate an extensive, vibrant, multi-turn debate? This task is challenging, as judging a debate involves grappling with lengthy texts, intricate argument relationships, and multi-dimensional assessments.At the same time, current research mainly focuses on short dialogues, rarely touching upon the evaluation of an entire debate.In this paper, by leveraging Large Language Models (LLMs), we propose Debatrix, which makes the analysis and assessment of multi-turn debates more aligned with majority preferences. Specifically, Debatrix features a vertical, iterative chronological analysis and a horizontal, multi-dimensional evaluation collaboration.To align with real-world debate scenarios, we introduced the PanelBench benchmark, comparing our system{'}s performance to actual debate outcomes.The findings indicate a notable enhancement over directly using LLMs for debate evaluation.Source code and benchmark data are available at https://github.com/ljcleo/debatrix.",
}
|
How can we construct an automated debate judge to evaluate an extensive, vibrant, multi-turn debate? This task is challenging, as judging a debate involves grappling with lengthy texts, intricate argument relationships, and multi-dimensional assessments.At the same time, current research mainly focuses on short dialogues, rarely touching upon the evaluation of an entire debate.In this paper, by leveraging Large Language Models (LLMs), we propose Debatrix, which makes the analysis and assessment of multi-turn debates more aligned with majority preferences. Specifically, Debatrix features a vertical, iterative chronological analysis and a horizontal, multi-dimensional evaluation collaboration.To align with real-world debate scenarios, we introduced the PanelBench benchmark, comparing our system{'}s performance to actual debate outcomes.The findings indicate a notable enhancement over directly using LLMs for debate evaluation.Source code and benchmark data are available at https://github.com/ljcleo/debatrix.
|
[
"Liang, Jingcong",
"Ye, Rong",
"Han, Meng",
"Lai, Ruofei",
"Zhang, Xinyu",
"Huang, Xuanjing",
"Wei, Zhongyu"
] |
Debatrix: Multi-dimensional Debate Judge with Iterative Chronological Analysis Based on {LLM}
|
findings-acl.868
|
Poster
|
2403.08010v3
|
https://aclanthology.org/2024.findings-acl.869.bib
|
@inproceedings{hong-etal-2024-cyclealign,
title = "{C}ycle{A}lign: Iterative Distillation from Black-box {LLM} to White-box Models for Better Human Alignment",
author = "Hong, Jixiang and
Tu, Quan and
Chen, Changyu and
Xing, Gao and
Zhang, Ji 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.869",
pages = "14596--14609",
abstract = "Language models trained on large-scale corpus often generate harmful responses that are harmful and contrary to human values. A prevalent approach for human alignment is reinforcement learning from human feedback (RLHF), utilizing algorithms such as proximal policy optimization (PPO). However, these methods are often characterized by complexity, instability, and substantial resource consumption. Considering that existing large language models (LLMs) like ChatGPT are already relatively well-aligned and cost-friendly, researchers propose to align the language model with human preferences from AI feedback. Nevertheless, the common practices, that unidirectionally distill the responses, are constrained by the inherent capability of LLMs. To address it, we introduce CycleAlign, a framework that distills alignment capabilities from the parameter-invisible LLMs (black-box) to the parameter-visible models (white-box) in an iterative manner. CycleAlign iteratively improves both the white-box and black-box models by integrating static and dynamic in-context learning and a belief alignment method.Empirical results illustrate that the model fine-tuned by CycleAlign remarkably exceeds existing methods, and achieves the state-of-the-art performance in alignment with human value.",
}
|
Language models trained on large-scale corpus often generate harmful responses that are harmful and contrary to human values. A prevalent approach for human alignment is reinforcement learning from human feedback (RLHF), utilizing algorithms such as proximal policy optimization (PPO). However, these methods are often characterized by complexity, instability, and substantial resource consumption. Considering that existing large language models (LLMs) like ChatGPT are already relatively well-aligned and cost-friendly, researchers propose to align the language model with human preferences from AI feedback. Nevertheless, the common practices, that unidirectionally distill the responses, are constrained by the inherent capability of LLMs. To address it, we introduce CycleAlign, a framework that distills alignment capabilities from the parameter-invisible LLMs (black-box) to the parameter-visible models (white-box) in an iterative manner. CycleAlign iteratively improves both the white-box and black-box models by integrating static and dynamic in-context learning and a belief alignment method.Empirical results illustrate that the model fine-tuned by CycleAlign remarkably exceeds existing methods, and achieves the state-of-the-art performance in alignment with human value.
|
[
"Hong, Jixiang",
"Tu, Quan",
"Chen, Changyu",
"Xing, Gao",
"Zhang, Ji",
"Yan, Rui"
] |
{C}ycle{A}lign: Iterative Distillation from Black-box {LLM} to White-box Models for Better Human Alignment
|
findings-acl.869
|
Poster
|
2310.16271v1
|
https://aclanthology.org/2024.findings-acl.870.bib
|
@inproceedings{nourbakhsh-etal-2024-towards,
title = "Towards a new research agenda for multimodal enterprise document understanding: What are we missing?",
author = "Nourbakhsh, Armineh and
Shah, Sameena and
Rose, Carolyn",
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.870",
pages = "14610--14622",
abstract = "The field of multimodal document understanding has produced a suite of models that have achieved stellar performance across several tasks, even coming close to human performance on certain benchmarks. Nevertheless, the application of these models to real-world enterprise datasets remains constrained by a number of limitations. In this position paper, we discuss these limitations in the context of three key aspects of research: dataset curation, model development, and evaluation on downstream tasks. By analyzing 14 datasets and 7 SotA models, we identify major gaps in their utility in the context of a real-world scenario. We demonstrate how each limitation impedes the widespread use of SotA models in enterprise settings, and present a set of research challenges that are motivated by these limitations. Lastly, we propose a research agenda that is aimed at driving the field towards higher impact in enterprise applications.",
}
|
The field of multimodal document understanding has produced a suite of models that have achieved stellar performance across several tasks, even coming close to human performance on certain benchmarks. Nevertheless, the application of these models to real-world enterprise datasets remains constrained by a number of limitations. In this position paper, we discuss these limitations in the context of three key aspects of research: dataset curation, model development, and evaluation on downstream tasks. By analyzing 14 datasets and 7 SotA models, we identify major gaps in their utility in the context of a real-world scenario. We demonstrate how each limitation impedes the widespread use of SotA models in enterprise settings, and present a set of research challenges that are motivated by these limitations. Lastly, we propose a research agenda that is aimed at driving the field towards higher impact in enterprise applications.
|
[
"Nourbakhsh, Armineh",
"Shah, Sameena",
"Rose, Carolyn"
] |
Towards a new research agenda for multimodal enterprise document understanding: What are we missing?
|
findings-acl.870
|
Poster
|
2406.13264v1
|
https://aclanthology.org/2024.findings-acl.871.bib
|
@inproceedings{abolghasemi-etal-2024-cause,
title = "{CAUSE}: Counterfactual Assessment of User Satisfaction Estimation in Task-Oriented Dialogue Systems",
author = "Abolghasemi, Amin and
Ren, Zhaochun and
Askari, Arian and
Aliannejadi, Mohammad and
Rijke, Maarten and
Verberne, Suzan",
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.871",
pages = "14623--14635",
abstract = "An important unexplored aspect in previous work on user satisfaction estimation for Task-Oriented Dialogue (TOD) systems is their evaluation in terms of robustness for the identification of user dissatisfaction: current benchmarks for user satisfaction estimation in TOD systems are highly skewed towards dialogues for which the user is satisfied. The effect of having a more balanced set of satisfaction labels on performance is unknown. However, balancing the data with more dissatisfactory dialogue samples requires further data collection and human annotation, which is costly and time-consuming. In this work, we leverage large language models (LLMs) and unlock their ability to generate satisfaction-aware counterfactual dialogues to augment the set of original dialogues of a test collection. We gather human annotations to ensure the reliability of the generated samples. We evaluate two open-source LLMs as user satisfaction estimators on our augmented collection against state-of-the-art fine-tuned models. Our experiments show that when used as few-shot user satisfaction estimators, open-source LLMs show higher robustness to the increase in the number of dissatisfaction labels in the test collection than the fine-tuned state-of-the-art models. Our results shed light on the need for data augmentation approaches for user satisfaction estimation in TOD systems. We release our aligned counterfactual dialogues, which are curated by human annotation, to facilitate further research on this topic.",
}
|
An important unexplored aspect in previous work on user satisfaction estimation for Task-Oriented Dialogue (TOD) systems is their evaluation in terms of robustness for the identification of user dissatisfaction: current benchmarks for user satisfaction estimation in TOD systems are highly skewed towards dialogues for which the user is satisfied. The effect of having a more balanced set of satisfaction labels on performance is unknown. However, balancing the data with more dissatisfactory dialogue samples requires further data collection and human annotation, which is costly and time-consuming. In this work, we leverage large language models (LLMs) and unlock their ability to generate satisfaction-aware counterfactual dialogues to augment the set of original dialogues of a test collection. We gather human annotations to ensure the reliability of the generated samples. We evaluate two open-source LLMs as user satisfaction estimators on our augmented collection against state-of-the-art fine-tuned models. Our experiments show that when used as few-shot user satisfaction estimators, open-source LLMs show higher robustness to the increase in the number of dissatisfaction labels in the test collection than the fine-tuned state-of-the-art models. Our results shed light on the need for data augmentation approaches for user satisfaction estimation in TOD systems. We release our aligned counterfactual dialogues, which are curated by human annotation, to facilitate further research on this topic.
|
[
"Abolghasemi, Amin",
"Ren, Zhaochun",
"Askari, Arian",
"Aliannejadi, Mohammad",
"Rijke, Maarten",
"Verberne, Suzan"
] |
{CAUSE}: Counterfactual Assessment of User Satisfaction Estimation in Task-Oriented Dialogue Systems
|
findings-acl.871
|
Poster
|
2403.19056v1
|
https://aclanthology.org/2024.findings-acl.872.bib
|
@inproceedings{gabburo-etal-2024-measuring,
title = "Measuring Retrieval Complexity in Question Answering Systems",
author = "Gabburo, Matteo and
Jedema, Nicolaas and
Garg, Siddhant and
Ribeiro, Leonardo and
Moschitti, Alessandro",
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.872",
pages = "14636--14650",
abstract = "In this paper, we investigate which questions are challenging for retrieval-based Question Answering (QA). We (i) propose retrieval complexity (RC), a novel metric conditioned on the completeness of retrieved documents, which measures the difficulty of answering questions, and (ii) propose an unsupervised pipeline to measure RC given an arbitrary retrieval system.Our proposed pipeline measures RC more accurately than alternative estimators, including LLMs, on six challenging QA benchmarks. Further investigation reveals that RC scores strongly correlate with both QA performance and expert judgment across five of the six studied benchmarks, indicating that RC is an effective measure of question difficulty.Subsequent categorization of high-RC questions shows that they span a broad set of question shapes, including multi-hop, compositional, and temporal QA, indicating that RC scores can categorize a new subset of complex questions. Our system can also have a major impact on retrieval-based systems by helping to identify more challenging questions on existing datasets.",
}
|
In this paper, we investigate which questions are challenging for retrieval-based Question Answering (QA). We (i) propose retrieval complexity (RC), a novel metric conditioned on the completeness of retrieved documents, which measures the difficulty of answering questions, and (ii) propose an unsupervised pipeline to measure RC given an arbitrary retrieval system.Our proposed pipeline measures RC more accurately than alternative estimators, including LLMs, on six challenging QA benchmarks. Further investigation reveals that RC scores strongly correlate with both QA performance and expert judgment across five of the six studied benchmarks, indicating that RC is an effective measure of question difficulty.Subsequent categorization of high-RC questions shows that they span a broad set of question shapes, including multi-hop, compositional, and temporal QA, indicating that RC scores can categorize a new subset of complex questions. Our system can also have a major impact on retrieval-based systems by helping to identify more challenging questions on existing datasets.
|
[
"Gabburo, Matteo",
"Jedema, Nicolaas",
"Garg, Siddhant",
"Ribeiro, Leonardo",
"Moschitti, Aless",
"ro"
] |
Measuring Retrieval Complexity in Question Answering Systems
|
findings-acl.872
|
Poster
|
2406.03592v1
|
https://aclanthology.org/2024.findings-acl.873.bib
|
@inproceedings{song-etal-2024-combining,
title = "Combining Hierachical {VAE}s with {LLM}s for clinically meaningful timeline summarisation in social media",
author = "Song, Jiayu and
Chim, Jenny and
Tsakalidis, Adam and
Ive, Julia and
Atzil-Slonim, Dana and
Liakata, Maria",
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.873",
pages = "14651--14672",
abstract = "We introduce a hybrid abstractive summarisation approach combining hierarchical VAEs with LLMs to produce clinically meaningful summaries from social media user timelines, appropriate for mental health monitoring. The summaries combine two different narrative points of view: (a) clinical insights in third person, generated by feeding into an LLM clinical expert-guided prompts, and importantly, (b) a temporally sensitive abstractive summary of the user{'}s timeline in first person, generated by a novel hierarchical variational autoencoder, TH-VAE. We assess the generated summaries via automatic evaluation against expert summaries and via human evaluation with clinical experts, showing that timeline summarisation by TH-VAE results in more factual and logically coherent summaries rich in clinical utility and superior to LLM-only approaches in capturing changes over time.",
}
|
We introduce a hybrid abstractive summarisation approach combining hierarchical VAEs with LLMs to produce clinically meaningful summaries from social media user timelines, appropriate for mental health monitoring. The summaries combine two different narrative points of view: (a) clinical insights in third person, generated by feeding into an LLM clinical expert-guided prompts, and importantly, (b) a temporally sensitive abstractive summary of the user{'}s timeline in first person, generated by a novel hierarchical variational autoencoder, TH-VAE. We assess the generated summaries via automatic evaluation against expert summaries and via human evaluation with clinical experts, showing that timeline summarisation by TH-VAE results in more factual and logically coherent summaries rich in clinical utility and superior to LLM-only approaches in capturing changes over time.
|
[
"Song, Jiayu",
"Chim, Jenny",
"Tsakalidis, Adam",
"Ive, Julia",
"Atzil-Slonim, Dana",
"Liakata, Maria"
] |
Combining Hierachical {VAE}s with {LLM}s for clinically meaningful timeline summarisation in social media
|
findings-acl.873
|
Poster
|
2401.16240v2
|
https://aclanthology.org/2024.findings-acl.874.bib
|
@inproceedings{tai-etal-2024-pixar,
title = "{PIXAR}: Auto-Regressive Language Modeling in Pixel Space",
author = "Tai, Yintao and
Liao, Xiyang and
Suglia, Alessandro and
Vergari, Antonio",
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.874",
pages = "14673--14695",
abstract = "Recent work showed the possibility of building open-vocabulary large language models (LLMs) that directly operate on pixel representations. These models are implemented as autoencoders that reconstruct masked patches of rendered text.However, these pixel-based LLMs are limited to discriminative tasks (e.g., classification) and, similar to BERT, cannot be used to generate text.Therefore, they cannot be used for generative tasks such as free-form question answering. In this work, we introduce PIXAR, the first pixel-based autoregressive LLM that performs text generation. Consisting of only a decoder, PIXAR can perform free-form generative tasks while keeping the number of parameters on par with previous encoder-decoder models.Furthermore, we highlight the challenges of generating text as non-noisy images and show this is due to using a maximum likelihood objective. To overcome this problem, we propose an adversarial pretraining stage that improves the readability and accuracy of PIXAR by 8.1 on LAMBADA and 8.5 on bAbI{---} making it comparable to GPT-2 on text generation tasks.This paves the way to build open-vocabulary LLMs that operate on perceptual input only and calls into question the necessity of the usual symbolic input representation, i.e., text as (sub)tokens.",
}
|
Recent work showed the possibility of building open-vocabulary large language models (LLMs) that directly operate on pixel representations. These models are implemented as autoencoders that reconstruct masked patches of rendered text.However, these pixel-based LLMs are limited to discriminative tasks (e.g., classification) and, similar to BERT, cannot be used to generate text.Therefore, they cannot be used for generative tasks such as free-form question answering. In this work, we introduce PIXAR, the first pixel-based autoregressive LLM that performs text generation. Consisting of only a decoder, PIXAR can perform free-form generative tasks while keeping the number of parameters on par with previous encoder-decoder models.Furthermore, we highlight the challenges of generating text as non-noisy images and show this is due to using a maximum likelihood objective. To overcome this problem, we propose an adversarial pretraining stage that improves the readability and accuracy of PIXAR by 8.1 on LAMBADA and 8.5 on bAbI{---} making it comparable to GPT-2 on text generation tasks.This paves the way to build open-vocabulary LLMs that operate on perceptual input only and calls into question the necessity of the usual symbolic input representation, i.e., text as (sub)tokens.
|
[
"Tai, Yintao",
"Liao, Xiyang",
"Suglia, Aless",
"ro",
"Vergari, Antonio"
] |
{PIXAR}: Auto-Regressive Language Modeling in Pixel Space
|
findings-acl.874
|
Poster
|
2305.01040v1
|
https://aclanthology.org/2024.findings-acl.875.bib
|
@inproceedings{ma-etal-2024-sparsity,
title = "Sparsity-Accelerated Training for Large Language Models",
author = "Ma, Da and
Chen, Lu and
Wang, Pengyu and
Xu, Hongshen and
Li, Hanqi and
Sun, Liangtai and
Zhu, Su and
Fan, Shuai and
Yu, Kai",
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.875",
pages = "14696--14707",
abstract = "Large language models (LLMs) have demonstrated proficiency across various natural language processing (NLP) tasks but often require additional training, such as continual pre-training and supervised fine-tuning. However, the costs associated with this, primarily due to their large parameter count, remain high. This paper proposes leveraging \textit{sparsity} in pre-trained LLMs to expedite this training process. By observing sparsity in activated neurons during forward iterations, we identify the potential for computational speed-ups by excluding inactive neurons. We address associated challenges by extending existing neuron importance evaluation metrics and introducing a ladder omission rate scheduler. Our experiments on Llama-2 demonstrate that Sparsity-Accelerated Training (SAT) achieves comparable or superior performance to standard training while significantly accelerating the process. Specifically, SAT achieves a 45{\%} throughput improvement in continual pre-training and saves 38{\%} training time in supervised fine-tuning. It offers a simple, hardware-agnostic, and easily deployable framework for additional LLM training.",
}
|
Large language models (LLMs) have demonstrated proficiency across various natural language processing (NLP) tasks but often require additional training, such as continual pre-training and supervised fine-tuning. However, the costs associated with this, primarily due to their large parameter count, remain high. This paper proposes leveraging \textit{sparsity} in pre-trained LLMs to expedite this training process. By observing sparsity in activated neurons during forward iterations, we identify the potential for computational speed-ups by excluding inactive neurons. We address associated challenges by extending existing neuron importance evaluation metrics and introducing a ladder omission rate scheduler. Our experiments on Llama-2 demonstrate that Sparsity-Accelerated Training (SAT) achieves comparable or superior performance to standard training while significantly accelerating the process. Specifically, SAT achieves a 45{\%} throughput improvement in continual pre-training and saves 38{\%} training time in supervised fine-tuning. It offers a simple, hardware-agnostic, and easily deployable framework for additional LLM training.
|
[
"Ma, Da",
"Chen, Lu",
"Wang, Pengyu",
"Xu, Hongshen",
"Li, Hanqi",
"Sun, Liangtai",
"Zhu, Su",
"Fan, Shuai",
"Yu, Kai"
] |
Sparsity-Accelerated Training for Large Language Models
|
findings-acl.875
|
Poster
|
2202.03371v1
|
https://aclanthology.org/2024.findings-acl.876.bib
|
@inproceedings{xu-etal-2024-preemptive,
title = "Preemptive Answer {``}Attacks{''} on Chain-of-Thought Reasoning",
author = "Xu, Rongwu and
Qi, Zehan and
Xu, 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.876",
pages = "14708--14726",
abstract = "Large language models (LLMs) showcase impressive reasoning capabilities when coupled with Chain-of-Thought (CoT) prompting. However, the robustness of this approach warrants further investigation. In this paper, we introduce a novel scenario termed preemptive answers, where the LLM obtains an answer before engaging in reasoning. This situation can arise inadvertently or induced by malicious users by prompt injection attacks. Experiments reveal that preemptive answers significantly impair the model{'}s reasoning capability across various CoT methods and a broad spectrum of datasets. To bolster the robustness of reasoning, we propose two measures aimed at mitigating this issue to some extent.",
}
|
Large language models (LLMs) showcase impressive reasoning capabilities when coupled with Chain-of-Thought (CoT) prompting. However, the robustness of this approach warrants further investigation. In this paper, we introduce a novel scenario termed preemptive answers, where the LLM obtains an answer before engaging in reasoning. This situation can arise inadvertently or induced by malicious users by prompt injection attacks. Experiments reveal that preemptive answers significantly impair the model{'}s reasoning capability across various CoT methods and a broad spectrum of datasets. To bolster the robustness of reasoning, we propose two measures aimed at mitigating this issue to some extent.
|
[
"Xu, Rongwu",
"Qi, Zehan",
"Xu, Wei"
] |
Preemptive Answer {``}Attacks{''} on Chain-of-Thought Reasoning
|
findings-acl.876
|
Poster
|
2405.20902v1
|
https://aclanthology.org/2024.findings-acl.877.bib
|
@inproceedings{jumelet-etal-2024-language,
title = "Do Language Models Exhibit Human-like Structural Priming Effects?",
author = "Jumelet, Jaap and
Zuidema, Willem and
Sinclair, Arabella",
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.877",
pages = "14727--14742",
abstract = "We explore which linguistic factors{---}at the sentence and token level{---}play an important role in influencing language model predictions, and investigate whether these are reflective of results found in humans and human corpora (Gries and Kootstra, 2017). We make use of the structural priming paradigm{---}where recent exposure to a structure facilitates processing of the same structure{---}to investigate where priming effects manifest, and what factors predict them. We find these effects can be explained via the inverse frequency effect found in human priming, where rarer elements within a prime increase priming effects, as well as lexical dependence between prime and target. Our results provide an important piece in the puzzle of understanding how properties within their context affect structural prediction in language models.",
}
|
We explore which linguistic factors{---}at the sentence and token level{---}play an important role in influencing language model predictions, and investigate whether these are reflective of results found in humans and human corpora (Gries and Kootstra, 2017). We make use of the structural priming paradigm{---}where recent exposure to a structure facilitates processing of the same structure{---}to investigate where priming effects manifest, and what factors predict them. We find these effects can be explained via the inverse frequency effect found in human priming, where rarer elements within a prime increase priming effects, as well as lexical dependence between prime and target. Our results provide an important piece in the puzzle of understanding how properties within their context affect structural prediction in language models.
|
[
"Jumelet, Jaap",
"Zuidema, Willem",
"Sinclair, Arabella"
] |
Do Language Models Exhibit Human-like Structural Priming Effects?
|
findings-acl.877
|
Poster
|
2406.04847v1
|
https://aclanthology.org/2024.findings-acl.878.bib
|
@inproceedings{wang-etal-2024-rolellm,
title = "{R}ole{LLM}: Benchmarking, Eliciting, and Enhancing Role-Playing Abilities of Large Language Models",
author = "Wang, Noah and
Peng, Z.y. and
Que, Haoran and
Liu, Jiaheng and
Zhou, Wangchunshu and
Wu, Yuhan and
Guo, Hongcheng and
Gan, Ruitong and
Ni, Zehao and
Yang, Jian and
Zhang, Man and
Zhang, Zhaoxiang and
Ouyang, Wanli and
Xu, Ke and
Huang, Wenhao and
Fu, Jie and
Peng, Junran",
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.878",
pages = "14743--14777",
abstract = "The advent of Large Language Models (LLMs) has paved the way for complex tasks such as role-playing, which enhances user interactions by enabling models to imitate various characters. However, the closed-source nature of state-of-the-art LLMs and their general-purpose training limit role-playing optimization. In this paper, we introduce RoleLLM, a framework to benchmark, elicit, and enhance role-playing abilities in LLMs. RoleLLM comprises four stages: (1) Role Profile Construction for 100 roles; (2) Context-Based Instruction Generation (Context-Instruct) for role-specific knowledge extraction; (3) Role Prompting using GPT (RoleGPT) for speaking style imitation; and (4) Role-Conditioned Instruction Tuning (RoCIT) for fine-tuning open-source models along with role customization. By Context-Instruct and RoleGPT, we create RoleBench, the first systematic and fine-grained character-level benchmark dataset for role-playing with 168,093 samples. Moreover, RoCIT on RoleBench yields RoleLLaMA (English) and RoleGLM (Chinese), significantly enhancing role-playing abilities and even achieving comparable results with RoleGPT (using GPT-4).",
}
|
The advent of Large Language Models (LLMs) has paved the way for complex tasks such as role-playing, which enhances user interactions by enabling models to imitate various characters. However, the closed-source nature of state-of-the-art LLMs and their general-purpose training limit role-playing optimization. In this paper, we introduce RoleLLM, a framework to benchmark, elicit, and enhance role-playing abilities in LLMs. RoleLLM comprises four stages: (1) Role Profile Construction for 100 roles; (2) Context-Based Instruction Generation (Context-Instruct) for role-specific knowledge extraction; (3) Role Prompting using GPT (RoleGPT) for speaking style imitation; and (4) Role-Conditioned Instruction Tuning (RoCIT) for fine-tuning open-source models along with role customization. By Context-Instruct and RoleGPT, we create RoleBench, the first systematic and fine-grained character-level benchmark dataset for role-playing with 168,093 samples. Moreover, RoCIT on RoleBench yields RoleLLaMA (English) and RoleGLM (Chinese), significantly enhancing role-playing abilities and even achieving comparable results with RoleGPT (using GPT-4).
|
[
"Wang, Noah",
"Peng, Z.y.",
"Que, Haoran",
"Liu, Jiaheng",
"Zhou, Wangchunshu",
"Wu, Yuhan",
"Guo, Hongcheng",
"Gan, Ruitong",
"Ni, Zehao",
"Yang, Jian",
"Zhang, Man",
"Zhang, Zhaoxiang",
"Ouyang, Wanli",
"Xu, Ke",
"Huang, Wenhao",
"Fu, Jie",
"Peng, Junran"
] |
{R}ole{LLM}: Benchmarking, Eliciting, and Enhancing Role-Playing Abilities of Large Language Models
|
findings-acl.878
|
Poster
|
2406.00627v1
|
https://aclanthology.org/2024.findings-acl.879.bib
|
@inproceedings{jia-etal-2024-langsuit,
title = "{L}ang{S}uit{\mbox{$\cdot$}}{E}: Planning, Controlling and Interacting with Large Language Models in Embodied Text Environments",
author = "Jia, Zixia and
Wang, Mengmeng and
Tong, Baichen and
Zhu, Song-Chun and
Zheng, Zilong",
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.879",
pages = "14778--14814",
abstract = "Recent advances in Large Language Models (LLMs) have shown inspiring achievements in constructing autonomous agents that rely onlanguage descriptions as inputs. However, it remains unclear how well LLMs can function as few-shot or zero-shot embodied agents in dynamic interactive environments. To address this gap, we introduce LangSuit{\mbox{$\cdot$}}E, a versatile and simulation-free testbed featuring 6 representative embodied tasks in textual embodied worlds. Compared with previous LLM-based testbeds, LangSuit{\mbox{$\cdot$}}E (i) offers adaptability to diverse environments without multiple simulation engines, (ii) evaluates agents{'} capacity to develop {``}internalized world knowledge{''} with embodied observations, and (iii) allows easy customization of communication and action strategies. To address the embodiment challenge, we devise a novel chain-of-thought (CoT) schema, EmMem, which summarizes embodied states w.r.t. history information. Comprehensive benchmark results illustrate challenges and insights of embodied planning. LangSuit{\mbox{$\cdot$}}E represents a significant step toward building embodied generalists in the context of language models.",
}
|
Recent advances in Large Language Models (LLMs) have shown inspiring achievements in constructing autonomous agents that rely onlanguage descriptions as inputs. However, it remains unclear how well LLMs can function as few-shot or zero-shot embodied agents in dynamic interactive environments. To address this gap, we introduce LangSuit{\mbox{$\cdot$}}E, a versatile and simulation-free testbed featuring 6 representative embodied tasks in textual embodied worlds. Compared with previous LLM-based testbeds, LangSuit{\mbox{$\cdot$}}E (i) offers adaptability to diverse environments without multiple simulation engines, (ii) evaluates agents{'} capacity to develop {``}internalized world knowledge{''} with embodied observations, and (iii) allows easy customization of communication and action strategies. To address the embodiment challenge, we devise a novel chain-of-thought (CoT) schema, EmMem, which summarizes embodied states w.r.t. history information. Comprehensive benchmark results illustrate challenges and insights of embodied planning. LangSuit{\mbox{$\cdot$}}E represents a significant step toward building embodied generalists in the context of language models.
|
[
"Jia, Zixia",
"Wang, Mengmeng",
"Tong, Baichen",
"Zhu, Song-Chun",
"Zheng, Zilong"
] |
{L}ang{S}uit{\mbox{$\cdot$}}{E}: Planning, Controlling and Interacting with Large Language Models in Embodied Text Environments
|
findings-acl.879
|
Poster
|
2406.16294v1
|
https://aclanthology.org/2024.findings-acl.880.bib
|
@inproceedings{soubki-etal-2024-views,
title = "Views Are My Own, but Also Yours: Benchmarking Theory of Mind Using Common Ground",
author = "Soubki, Adil and
Murzaku, John and
Yousefi Jordehi, Arash and
Zeng, Peter and
Markowska, Magdalena and
Mirroshandel, Seyed Abolghasem and
Rambow, Owen",
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.880",
pages = "14815--14823",
abstract = "Evaluating the theory of mind (ToM) capabilities of language models (LMs) has recently received a great deal of attention. However, many existing benchmarks rely on synthetic data, which risks misaligning the resulting experiments with human behavior. We introduce the first ToM dataset based on naturally occurring spoken dialogs, Common-ToM, and show that LMs struggle to demonstrate ToM. We then show that integrating a simple, explicit representation of beliefs improves LM performance on Common-ToM.",
}
|
Evaluating the theory of mind (ToM) capabilities of language models (LMs) has recently received a great deal of attention. However, many existing benchmarks rely on synthetic data, which risks misaligning the resulting experiments with human behavior. We introduce the first ToM dataset based on naturally occurring spoken dialogs, Common-ToM, and show that LMs struggle to demonstrate ToM. We then show that integrating a simple, explicit representation of beliefs improves LM performance on Common-ToM.
|
[
"Soubki, Adil",
"Murzaku, John",
"Yousefi Jordehi, Arash",
"Zeng, Peter",
"Markowska, Magdalena",
"Mirrosh",
"el, Seyed Abolghasem",
"Rambow, Owen"
] |
Views Are My Own, but Also Yours: Benchmarking Theory of Mind Using Common Ground
|
findings-acl.880
|
Poster
|
2403.02451v2
|
https://aclanthology.org/2024.findings-acl.881.bib
|
@inproceedings{aggarwal-etal-2024-maple,
title = "{MAPLE}: Multilingual Evaluation of Parameter Efficient Finetuning of Large Language Models",
author = "Aggarwal, Divyanshu and
Sathe, Ashutosh and
Watts, Ishaan and
Sitaram, Sunayana",
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.881",
pages = "14824--14867",
abstract = "Parameter efficient finetuning has emerged as a viable solution for improving the performance of Large Language Models without requiring massive resources and compute. Prior work on multilingual evaluation has shown that there is a large gap between the performance of LLMs on English and other languages. Further, there is also a large gap between the performance of smaller open-source models and larger LLMs. Finetuning can be an effective way to bridge this gap and make language models more equitable. In this work, we finetune the Llama-2 and Mistral models on two synthetic multilingual instruction tuning datasets to determine its effect on model performance on six downstream tasks covering forty one languages in all. Additionally, we experiment with various parameters, such as rank for low-rank adaptation and values of quantisation to determine their effects on downstream performance and find that higher rank and higher quantisation values benefit low-resource languages. We find that parameter efficient finetuning of smaller open-source models sometimes bridges the gap between the performance of these models and the larger ones, however, English performance can take a hit. We also find that finetuning sometimes improves performance on low-resource languages, while degrading performance on high-resource languages.",
}
|
Parameter efficient finetuning has emerged as a viable solution for improving the performance of Large Language Models without requiring massive resources and compute. Prior work on multilingual evaluation has shown that there is a large gap between the performance of LLMs on English and other languages. Further, there is also a large gap between the performance of smaller open-source models and larger LLMs. Finetuning can be an effective way to bridge this gap and make language models more equitable. In this work, we finetune the Llama-2 and Mistral models on two synthetic multilingual instruction tuning datasets to determine its effect on model performance on six downstream tasks covering forty one languages in all. Additionally, we experiment with various parameters, such as rank for low-rank adaptation and values of quantisation to determine their effects on downstream performance and find that higher rank and higher quantisation values benefit low-resource languages. We find that parameter efficient finetuning of smaller open-source models sometimes bridges the gap between the performance of these models and the larger ones, however, English performance can take a hit. We also find that finetuning sometimes improves performance on low-resource languages, while degrading performance on high-resource languages.
|
[
"Aggarwal, Divyanshu",
"Sathe, Ashutosh",
"Watts, Ishaan",
"Sitaram, Sunayana"
] |
{MAPLE}: Multilingual Evaluation of Parameter Efficient Finetuning of Large Language Models
|
findings-acl.881
|
Poster
|
2402.17193v1
|
https://aclanthology.org/2024.findings-acl.882.bib
|
@inproceedings{cheng-etal-2024-moe,
title = "{M}o{E}-{SLU}: Towards {ASR}-Robust Spoken Language Understanding via Mixture-of-Experts",
author = "Cheng, Xuxin and
Zhu, Zhihong and
Zhuang, Xianwei and
Chen, Zhanpeng and
Huang, Zhiqi 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.882",
pages = "14868--14879",
abstract = "As a crucial task in the task-oriented dialogue systems, spoken language understanding (SLU) has garnered increasing attention. However, errors from automatic speech recognition (ASR) often hinder the performance of understanding. To tackle this problem, we propose MoE-SLU, an ASR-Robust SLU framework based on the mixture-of-experts technique. Specifically, we first introduce three strategies to generate additional transcripts from clean transcripts. Then, we employ the mixture-of-experts technique to weigh the representations of the generated transcripts, ASR transcripts, and the corresponding clean manual transcripts. Additionally, we also regularize the weighted average of predictions and the predictions of ASR transcripts by minimizing the Jensen-Shannon Divergence (JSD) between these two output distributions. Experiment results on three benchmark SLU datasets demonstrate that our MoE-SLU achieves state-of-the-art performance. Further model analysis also verifies the superiority of our method.",
}
|
As a crucial task in the task-oriented dialogue systems, spoken language understanding (SLU) has garnered increasing attention. However, errors from automatic speech recognition (ASR) often hinder the performance of understanding. To tackle this problem, we propose MoE-SLU, an ASR-Robust SLU framework based on the mixture-of-experts technique. Specifically, we first introduce three strategies to generate additional transcripts from clean transcripts. Then, we employ the mixture-of-experts technique to weigh the representations of the generated transcripts, ASR transcripts, and the corresponding clean manual transcripts. Additionally, we also regularize the weighted average of predictions and the predictions of ASR transcripts by minimizing the Jensen-Shannon Divergence (JSD) between these two output distributions. Experiment results on three benchmark SLU datasets demonstrate that our MoE-SLU achieves state-of-the-art performance. Further model analysis also verifies the superiority of our method.
|
[
"Cheng, Xuxin",
"Zhu, Zhihong",
"Zhuang, Xianwei",
"Chen, Zhanpeng",
"Huang, Zhiqi",
"Zou, Yuexian"
] |
{M}o{E}-{SLU}: Towards {ASR}-Robust Spoken Language Understanding via Mixture-of-Experts
|
findings-acl.882
|
Poster
|
2102.02270v2
|
https://aclanthology.org/2024.findings-acl.883.bib
|
@inproceedings{mueller-etal-2024-multi,
title = "Multi-Task Transfer Matters During Instruction-Tuning",
author = "Mueller, David and
Dredze, Mark and
Andrews, Nicholas",
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.883",
pages = "14880--14891",
abstract = "Instruction-tuning trains a language model on hundreds of tasks jointly to improve a model{'}s ability to learn in-context;however, the mechanisms that drive in-context learning are poorly understood and, as a result, the role of instruction-tuning on in-context generalization is poorly understood as well.In this work, we study the impact of instruction-tuning on multi-task transfer: how well a model{'}s parameters adapt to an unseen task via fine-tuning.We find that instruction-tuning negatively impacts a model{'}s transfer to unseen tasks, and that model transfer and in-context generalization are highly correlated, suggesting that this catastrophic forgetting may impact in-context learning.We study methods to improve model transfer, finding that multi-task training{---}how well the training tasks are optimized{---}can significantly impact ICL generalization; additionally, we find that continual training on unsupervised pre-training data can mitigate forgetting and improve ICL generalization as well.Finally, we demonstrate that, early into training, the impact of instruction-tuning on model transfer to tasks impacts in-context generalization on that task.Overall, we provide significant evidence that multi-task transfer is deeply connected to a model{'}s ability to learn a task in-context.",
}
|
Instruction-tuning trains a language model on hundreds of tasks jointly to improve a model{'}s ability to learn in-context;however, the mechanisms that drive in-context learning are poorly understood and, as a result, the role of instruction-tuning on in-context generalization is poorly understood as well.In this work, we study the impact of instruction-tuning on multi-task transfer: how well a model{'}s parameters adapt to an unseen task via fine-tuning.We find that instruction-tuning negatively impacts a model{'}s transfer to unseen tasks, and that model transfer and in-context generalization are highly correlated, suggesting that this catastrophic forgetting may impact in-context learning.We study methods to improve model transfer, finding that multi-task training{---}how well the training tasks are optimized{---}can significantly impact ICL generalization; additionally, we find that continual training on unsupervised pre-training data can mitigate forgetting and improve ICL generalization as well.Finally, we demonstrate that, early into training, the impact of instruction-tuning on model transfer to tasks impacts in-context generalization on that task.Overall, we provide significant evidence that multi-task transfer is deeply connected to a model{'}s ability to learn a task in-context.
|
[
"Mueller, David",
"Dredze, Mark",
"Andrews, Nicholas"
] |
Multi-Task Transfer Matters During Instruction-Tuning
|
findings-acl.883
|
Poster
|
1312.4092v1
|
https://aclanthology.org/2024.findings-acl.884.bib
|
@inproceedings{guo-etal-2024-makes,
title = "What Makes a Good Order of Examples in In-Context Learning",
author = "Guo, Qi and
Wang, Leiyu and
Wang, Yidong and
Ye, Wei and
Zhang, Shikun",
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.884",
pages = "14892--14904",
abstract = "Although large language models (LLMs) have demonstrated impressive few-shot learning capabilities via in-context learning (ICL), ICL performance is known to be highly sensitive to the order of examples provided. To identify appropriate orders, recent studies propose heuristic methods to evaluate order performance using a set of unlabeled data. However, the requirement of in-domain data limits their utility in real-world scenarios where additional annotated data is challenging to acquire. Additionally, these dataset-based approaches are prone to being sub-optimal for a lack of consideration for individual differences. To address the problems, we first analyze the properties of performant example orders at both corpus level and instance level. Based on the analysis we propose **DEmO** to adaptively identify performant example order for each instance without extra data. DEmO works by filtering out a subset of orders featuring label fairness, then selecting the most influential order for each test instance. The employment of a content-free metric makes DEmO independent of in-domain data. Extensive experiments indicate the superiority of DEmO over a wide range of strong baselines. Further analysis validates the generalizability across various settings.",
}
|
Although large language models (LLMs) have demonstrated impressive few-shot learning capabilities via in-context learning (ICL), ICL performance is known to be highly sensitive to the order of examples provided. To identify appropriate orders, recent studies propose heuristic methods to evaluate order performance using a set of unlabeled data. However, the requirement of in-domain data limits their utility in real-world scenarios where additional annotated data is challenging to acquire. Additionally, these dataset-based approaches are prone to being sub-optimal for a lack of consideration for individual differences. To address the problems, we first analyze the properties of performant example orders at both corpus level and instance level. Based on the analysis we propose **DEmO** to adaptively identify performant example order for each instance without extra data. DEmO works by filtering out a subset of orders featuring label fairness, then selecting the most influential order for each test instance. The employment of a content-free metric makes DEmO independent of in-domain data. Extensive experiments indicate the superiority of DEmO over a wide range of strong baselines. Further analysis validates the generalizability across various settings.
|
[
"Guo, Qi",
"Wang, Leiyu",
"Wang, Yidong",
"Ye, Wei",
"Zhang, Shikun"
] |
What Makes a Good Order of Examples in In-Context Learning
|
findings-acl.884
|
Poster
|
2305.09731v1
|
https://aclanthology.org/2024.findings-acl.885.bib
|
@inproceedings{gong-etal-2024-bloomvqa,
title = "{B}loom{VQA}: Assessing Hierarchical Multi-modal Comprehension",
author = "Gong, Yunye and
Shrestha, Robik and
Claypoole, Jared and
Cogswell, Michael and
Ray, Arijit and
Kanan, Christopher and
Divakaran, Ajay",
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.885",
pages = "14905--14918",
abstract = "We propose a novel VQA dataset, BloomVQA, to facilitate comprehensive evaluation of large vision-language models on comprehension tasks. Unlike current benchmarks that often focus on fact-based memorization and simple reasoning tasks without theoretical grounding, we collect multiple-choice samples based on picture stories that reflect different levels of comprehension, as laid out in Bloom{'}s Taxonomy, a classic framework for learning assessment widely adopted in education research. Our data maps to a novel hierarchical graph representation which enables automatic data augmentation and novel measures characterizing model consistency. We perform graded evaluation and reliability analysis on recent multi-modal models. In comparison to low-level tasks, we observe decreased performance on tasks requiring advanced comprehension and cognitive skills with up to 38.0{\%} drop in VQA accuracy. In comparison to earlier models, GPT-4V demonstrates improved accuracy over all comprehension levels and also shows a tendency of bypassing visual inputs especially for higher-level tasks. Current models also show consistency patterns misaligned with human comprehension in various scenarios, demonstrating the need for improvement based on theoretically-grounded criteria. The dataset can be accessed at https://huggingface.co/datasets/ygong/BloomVQA.",
}
|
We propose a novel VQA dataset, BloomVQA, to facilitate comprehensive evaluation of large vision-language models on comprehension tasks. Unlike current benchmarks that often focus on fact-based memorization and simple reasoning tasks without theoretical grounding, we collect multiple-choice samples based on picture stories that reflect different levels of comprehension, as laid out in Bloom{'}s Taxonomy, a classic framework for learning assessment widely adopted in education research. Our data maps to a novel hierarchical graph representation which enables automatic data augmentation and novel measures characterizing model consistency. We perform graded evaluation and reliability analysis on recent multi-modal models. In comparison to low-level tasks, we observe decreased performance on tasks requiring advanced comprehension and cognitive skills with up to 38.0{\%} drop in VQA accuracy. In comparison to earlier models, GPT-4V demonstrates improved accuracy over all comprehension levels and also shows a tendency of bypassing visual inputs especially for higher-level tasks. Current models also show consistency patterns misaligned with human comprehension in various scenarios, demonstrating the need for improvement based on theoretically-grounded criteria. The dataset can be accessed at https://huggingface.co/datasets/ygong/BloomVQA.
|
[
"Gong, Yunye",
"Shrestha, Robik",
"Claypoole, Jared",
"Cogswell, Michael",
"Ray, Arijit",
"Kanan, Christopher",
"Divakaran, Ajay"
] |
{B}loom{VQA}: Assessing Hierarchical Multi-modal Comprehension
|
findings-acl.885
|
Poster
|
2302.11891v2
|
https://aclanthology.org/2024.findings-acl.886.bib
|
@inproceedings{li-etal-2024-attributionbench,
title = "{A}ttribution{B}ench: How Hard is Automatic Attribution Evaluation?",
author = "Li, Yifei and
Yue, Xiang and
Liao, Zeyi and
Sun, Huan",
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.886",
pages = "14919--14935",
abstract = "Modern generative search engines enhance the reliability of large language model (LLM) responses by providing cited evidence. However, evaluating the answer{'}s attribution, i.e., whether every claim within the generated responses is fully supported by its cited evidence, remains an open problem. This verification, traditionally dependent on costly human evaluation, underscores the urgent need for automatic attribution evaluation methods. To bridge the gap in the absence of standardized benchmarks for these methods, we present AttributionBench, a comprehensive benchmark compiled from various existing attribution datasets. Our extensive experiments on AttributionBench reveal the challenges of automatic attribution evaluation, even for state-of-the-art LLMs. Specifically, our findings show that even a fine-tuned GPT-3.5 only achieves around 80{\%} macro-F1 under a binary classification formulation. A detailed analysis of more than 300 error cases indicates that a majority of failures stem from the model{'}s inability to process nuanced information, and the discrepancy between the information the model has access to and that human annotators do.",
}
|
Modern generative search engines enhance the reliability of large language model (LLM) responses by providing cited evidence. However, evaluating the answer{'}s attribution, i.e., whether every claim within the generated responses is fully supported by its cited evidence, remains an open problem. This verification, traditionally dependent on costly human evaluation, underscores the urgent need for automatic attribution evaluation methods. To bridge the gap in the absence of standardized benchmarks for these methods, we present AttributionBench, a comprehensive benchmark compiled from various existing attribution datasets. Our extensive experiments on AttributionBench reveal the challenges of automatic attribution evaluation, even for state-of-the-art LLMs. Specifically, our findings show that even a fine-tuned GPT-3.5 only achieves around 80{\%} macro-F1 under a binary classification formulation. A detailed analysis of more than 300 error cases indicates that a majority of failures stem from the model{'}s inability to process nuanced information, and the discrepancy between the information the model has access to and that human annotators do.
|
[
"Li, Yifei",
"Yue, Xiang",
"Liao, Zeyi",
"Sun, Huan"
] |
{A}ttribution{B}ench: How Hard is Automatic Attribution Evaluation?
|
findings-acl.886
|
Poster
|
2002.00460v1
|
https://aclanthology.org/2024.findings-acl.887.bib
|
@inproceedings{lovelace-etal-2024-diffusion,
title = "Diffusion Guided Language Modeling",
author = "Lovelace, Justin and
Kishore, Varsha and
Chen, Yiwei and
Weinberger, Kilian",
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.887",
pages = "14936--14952",
abstract = "Current language models demonstrate remarkable proficiency in text generation. However, for many applications it is desirable to control attributes, such as sentiment, or toxicity, of the generated language{---}ideally tailored towards each specific use case and target audience. For auto-regressive language models, existing guidance methods are prone to decoding errors that cascade during generation and degrade performance. In contrast, text diffusion models can easily be guided with, for example, a simple linear sentiment classifier{---}however they do suffer from significantly higher perplexity than auto-regressive alternatives. In this paper we use a guided diffusion model to produce a latent proposal that steers an auto-regressive language model to generate text with desired properties. Our model inherits the unmatched fluency of the auto-regressive approach and the plug-and-play flexibility of diffusion. We show that it outperforms previous plug-and-play guidance methods across a wide range of benchmark data sets. Further, controlling a new attribute in our framework is reduced to training a single logistic regression classifier.",
}
|
Current language models demonstrate remarkable proficiency in text generation. However, for many applications it is desirable to control attributes, such as sentiment, or toxicity, of the generated language{---}ideally tailored towards each specific use case and target audience. For auto-regressive language models, existing guidance methods are prone to decoding errors that cascade during generation and degrade performance. In contrast, text diffusion models can easily be guided with, for example, a simple linear sentiment classifier{---}however they do suffer from significantly higher perplexity than auto-regressive alternatives. In this paper we use a guided diffusion model to produce a latent proposal that steers an auto-regressive language model to generate text with desired properties. Our model inherits the unmatched fluency of the auto-regressive approach and the plug-and-play flexibility of diffusion. We show that it outperforms previous plug-and-play guidance methods across a wide range of benchmark data sets. Further, controlling a new attribute in our framework is reduced to training a single logistic regression classifier.
|
[
"Lovelace, Justin",
"Kishore, Varsha",
"Chen, Yiwei",
"Weinberger, Kilian"
] |
Diffusion Guided Language Modeling
|
findings-acl.887
|
Poster
|
2408.04220v1
|
https://aclanthology.org/2024.findings-acl.888.bib
|
@inproceedings{han-etal-2024-instructed,
title = "{I}nstruct{E}d: Soft-Instruction Tuning for Model Editing with Hops",
author = "Han, XiaoQi and
Li, Ru and
Li, Xiaoli and
Liang, Jiye and
Zhang, Zifang and
Pan, Jeff",
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.888",
pages = "14953--14968",
abstract = "The task of model editing becomes popular for correcting inaccurate or outdated parametric knowledge in Large Language Models (LLMs). However, there are major limitations of state of the art (SOTA) model editing methods, including the excessive memorization issue caused by the direct editing methods, as well as the error propagation and knowledge conflict issues from the memory enhancement methods, resulting in hindering models{'} *portability*, e.g., the ability to transfer the new knowledge to related one-hop or multi-hop content. To address these issues, we propose the InstructEd method, the idea of which is to insert soft instructions into the attention module so as to facilitate interactions between instructions and questions and to understand and utilize new facts. Our main findings are: (i) InstructEd has achieved SOTA performance on three datasets for one-hop/multi-hop evaluation with LLaMAs and GPT2, achieving 10{\%} (5{\%}) improvement in one-hop (multi-hop) model editing.(ii) Different from earlier methods on editing parameters in FFN, we show that editing attention can also help. (iii) Model editing is highly related to retrieval augmented methods, which can help improve the locality of model editing while slightly decrease the editing performance with hops.",
}
|
The task of model editing becomes popular for correcting inaccurate or outdated parametric knowledge in Large Language Models (LLMs). However, there are major limitations of state of the art (SOTA) model editing methods, including the excessive memorization issue caused by the direct editing methods, as well as the error propagation and knowledge conflict issues from the memory enhancement methods, resulting in hindering models{'} *portability*, e.g., the ability to transfer the new knowledge to related one-hop or multi-hop content. To address these issues, we propose the InstructEd method, the idea of which is to insert soft instructions into the attention module so as to facilitate interactions between instructions and questions and to understand and utilize new facts. Our main findings are: (i) InstructEd has achieved SOTA performance on three datasets for one-hop/multi-hop evaluation with LLaMAs and GPT2, achieving 10{\%} (5{\%}) improvement in one-hop (multi-hop) model editing.(ii) Different from earlier methods on editing parameters in FFN, we show that editing attention can also help. (iii) Model editing is highly related to retrieval augmented methods, which can help improve the locality of model editing while slightly decrease the editing performance with hops.
|
[
"Han, XiaoQi",
"Li, Ru",
"Li, Xiaoli",
"Liang, Jiye",
"Zhang, Zifang",
"Pan, Jeff"
] |
{I}nstruct{E}d: Soft-Instruction Tuning for Model Editing with Hops
|
findings-acl.888
|
Poster
|
2312.15194v2
|
https://aclanthology.org/2024.findings-acl.889.bib
|
@inproceedings{yoon-etal-2024-tlcr,
title = "{TLCR}: Token-Level Continuous Reward for Fine-grained Reinforcement Learning from Human Feedback",
author = "Yoon, Eunseop and
Yoon, Hee Suk and
Eom, SooHwan and
Han, Gunsoo and
Nam, Daniel and
Jo, Daejin and
On, Kyoung-Woon and
Hasegawa-Johnson, Mark and
Kim, Sungwoong and
Yoo, Chang",
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.889",
pages = "14969--14981",
abstract = "Reinforcement Learning from Human Feedback (RLHF) leverages human preference data to train language models to align more closely with human essence. These human preference data, however, are labeled at the sequence level, creating a mismatch between sequence-level preference labels and tokens, which are autoregressively generated from the language model. Although several recent approaches have tried to provide token-level (i.e., dense) rewards for each individual token, these typically rely on predefined discrete reward values (e.g., positive: +1, negative: -1, neutral: 0), failing to account for varying degrees of preference inherent to each token. To address this limitation, we introduce TLCR (Token-Level Continuous Reward) for RLHF, which incorporates a discriminator trained to distinguish positive and negative tokens, and the confidence of the discriminator is used to assign continuous rewards to each token considering the context. Extensive experiments show that our proposed TLCR leads to consistent performance improvements over previous sequence-level or token-level discrete rewards on open-ended generation benchmarks.",
}
|
Reinforcement Learning from Human Feedback (RLHF) leverages human preference data to train language models to align more closely with human essence. These human preference data, however, are labeled at the sequence level, creating a mismatch between sequence-level preference labels and tokens, which are autoregressively generated from the language model. Although several recent approaches have tried to provide token-level (i.e., dense) rewards for each individual token, these typically rely on predefined discrete reward values (e.g., positive: +1, negative: -1, neutral: 0), failing to account for varying degrees of preference inherent to each token. To address this limitation, we introduce TLCR (Token-Level Continuous Reward) for RLHF, which incorporates a discriminator trained to distinguish positive and negative tokens, and the confidence of the discriminator is used to assign continuous rewards to each token considering the context. Extensive experiments show that our proposed TLCR leads to consistent performance improvements over previous sequence-level or token-level discrete rewards on open-ended generation benchmarks.
|
[
"Yoon, Eunseop",
"Yoon, Hee Suk",
"Eom, SooHwan",
"Han, Gunsoo",
"Nam, Daniel",
"Jo, Daejin",
"On, Kyoung-Woon",
"Hasegawa-Johnson, Mark",
"Kim, Sungwoong",
"Yoo, Chang"
] |
{TLCR}: Token-Level Continuous Reward for Fine-grained Reinforcement Learning from Human Feedback
|
findings-acl.889
|
Poster
|
2407.16574v1
|
https://aclanthology.org/2024.findings-acl.890.bib
|
@inproceedings{hsieh-etal-2024-found,
title = "Found in the middle: Calibrating Positional Attention Bias Improves Long Context Utilization",
author = "Hsieh, Cheng-Yu and
Chuang, Yung-Sung and
Li, Chun-Liang and
Wang, Zifeng and
Le, Long and
Kumar, Abhishek and
Glass, James and
Ratner, Alexander and
Lee, Chen-Yu and
Krishna, Ranjay and
Pfister, Tomas",
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.890",
pages = "14982--14995",
abstract = "Large language models (LLMs), even when specifically trained to process long input contexts, struggle to capture relevant information located in the middle of their input. This phenomenon has been known as the lost-in-the-middle problem. In this work, we make three contributions. First, we set out to understand the factors that cause this phenomenon. In doing so, we establish a connection between lost-in-the-middle to LLMs{'} intrinsic attention bias: LLMs exhibit an U-shaped attention bias where the tokens at the beginning and at the end of its input receive higher attention, regardless of their relevance. Second, we mitigate this positional bias through a calibration mechanism, found-in-the-middle, that allows the model to attend to contexts faithfully according to their relevance, even though when they are in the middle. Third, we show found-in-the-middle not only achieves better performance in locating relevant information within a long context, but also eventually leads to improved retrieval-augmented generation (RAG) performance across various tasks, outperforming existing methods by up to 10 percentage point. These findings open up future directions in understanding LLM attention bias and its potential consequences.",
}
|
Large language models (LLMs), even when specifically trained to process long input contexts, struggle to capture relevant information located in the middle of their input. This phenomenon has been known as the lost-in-the-middle problem. In this work, we make three contributions. First, we set out to understand the factors that cause this phenomenon. In doing so, we establish a connection between lost-in-the-middle to LLMs{'} intrinsic attention bias: LLMs exhibit an U-shaped attention bias where the tokens at the beginning and at the end of its input receive higher attention, regardless of their relevance. Second, we mitigate this positional bias through a calibration mechanism, found-in-the-middle, that allows the model to attend to contexts faithfully according to their relevance, even though when they are in the middle. Third, we show found-in-the-middle not only achieves better performance in locating relevant information within a long context, but also eventually leads to improved retrieval-augmented generation (RAG) performance across various tasks, outperforming existing methods by up to 10 percentage point. These findings open up future directions in understanding LLM attention bias and its potential consequences.
|
[
"Hsieh, Cheng-Yu",
"Chuang, Yung-Sung",
"Li, Chun-Liang",
"Wang, Zifeng",
"Le, Long",
"Kumar, Abhishek",
"Glass, James",
"Ratner, Alex",
"er",
"Lee, Chen-Yu",
"Krishna, Ranjay",
"Pfister, Tomas"
] |
Found in the middle: Calibrating Positional Attention Bias Improves Long Context Utilization
|
findings-acl.890
|
Poster
|
2406.16008v2
|
https://aclanthology.org/2024.findings-acl.891.bib
|
@inproceedings{das-etal-2024-s3,
title = "S3-{DST}: Structured Open-Domain Dialogue Segmentation and State Tracking in the Era of {LLM}s",
author = "Das, Sarkar Snigdha Sarathi and
Shah, Chirag and
Wan, Mengting and
Neville, Jennifer and
Yang, Longqi and
Andersen, Reid and
Buscher, Georg and
Safavi, Tara",
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.891",
pages = "14996--15014",
abstract = "Traditional Dialogue State Tracking (DST) has focused on tracking preferences and intents in conversations centered around specific tasks (e.g. booking services). These conventional systems assume a relatively restricted conversation flow in which each turn gradually offers new information. However, advancements in Large Language Models (LLMs) have ushered in more versatile open-domain chat systems in which extended dialogue sessions encompassing numerous tasks and topics are common{---}in turn requiring new conversational tracking tools in order to successfully orchestrate such systems. Addressing these challenges, we introduce a novel approach combining dialogue segmentation and state tracking within open-domain dialogues, tailored for zero-shot applications appropriate to a true open-domain dialogue system. Our proposed method S3-DST employs a unique structured prompting technique and *Pre-Analytical Recollection*, a novel grounding mechanism we designed for improving long context tracking. Tested on proprietary anonymized open-domain dialogue datasets as well as publicly available DST and segmentation datasets, S3-DST consistently outperforms the state-of-the-art, showcasing its effectiveness and adaptability state tracking in the next wave of LLM-based chat systems. We also release S3-DST annotations with GPT-4 on a curated subset of LMSYS-Chat-1M to be used as a testbed to fuel research in this direction.",
}
|
Traditional Dialogue State Tracking (DST) has focused on tracking preferences and intents in conversations centered around specific tasks (e.g. booking services). These conventional systems assume a relatively restricted conversation flow in which each turn gradually offers new information. However, advancements in Large Language Models (LLMs) have ushered in more versatile open-domain chat systems in which extended dialogue sessions encompassing numerous tasks and topics are common{---}in turn requiring new conversational tracking tools in order to successfully orchestrate such systems. Addressing these challenges, we introduce a novel approach combining dialogue segmentation and state tracking within open-domain dialogues, tailored for zero-shot applications appropriate to a true open-domain dialogue system. Our proposed method S3-DST employs a unique structured prompting technique and *Pre-Analytical Recollection*, a novel grounding mechanism we designed for improving long context tracking. Tested on proprietary anonymized open-domain dialogue datasets as well as publicly available DST and segmentation datasets, S3-DST consistently outperforms the state-of-the-art, showcasing its effectiveness and adaptability state tracking in the next wave of LLM-based chat systems. We also release S3-DST annotations with GPT-4 on a curated subset of LMSYS-Chat-1M to be used as a testbed to fuel research in this direction.
|
[
"Das, Sarkar Snigdha Sarathi",
"Shah, Chirag",
"Wan, Mengting",
"Neville, Jennifer",
"Yang, Longqi",
"Andersen, Reid",
"Buscher, Georg",
"Safavi, Tara"
] |
S3-{DST}: Structured Open-Domain Dialogue Segmentation and State Tracking in the Era of {LLM}s
|
findings-acl.891
|
Poster
|
2309.08827v1
|
https://aclanthology.org/2024.findings-acl.892.bib
|
@inproceedings{zhao-etal-2024-set,
title = "Set the Clock: Temporal Alignment of Pretrained Language Models",
author = "Zhao, Bowen and
Brumbaugh, Zander and
Wang, Yizhong and
Hajishirzi, Hannaneh and
Smith, Noah",
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.892",
pages = "15015--15040",
abstract = "Language models (LMs) are trained on web text originating from many points in time and, in general, without any explicit temporal grounding. This work investigates the temporal chaos of pretrained LMs and explores various methods to align their internal knowledge to a target time, which we call {``}temporal alignment.{''} To do this, we first automatically construct a dataset containing 20K time-sensitive questions and their answers for each year from 2000 to 2023. Based on this dataset, we empirically show that pretrained LMs (e.g., LLaMa2), despite having a recent pretraining cutoff (e.g., 2022), mostly answer questions using earlier knowledge (e.g., in 2019). We then develop several methods, from prompting to finetuning, to align LMs to use their most recent knowledge when answering questions, and investigate various factors in this alignment. Our experiments demonstrate that aligning LLaMa2 to the year 2022 can enhance its performance by up to 62{\%} according to that year{'}s answers. This improvement occurs even without explicitly mentioning time information, indicating the possibility of aligning models{'} internal sense of time after pretraining. Finally, we find that alignment to a historical time is also possible, with up to $2.8\times$ the performance of the unaligned LM in 2010 if finetuning models to that year. These findings hint at the sophistication of LMs{'} internal knowledge organization and the necessity of tuning them properly.",
}
|
Language models (LMs) are trained on web text originating from many points in time and, in general, without any explicit temporal grounding. This work investigates the temporal chaos of pretrained LMs and explores various methods to align their internal knowledge to a target time, which we call {``}temporal alignment.{''} To do this, we first automatically construct a dataset containing 20K time-sensitive questions and their answers for each year from 2000 to 2023. Based on this dataset, we empirically show that pretrained LMs (e.g., LLaMa2), despite having a recent pretraining cutoff (e.g., 2022), mostly answer questions using earlier knowledge (e.g., in 2019). We then develop several methods, from prompting to finetuning, to align LMs to use their most recent knowledge when answering questions, and investigate various factors in this alignment. Our experiments demonstrate that aligning LLaMa2 to the year 2022 can enhance its performance by up to 62{\%} according to that year{'}s answers. This improvement occurs even without explicitly mentioning time information, indicating the possibility of aligning models{'} internal sense of time after pretraining. Finally, we find that alignment to a historical time is also possible, with up to $2.8\times$ the performance of the unaligned LM in 2010 if finetuning models to that year. These findings hint at the sophistication of LMs{'} internal knowledge organization and the necessity of tuning them properly.
|
[
"Zhao, Bowen",
"Brumbaugh, Z",
"er",
"Wang, Yizhong",
"Hajishirzi, Hannaneh",
"Smith, Noah"
] |
Set the Clock: Temporal Alignment of Pretrained Language Models
|
findings-acl.892
|
Poster
|
2407.16222v1
|
https://aclanthology.org/2024.findings-acl.893.bib
|
@inproceedings{ermis-etal-2024-one,
title = "From One to Many: Expanding the Scope of Toxicity Mitigation in Language Models",
author = "Ermis, Beyza and
Pozzobon, Luiza and
Hooker, Sara and
Lewis, Patrick",
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.893",
pages = "15041--15058",
abstract = "To date, toxicity mitigation in language models has almost entirely been focused on single-language settings. As language models embrace multilingual capabilities, it{'}s crucial our safety measures keep pace. Recognizing this research gap, our approach expands the scope of conventional toxicity mitigation to address the complexities presented by multiple languages. In the absence of sufficient annotated datasets across languages, we employ translated data to evaluate and enhance our mitigation techniques. We also compare finetuning mitigation approaches against retrieval-augmented techniques under both static and continual toxicity mitigation scenarios. This allows us to examine the effects of translation quality and the cross-lingual transfer on toxicity mitigation. We also explore how model size and data quantity affect the success of these mitigation efforts. Covering nine languages, our study represents a broad array of linguistic families and levels of resource availability, ranging from high to mid-resource languages. Through comprehensive experiments, we provide insights into the complexities of multilingual toxicity mitigation, offering valuable insights and paving the way for future research in this increasingly important field.",
}
|
To date, toxicity mitigation in language models has almost entirely been focused on single-language settings. As language models embrace multilingual capabilities, it{'}s crucial our safety measures keep pace. Recognizing this research gap, our approach expands the scope of conventional toxicity mitigation to address the complexities presented by multiple languages. In the absence of sufficient annotated datasets across languages, we employ translated data to evaluate and enhance our mitigation techniques. We also compare finetuning mitigation approaches against retrieval-augmented techniques under both static and continual toxicity mitigation scenarios. This allows us to examine the effects of translation quality and the cross-lingual transfer on toxicity mitigation. We also explore how model size and data quantity affect the success of these mitigation efforts. Covering nine languages, our study represents a broad array of linguistic families and levels of resource availability, ranging from high to mid-resource languages. Through comprehensive experiments, we provide insights into the complexities of multilingual toxicity mitigation, offering valuable insights and paving the way for future research in this increasingly important field.
|
[
"Ermis, Beyza",
"Pozzobon, Luiza",
"Hooker, Sara",
"Lewis, Patrick"
] |
From One to Many: Expanding the Scope of Toxicity Mitigation in Language Models
|
findings-acl.893
|
Poster
|
2403.03893v3
|
https://aclanthology.org/2024.findings-acl.894.bib
|
@inproceedings{arora-etal-2024-heres,
title = "Here{'}s a Free Lunch: Sanitizing Backdoored Models with Model Merge",
author = "Arora, Ansh and
He, Xuanli and
Mozes, Maximilian and
Swain, Srinibas and
Dras, Mark and
Xu, Qiongkai",
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.894",
pages = "15059--15075",
abstract = "The democratization of pre-trained language models through open-source initiatives has rapidly advanced innovation and expanded access to cutting-edge technologies. However, this openness also brings significant security risks, including backdoor attacks, where hidden malicious behaviors are triggered by specific inputs, compromising natural language processing (NLP) system integrity and reliability. This paper suggests that merging a backdoored model with other homogeneous models can significantly remediate backdoor vulnerabilities even if such models are not entirely secure. In our experiments, we verify our hypothesis on various models (BERT-Base, RoBERTa-Large, Llama2-7B, and Mistral-7B) and datasets (SST-2, OLID, AG News, and QNLI). Compared to multiple advanced defensive approaches, our method offers an effective and efficient inference-stage defense against backdoor attacks on classification and instruction-tuned tasks without additional resources or specific knowledge. Our approach consistently outperforms recent advanced baselines, leading to an average of about 75{\%} reduction in the attack success rate. Since model merging has been an established approach for improving model performance, the extra advantage it provides regarding defense can be seen as a cost-free bonus.",
}
|
The democratization of pre-trained language models through open-source initiatives has rapidly advanced innovation and expanded access to cutting-edge technologies. However, this openness also brings significant security risks, including backdoor attacks, where hidden malicious behaviors are triggered by specific inputs, compromising natural language processing (NLP) system integrity and reliability. This paper suggests that merging a backdoored model with other homogeneous models can significantly remediate backdoor vulnerabilities even if such models are not entirely secure. In our experiments, we verify our hypothesis on various models (BERT-Base, RoBERTa-Large, Llama2-7B, and Mistral-7B) and datasets (SST-2, OLID, AG News, and QNLI). Compared to multiple advanced defensive approaches, our method offers an effective and efficient inference-stage defense against backdoor attacks on classification and instruction-tuned tasks without additional resources or specific knowledge. Our approach consistently outperforms recent advanced baselines, leading to an average of about 75{\%} reduction in the attack success rate. Since model merging has been an established approach for improving model performance, the extra advantage it provides regarding defense can be seen as a cost-free bonus.
|
[
"Arora, Ansh",
"He, Xuanli",
"Mozes, Maximilian",
"Swain, Srinibas",
"Dras, Mark",
"Xu, Qiongkai"
] |
Here{'}s a Free Lunch: Sanitizing Backdoored Models with Model Merge
|
findings-acl.894
|
Poster
|
2402.19334v2
|
https://aclanthology.org/2024.findings-acl.895.bib
|
@inproceedings{scalercio-etal-2024-enhancing,
title = "Enhancing Sentence Simplification in {P}ortuguese: Leveraging Paraphrases, Context, and Linguistic Features",
author = "Scalercio, Arthur and
Finatto, Maria and
Paes, Aline",
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.895",
pages = "15076--15091",
abstract = "Automatic text simplification focuses on transforming texts into a more comprehensible version without sacrificing their precision. However, automatic methods usually require (paired) datasets that can be rather scarce in languages other than English. This paper presents a new approach to automatic sentence simplification that leverages paraphrases, context, and linguistic attributes to overcome the absence of paired texts in Portuguese.We frame the simplification problem as a textual style transfer task and learn a style representation using the sentences around the target sentence in the document and its linguistic attributes. Moreover, unlike most unsupervised approaches that require style-labeled training data, we fine-tune strong pre-trained models using sentence-level paraphrases instead of annotated data. Our experiments show that our model achieves remarkable results, surpassing the current state-of-the-art (BART+ACCESS) while competitively matching a Large Language Model.",
}
|
Automatic text simplification focuses on transforming texts into a more comprehensible version without sacrificing their precision. However, automatic methods usually require (paired) datasets that can be rather scarce in languages other than English. This paper presents a new approach to automatic sentence simplification that leverages paraphrases, context, and linguistic attributes to overcome the absence of paired texts in Portuguese.We frame the simplification problem as a textual style transfer task and learn a style representation using the sentences around the target sentence in the document and its linguistic attributes. Moreover, unlike most unsupervised approaches that require style-labeled training data, we fine-tune strong pre-trained models using sentence-level paraphrases instead of annotated data. Our experiments show that our model achieves remarkable results, surpassing the current state-of-the-art (BART+ACCESS) while competitively matching a Large Language Model.
|
[
"Scalercio, Arthur",
"Finatto, Maria",
"Paes, Aline"
] |
Enhancing Sentence Simplification in {P}ortuguese: Leveraging Paraphrases, Context, and Linguistic Features
|
findings-acl.895
|
Poster
|
2310.00355v1
|
https://aclanthology.org/2024.findings-acl.896.bib
|
@inproceedings{wu-etal-2024-far,
title = "How Far can 100 Samples Go? Unlocking Zero-Shot Translation with Tiny Multi-Parallel Data",
author = "Wu, Di and
Tan, Shaomu and
Meng, Yan and
Stap, David and
Monz, Christof",
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.896",
pages = "15092--15108",
abstract = "Zero-shot translation aims to translate between language pairs not seen during training in Multilingual Machine Translation (MMT) and is widely considered an open problem. A common, albeit resource-consuming, solution is to add as many related translation directions as possible to the training corpus. In this paper, we show that for an English-centric model, surprisingly large zero-shot improvements can be achieved by simply fine-tuning with a very small amount of multi-parallel data. For example, on the EC30 dataset, we obtain up to +21.7 ChrF++ non-English overall improvements (870 directions) by using only 100 multi-parallel samples while preserving English-centric translation quality. This performance exceeds M2M100 by an average of 5.9 ChrF++ in the involved non-English directions. When investigating the size effect of fine-tuning data on translation quality, we found that already a small, randomly sampled set of fine-tuning directions is sufficient to achieve comparable improvements. The resulting non-English performance is close to the complete translation upper bound. Even in a minimal setting{---}fine-tuning with only one single sample{---}the well-known off-target issue is almost completely resolved, explaining parts{---}but not all{---}of the observed improvements in translation quality.",
}
|
Zero-shot translation aims to translate between language pairs not seen during training in Multilingual Machine Translation (MMT) and is widely considered an open problem. A common, albeit resource-consuming, solution is to add as many related translation directions as possible to the training corpus. In this paper, we show that for an English-centric model, surprisingly large zero-shot improvements can be achieved by simply fine-tuning with a very small amount of multi-parallel data. For example, on the EC30 dataset, we obtain up to +21.7 ChrF++ non-English overall improvements (870 directions) by using only 100 multi-parallel samples while preserving English-centric translation quality. This performance exceeds M2M100 by an average of 5.9 ChrF++ in the involved non-English directions. When investigating the size effect of fine-tuning data on translation quality, we found that already a small, randomly sampled set of fine-tuning directions is sufficient to achieve comparable improvements. The resulting non-English performance is close to the complete translation upper bound. Even in a minimal setting{---}fine-tuning with only one single sample{---}the well-known off-target issue is almost completely resolved, explaining parts{---}but not all{---}of the observed improvements in translation quality.
|
[
"Wu, Di",
"Tan, Shaomu",
"Meng, Yan",
"Stap, David",
"Monz, Christof"
] |
How Far can 100 Samples Go? Unlocking Zero-Shot Translation with Tiny Multi-Parallel Data
|
findings-acl.896
|
Poster
|
2401.12413v2
|
https://aclanthology.org/2024.findings-acl.897.bib
|
@inproceedings{ghosh-etal-2024-toward,
title = "Toward Reliable Ad-hoc Scientific Information Extraction: A Case Study on Two Materials Dataset",
author = "Ghosh, Satanu and
Brodnik, Neal and
Frey, Carolina and
Holgate, Collin and
Pollock, Tresa and
Daly, Samantha and
Carton, Samuel",
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.897",
pages = "15109--15123",
abstract = "We explore the ability of GPT-4 to perform ad-hoc schema-based information extraction from scientific literature. We assess specifically whether it can, with a basic one-shot prompting approach over the full text of the included manusciprts, replicate two existing material science datasets, one pertaining to multi-principal element alloys (MPEAs), and one to silicate diffusion. We collaborate with materials scientists to perform a detailed manual error analysis to assess where and why the model struggles to faithfully extract the desired information, and draw on their insights to suggest research directions to address this broadly important task.",
}
|
We explore the ability of GPT-4 to perform ad-hoc schema-based information extraction from scientific literature. We assess specifically whether it can, with a basic one-shot prompting approach over the full text of the included manusciprts, replicate two existing material science datasets, one pertaining to multi-principal element alloys (MPEAs), and one to silicate diffusion. We collaborate with materials scientists to perform a detailed manual error analysis to assess where and why the model struggles to faithfully extract the desired information, and draw on their insights to suggest research directions to address this broadly important task.
|
[
"Ghosh, Satanu",
"Brodnik, Neal",
"Frey, Carolina",
"Holgate, Collin",
"Pollock, Tresa",
"Daly, Samantha",
"Carton, Samuel"
] |
Toward Reliable Ad-hoc Scientific Information Extraction: A Case Study on Two Materials Dataset
|
findings-acl.897
|
Poster
|
2406.05348v1
|
https://aclanthology.org/2024.findings-acl.898.bib
|
@inproceedings{park-kim-2024-structural,
title = "Structural Optimization Ambiguity and Simplicity Bias in Unsupervised Neural Grammar Induction",
author = "Park, Jinwook and
Kim, Kangil",
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.898",
pages = "15124--15139",
abstract = "Neural parameterization has significantly advanced unsupervised grammar induction. However, training these models with a traditional likelihood loss for all possible parses exacerbates two issues: 1) *structural optimization ambiguity* that arbitrarily selects one among structurally ambiguous optimal grammars despite the specific preference of gold parses, and 2) *structural simplicity bias* that leads a model to underutilize rules to compose parse trees. These challenges subject unsupervised neural grammar induction (UNGI) to inevitable prediction errors, high variance, and the necessity for extensive grammars to achieve accurate predictions. This paper tackles these issues, offering a comprehensive analysis of their origins. As a solution, we introduce *sentence-wise parse-focusing* to reduce the parse pool per sentence for loss evaluation, using the structural bias from pre-trained parsers on the same dataset.In unsupervised parsing benchmark tests, our method significantly improves performance while effectively reducing variance and bias toward overly simplistic parses. Our research promotes learning more compact, accurate, and consistent explicit grammars, facilitating better interpretability.",
}
|
Neural parameterization has significantly advanced unsupervised grammar induction. However, training these models with a traditional likelihood loss for all possible parses exacerbates two issues: 1) *structural optimization ambiguity* that arbitrarily selects one among structurally ambiguous optimal grammars despite the specific preference of gold parses, and 2) *structural simplicity bias* that leads a model to underutilize rules to compose parse trees. These challenges subject unsupervised neural grammar induction (UNGI) to inevitable prediction errors, high variance, and the necessity for extensive grammars to achieve accurate predictions. This paper tackles these issues, offering a comprehensive analysis of their origins. As a solution, we introduce *sentence-wise parse-focusing* to reduce the parse pool per sentence for loss evaluation, using the structural bias from pre-trained parsers on the same dataset.In unsupervised parsing benchmark tests, our method significantly improves performance while effectively reducing variance and bias toward overly simplistic parses. Our research promotes learning more compact, accurate, and consistent explicit grammars, facilitating better interpretability.
|
[
"Park, Jinwook",
"Kim, Kangil"
] |
Structural Optimization Ambiguity and Simplicity Bias in Unsupervised Neural Grammar Induction
|
findings-acl.898
|
Poster
|
2407.16181v1
|
https://aclanthology.org/2024.findings-acl.899.bib
|
@inproceedings{perot-etal-2024-lmdx,
title = "{LMDX}: Language Model-based Document Information Extraction and Localization",
author = "Perot, Vincent and
Kang, Kai and
Luisier, Florian and
Su, Guolong and
Sun, Xiaoyu and
Boppana, Ramya Sree and
Wang, Zilong and
Wang, Zifeng and
Mu, Jiaqi and
Zhang, Hao and
Lee, Chen-Yu and
Hua, Nan",
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.899",
pages = "15140--15168",
abstract = "Large Language Models (LLM) have revolutionized Natural Language Processing (NLP), improving state-of-the-art and exhibiting emergent capabilities across various tasks. However, their application in extracting information from visually rich documents, which is at the core of many document processing workflows and involving the extraction of key entities from semi-structured documents, has not yet been successful. The main obstacles to adopting LLMs for this task include the absence of layout encoding within LLMs, which is critical for high quality extraction, and the lack of a grounding mechanism to localize the predicted entities within the document. In this paper, we introduce Language Model-based Document Information EXtraction and Localization (LMDX), a methodology to reframe the document information extraction task for a LLM. LMDX enables extraction of singular, repeated, and hierarchical entities, both with and without training data, while providing grounding guarantees and localizing the entities within the document. Finally, we apply LMDX to the PaLM 2-S and Gemini Pro LLMs and evaluate it on VRDU and CORD benchmarks, setting a new state-of-the-art and showing how LMDX enables the creation of high quality, data-efficient parsers.",
}
|
Large Language Models (LLM) have revolutionized Natural Language Processing (NLP), improving state-of-the-art and exhibiting emergent capabilities across various tasks. However, their application in extracting information from visually rich documents, which is at the core of many document processing workflows and involving the extraction of key entities from semi-structured documents, has not yet been successful. The main obstacles to adopting LLMs for this task include the absence of layout encoding within LLMs, which is critical for high quality extraction, and the lack of a grounding mechanism to localize the predicted entities within the document. In this paper, we introduce Language Model-based Document Information EXtraction and Localization (LMDX), a methodology to reframe the document information extraction task for a LLM. LMDX enables extraction of singular, repeated, and hierarchical entities, both with and without training data, while providing grounding guarantees and localizing the entities within the document. Finally, we apply LMDX to the PaLM 2-S and Gemini Pro LLMs and evaluate it on VRDU and CORD benchmarks, setting a new state-of-the-art and showing how LMDX enables the creation of high quality, data-efficient parsers.
|
[
"Perot, Vincent",
"Kang, Kai",
"Luisier, Florian",
"Su, Guolong",
"Sun, Xiaoyu",
"Boppana, Ramya Sree",
"Wang, Zilong",
"Wang, Zifeng",
"Mu, Jiaqi",
"Zhang, Hao",
"Lee, Chen-Yu",
"Hua, Nan"
] |
{LMDX}: Language Model-based Document Information Extraction and Localization
|
findings-acl.899
|
Poster
|
2206.11229v1
|
https://aclanthology.org/2024.findings-acl.900.bib
|
@inproceedings{nararatwong-etal-2024-dbqr,
title = "{DBQR}-{QA}: A Question Answering Dataset on a Hybrid of Database Querying and Reasoning",
author = "Nararatwong, Rungsiman and
Chen, Chung-Chi and
Kertkeidkachorn, Natthawut and
Takamura, Hiroya and
Ichise, Ryutaro",
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.900",
pages = "15169--15182",
abstract = "This paper introduces the Database Querying and Reasoning Dataset for Question Answering (DBQR-QA), aimed at addressing the gap in current question-answering (QA) research by emphasizing the essential processes of database querying and reasoning to answer questions. Specifically designed to accommodate sequential questions and multi-hop queries, DBQR-QA more accurately mirrors the dynamics of real-world information retrieval and analysis, with a particular focus on the financial reports of US companies. The dataset{'}s construction, the challenges encountered during its development, the performance of large language models on this dataset, and a human evaluation are thoroughly discussed to illustrate the dataset{'}s complexity and highlight future research directions in querying and reasoning tasks.",
}
|
This paper introduces the Database Querying and Reasoning Dataset for Question Answering (DBQR-QA), aimed at addressing the gap in current question-answering (QA) research by emphasizing the essential processes of database querying and reasoning to answer questions. Specifically designed to accommodate sequential questions and multi-hop queries, DBQR-QA more accurately mirrors the dynamics of real-world information retrieval and analysis, with a particular focus on the financial reports of US companies. The dataset{'}s construction, the challenges encountered during its development, the performance of large language models on this dataset, and a human evaluation are thoroughly discussed to illustrate the dataset{'}s complexity and highlight future research directions in querying and reasoning tasks.
|
[
"Nararatwong, Rungsiman",
"Chen, Chung-Chi",
"Kertkeidkachorn, Natthawut",
"Takamura, Hiroya",
"Ichise, Ryutaro"
] |
{DBQR}-{QA}: A Question Answering Dataset on a Hybrid of Database Querying and Reasoning
|
findings-acl.900
|
Poster
|
2108.02866v2
|
https://aclanthology.org/2024.findings-acl.901.bib
|
@inproceedings{wang-etal-2024-notechat,
title = "{N}ote{C}hat: A Dataset of Synthetic Patient-Physician Conversations Conditioned on Clinical Notes",
author = "Wang, Junda and
Yao, Zonghai and
Yang, Zhichao and
Zhou, Huixue and
Li, Rumeng and
Wang, Xun and
Xu, Yucheng and
Yu, Hong",
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.901",
pages = "15183--15201",
abstract = "We introduce NoteChat, a novel cooperative multi-agent framework leveraging Large Language Models (LLMs) to generate patient-physician dialogues. NoteChat embodies the principle that an ensemble of role-specific LLMs, through structured role-play and strategic prompting, can perform their assigned roles more effectively. The synergy among these role-playing LLMs results in a cohesive and efficient dialogue generation. Evaluation on MTS-dialogue, a benchmark dataset for patient-physician dialogues-note pairs, shows that models trained with the augmented synthetic patient-physician dialogues by NoteChat outperforms other state-of-the-art models for generating clinical notes. Our comprehensive automatic and human evaluation demonstrates that NoteChat substantially surpasses state-of-the-art models like ChatGPT and GPT-4 up to 22.78{\%} by domain experts in generating superior synthetic patient-physician dialogues based on clinical notes. NoteChat has the potential to engage patients directly and help clinical documentation, a leading cause of physician burnout.",
}
|
We introduce NoteChat, a novel cooperative multi-agent framework leveraging Large Language Models (LLMs) to generate patient-physician dialogues. NoteChat embodies the principle that an ensemble of role-specific LLMs, through structured role-play and strategic prompting, can perform their assigned roles more effectively. The synergy among these role-playing LLMs results in a cohesive and efficient dialogue generation. Evaluation on MTS-dialogue, a benchmark dataset for patient-physician dialogues-note pairs, shows that models trained with the augmented synthetic patient-physician dialogues by NoteChat outperforms other state-of-the-art models for generating clinical notes. Our comprehensive automatic and human evaluation demonstrates that NoteChat substantially surpasses state-of-the-art models like ChatGPT and GPT-4 up to 22.78{\%} by domain experts in generating superior synthetic patient-physician dialogues based on clinical notes. NoteChat has the potential to engage patients directly and help clinical documentation, a leading cause of physician burnout.
|
[
"Wang, Junda",
"Yao, Zonghai",
"Yang, Zhichao",
"Zhou, Huixue",
"Li, Rumeng",
"Wang, Xun",
"Xu, Yucheng",
"Yu, Hong"
] |
{N}ote{C}hat: A Dataset of Synthetic Patient-Physician Conversations Conditioned on Clinical Notes
|
findings-acl.901
|
Poster
|
2310.15959v3
|
https://aclanthology.org/2024.findings-acl.902.bib
|
@inproceedings{gupta-etal-2024-model,
title = "Model Editing at Scale leads to Gradual and Catastrophic Forgetting",
author = "Gupta, Akshat and
Rao, Anurag and
Anumanchipalli, Gopala",
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.902",
pages = "15202--15232",
abstract = "Editing knowledge in large language models is an attractive capability that allows us to correct incorrectly learned facts during pre-training, as well as update the model with an ever-growing list of new facts. While existing model editing techniques have shown promise, they are usually evaluated using metrics for reliability, specificity and generalization over one or few edits. We argue that for model editing to have practical utility, we must be able to make multiple edits to the same model. With this in mind, we evaluate current model editing methods at scale, focusing on two state of the art methods - ROME and MEMIT. With the lens of scalability, we evaluate model editing methods for three crucial properties - editing proficiency, fact forgetting and downstream performance. We find that as a model is edited sequentially with multiple facts, it continually becomes less editable, forgets previously edited facts and loses the ability to perform downstream tasks. For ROME and MEMIT, this {``}forgetting{''} happens in two phases - an initial gradual but progressive forgetting phase followed by an abrupt or catastrophic forgetting. Both gradual and catastrophic forgetting limit the usefulness of model editing methods at scale - the former makes model editing less effective as multiple edits are made to the model while the latter caps the scalability of such model editing methods. Our analysis also highlights other key limitations of ROME and MEMIT at scale. With our work, we push for better evaluation of model editing and development of model editing methods keeping scalability in mind.",
}
|
Editing knowledge in large language models is an attractive capability that allows us to correct incorrectly learned facts during pre-training, as well as update the model with an ever-growing list of new facts. While existing model editing techniques have shown promise, they are usually evaluated using metrics for reliability, specificity and generalization over one or few edits. We argue that for model editing to have practical utility, we must be able to make multiple edits to the same model. With this in mind, we evaluate current model editing methods at scale, focusing on two state of the art methods - ROME and MEMIT. With the lens of scalability, we evaluate model editing methods for three crucial properties - editing proficiency, fact forgetting and downstream performance. We find that as a model is edited sequentially with multiple facts, it continually becomes less editable, forgets previously edited facts and loses the ability to perform downstream tasks. For ROME and MEMIT, this {``}forgetting{''} happens in two phases - an initial gradual but progressive forgetting phase followed by an abrupt or catastrophic forgetting. Both gradual and catastrophic forgetting limit the usefulness of model editing methods at scale - the former makes model editing less effective as multiple edits are made to the model while the latter caps the scalability of such model editing methods. Our analysis also highlights other key limitations of ROME and MEMIT at scale. With our work, we push for better evaluation of model editing and development of model editing methods keeping scalability in mind.
|
[
"Gupta, Akshat",
"Rao, Anurag",
"Anumanchipalli, Gopala"
] |
Model Editing at Scale leads to Gradual and Catastrophic Forgetting
|
findings-acl.902
|
Poster
|
2401.07453v4
|
https://aclanthology.org/2024.findings-acl.903.bib
|
@inproceedings{ding-etal-2024-3mvrd,
title = "3{MVRD}: Multimodal Multi-task Multi-teacher Visually-Rich Form Document Understanding",
author = "Ding, Yihao and
Vaiani, Lorenzo and
Han, Caren and
Lee, Jean and
Garza, Paolo and
Poon, Josiah and
Cagliero, Luca",
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.903",
pages = "15233--15244",
abstract = "This paper presents a groundbreaking multimodal, multi-task, multi-teacher joint-grained knowledge distillation model for visually-rich form document understanding. The model is designed to leverage insights from both fine-grained and coarse-grained levels by facilitating a nuanced correlation between token and entity representations, addressing the complexities inherent in form documents. Additionally, we introduce new inter-grained and cross-grained loss functions to further refine diverse multi-teacher knowledge distillation transfer process, presenting distribution gaps and a harmonised understanding of form documents. Through a comprehensive evaluation across publicly available form document understanding datasets, our proposed model consistently outperforms existing baselines, showcasing its efficacy in handling the intricate structures and content of visually complex form documents.",
}
|
This paper presents a groundbreaking multimodal, multi-task, multi-teacher joint-grained knowledge distillation model for visually-rich form document understanding. The model is designed to leverage insights from both fine-grained and coarse-grained levels by facilitating a nuanced correlation between token and entity representations, addressing the complexities inherent in form documents. Additionally, we introduce new inter-grained and cross-grained loss functions to further refine diverse multi-teacher knowledge distillation transfer process, presenting distribution gaps and a harmonised understanding of form documents. Through a comprehensive evaluation across publicly available form document understanding datasets, our proposed model consistently outperforms existing baselines, showcasing its efficacy in handling the intricate structures and content of visually complex form documents.
|
[
"Ding, Yihao",
"Vaiani, Lorenzo",
"Han, Caren",
"Lee, Jean",
"Garza, Paolo",
"Poon, Josiah",
"Cagliero, Luca"
] |
3{MVRD}: Multimodal Multi-task Multi-teacher Visually-Rich Form Document Understanding
|
findings-acl.903
|
Poster
|
2305.09520v1
|
https://aclanthology.org/2024.findings-acl.904.bib
|
@inproceedings{jandaghi-etal-2024-faithful-persona,
title = "Faithful Persona-based Conversational Dataset Generation with Large Language Models",
author = "Jandaghi, Pegah and
Sheng, Xianghai and
Bai, Xinyi and
Pujara, Jay and
Sidahmed, Hakim",
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.904",
pages = "15245--15270",
abstract = "High-quality conversational datasets are essential for developing AI models that can communicate with users.One way to foster deeper interactions between a chatbot and its user is through *personas*, aspects of the user{'}s character that provide insights into their personality, motivations, and behaviors.Training Natural Language Processing (NLP) models on a diverse and comprehensive persona-based dataset can lead to conversational models that create a deeper connection with the user, and maintain their engagement. In this paper, we leverage the power of Large Language Models (LLMs) to create a large, high-quality conversational dataset from a seed dataset. We propose a Generator-Critic architecture framework to expand the initial dataset, while improving the quality of its conversations.The Generator is an LLM prompted to output conversations.The Critic consists of a mixture of expert LLMs that control the quality of the generated conversations.These experts select the best generated conversations, which we then use to improve the Generator.We release Synthetic-Persona-Chat, consisting of 20k conversations seeded from Persona-Chat.We evaluate the quality of Synthetic-Persona-Chat and our generation framework on different dimensions through extensive experiments, and observe that the losing rate of Synthetic-Persona-Chat against Persona-Chat during an AI detection test decreases from 17.2{\%} to 8.8{\%} over three iterations.",
}
|
High-quality conversational datasets are essential for developing AI models that can communicate with users.One way to foster deeper interactions between a chatbot and its user is through *personas*, aspects of the user{'}s character that provide insights into their personality, motivations, and behaviors.Training Natural Language Processing (NLP) models on a diverse and comprehensive persona-based dataset can lead to conversational models that create a deeper connection with the user, and maintain their engagement. In this paper, we leverage the power of Large Language Models (LLMs) to create a large, high-quality conversational dataset from a seed dataset. We propose a Generator-Critic architecture framework to expand the initial dataset, while improving the quality of its conversations.The Generator is an LLM prompted to output conversations.The Critic consists of a mixture of expert LLMs that control the quality of the generated conversations.These experts select the best generated conversations, which we then use to improve the Generator.We release Synthetic-Persona-Chat, consisting of 20k conversations seeded from Persona-Chat.We evaluate the quality of Synthetic-Persona-Chat and our generation framework on different dimensions through extensive experiments, and observe that the losing rate of Synthetic-Persona-Chat against Persona-Chat during an AI detection test decreases from 17.2{\%} to 8.8{\%} over three iterations.
|
[
"J",
"aghi, Pegah",
"Sheng, Xianghai",
"Bai, Xinyi",
"Pujara, Jay",
"Sidahmed, Hakim"
] |
Faithful Persona-based Conversational Dataset Generation with Large Language Models
|
findings-acl.904
|
Poster
|
2303.03278v1
|
https://aclanthology.org/2024.findings-acl.905.bib
|
@inproceedings{xu-etal-2024-vision,
title = "Vision-Flan: Scaling Human-Labeled Tasks in Visual Instruction Tuning",
author = "Xu, Zhiyang and
Feng, Chao and
Shao, Rulin and
Ashby, Trevor and
Shen, Ying and
Jin, Di and
Cheng, Yu and
Wang, Qifan and
Huang, Lifu",
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.905",
pages = "15271--15342",
abstract = "Despite vision-language models{'} (VLMs) remarkable capabilities as versatile visual assistants, two substantial challenges persist within the existing VLM frameworks: (1) lacking task diversity in pretraining and visual instruction tuning, and (2) annotation error and bias in GPT-4 synthesized instruction tuning data. Both challenges lead to issues such as poor generalizability, hallucination, and catastrophic forgetting. To address these challenges, we construct Vision-Flan, the most diverse publicly available visual instruction tuning dataset to date, comprising 187 diverse tasks and 1,664,261 instances sourced from academic datasets, and each task is accompanied by an expert-written instruction. In addition, we propose a two-stage instruction tuning framework, in which VLMs are firstly finetuned on Vision-Flan and further tuned on GPT-4 synthesized data. We find this two-stage tuning framework significantly outperforms the traditional single-stage visual instruction tuning framework and achieves the state-of-the-art performance across a wide range of multi-modal evaluation benchmarks. Finally, we conduct in-depth analyses to understand visual instruction tuning and our findings reveal that: (1) GPT-4 synthesized data does not substantially enhance VLMs{'} capabilities but rather modulates the model{'}s responses to human-preferred formats; (2) A minimal quantity (e.g., 1,000) of GPT-4 synthesized data can effectively align VLM responses with human-preference; (3) Visual instruction tuning mainly helps large-language models (LLMs) to understand visual features.",
}
|
Despite vision-language models{'} (VLMs) remarkable capabilities as versatile visual assistants, two substantial challenges persist within the existing VLM frameworks: (1) lacking task diversity in pretraining and visual instruction tuning, and (2) annotation error and bias in GPT-4 synthesized instruction tuning data. Both challenges lead to issues such as poor generalizability, hallucination, and catastrophic forgetting. To address these challenges, we construct Vision-Flan, the most diverse publicly available visual instruction tuning dataset to date, comprising 187 diverse tasks and 1,664,261 instances sourced from academic datasets, and each task is accompanied by an expert-written instruction. In addition, we propose a two-stage instruction tuning framework, in which VLMs are firstly finetuned on Vision-Flan and further tuned on GPT-4 synthesized data. We find this two-stage tuning framework significantly outperforms the traditional single-stage visual instruction tuning framework and achieves the state-of-the-art performance across a wide range of multi-modal evaluation benchmarks. Finally, we conduct in-depth analyses to understand visual instruction tuning and our findings reveal that: (1) GPT-4 synthesized data does not substantially enhance VLMs{'} capabilities but rather modulates the model{'}s responses to human-preferred formats; (2) A minimal quantity (e.g., 1,000) of GPT-4 synthesized data can effectively align VLM responses with human-preference; (3) Visual instruction tuning mainly helps large-language models (LLMs) to understand visual features.
|
[
"Xu, Zhiyang",
"Feng, Chao",
"Shao, Rulin",
"Ashby, Trevor",
"Shen, Ying",
"Jin, Di",
"Cheng, Yu",
"Wang, Qifan",
"Huang, Lifu"
] |
Vision-Flan: Scaling Human-Labeled Tasks in Visual Instruction Tuning
|
findings-acl.905
|
Poster
|
2403.04343v1
|
https://aclanthology.org/2024.findings-acl.906.bib
|
@inproceedings{powell-etal-2024-taxi,
title = "{TAXI}: Evaluating Categorical Knowledge Editing for Language Models",
author = "Powell, Derek and
Gerych, Walter and
Hartvigsen, 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.906",
pages = "15343--15352",
abstract = "Humans rarely learn one fact in isolation. Instead, learning a new fact induces knowledge of other facts about the world. For example, in learning a korat is a type of cat, you also infer it is a mammal and has claws, ensuring your model of the world is consistent. Knowledge editing aims to inject new facts into language models to improve their factuality, but current benchmarks fail to evaluate consistency, which is critical to ensure efficient, accurate, and generalizable edits. We manually create TAXI, a new benchmark dataset specifically created to evaluate consistency in categorical knowledge edits. TAXI contains 11,120 multiple-choice queries for 976 edits spanning 41 categories (e.g., Dogs), 164 subjects (e.g., Labrador), and 183 properties (e.g., is a mammal). We then use TAXI to evaluate popular editors{'} categorical consistency, measuring how often editing a subject{'}s category appropriately edits its properties. We find that 1) the editors achieve marginal, yet non-random consistency, 2) their consistency far underperforms human baselines, and 3) consistency is more achievable when editing atypical subjects.",
}
|
Humans rarely learn one fact in isolation. Instead, learning a new fact induces knowledge of other facts about the world. For example, in learning a korat is a type of cat, you also infer it is a mammal and has claws, ensuring your model of the world is consistent. Knowledge editing aims to inject new facts into language models to improve their factuality, but current benchmarks fail to evaluate consistency, which is critical to ensure efficient, accurate, and generalizable edits. We manually create TAXI, a new benchmark dataset specifically created to evaluate consistency in categorical knowledge edits. TAXI contains 11,120 multiple-choice queries for 976 edits spanning 41 categories (e.g., Dogs), 164 subjects (e.g., Labrador), and 183 properties (e.g., is a mammal). We then use TAXI to evaluate popular editors{'} categorical consistency, measuring how often editing a subject{'}s category appropriately edits its properties. We find that 1) the editors achieve marginal, yet non-random consistency, 2) their consistency far underperforms human baselines, and 3) consistency is more achievable when editing atypical subjects.
|
[
"Powell, Derek",
"Gerych, Walter",
"Hartvigsen, Thomas"
] |
{TAXI}: Evaluating Categorical Knowledge Editing for Language Models
|
findings-acl.906
|
Poster
|
2404.15004v2
|
https://aclanthology.org/2024.findings-acl.907.bib
|
@inproceedings{jin-etal-2024-automatic,
title = "Automatic Bug Detection in {LLM}-Powered Text-Based Games Using {LLM}s",
author = "Jin, Claire and
Rao, Sudha and
Peng, Xiangyu and
Botchway, Portia and
Quaye, Jessica and
Brockett, Chris and
Dolan, Bill",
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.907",
pages = "15353--15368",
abstract = "Advancements in large language models (LLMs) are revolutionizing interactive game design, enabling dynamic plotlines and interactions between players and non-player characters (NPCs). However, LLMs may exhibit flaws such as hallucinations, forgetfulness, or misinterpretations of prompts, causing logical inconsistencies and unexpected deviations from intended designs. Automated techniques for detecting such game bugs are still lacking. To address this, we propose a systematic LLM-based method for automatically identifying such bugs from player game logs, eliminating the need for collecting additional data such as post-play surveys. Applied to a text-based game DejaBoom!, our approach effectively identifies bugs inherent in LLM-powered interactive games, surpassing unstructured LLM-powered bug-catching methods and filling the gap in automated detection of logical and design flaws.",
}
|
Advancements in large language models (LLMs) are revolutionizing interactive game design, enabling dynamic plotlines and interactions between players and non-player characters (NPCs). However, LLMs may exhibit flaws such as hallucinations, forgetfulness, or misinterpretations of prompts, causing logical inconsistencies and unexpected deviations from intended designs. Automated techniques for detecting such game bugs are still lacking. To address this, we propose a systematic LLM-based method for automatically identifying such bugs from player game logs, eliminating the need for collecting additional data such as post-play surveys. Applied to a text-based game DejaBoom!, our approach effectively identifies bugs inherent in LLM-powered interactive games, surpassing unstructured LLM-powered bug-catching methods and filling the gap in automated detection of logical and design flaws.
|
[
"Jin, Claire",
"Rao, Sudha",
"Peng, Xiangyu",
"Botchway, Portia",
"Quaye, Jessica",
"Brockett, Chris",
"Dolan, Bill"
] |
Automatic Bug Detection in {LLM}-Powered Text-Based Games Using {LLM}s
|
findings-acl.907
|
Poster
|
2406.04482v1
|
https://aclanthology.org/2024.findings-acl.908.bib
|
@inproceedings{amin-rayz-2024-embodied,
title = "Embodied Language Learning: Opportunities, Challenges, and Future Directions",
author = "Amin, Nadine and
Rayz, Julia",
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.908",
pages = "15369--15379",
abstract = "While large language and vision-language models showcase impressive capabilities, they face a notable limitation: the inability to connect language with the physical world. To bridge this gap, research has focused on embodied language learning, where the language learner is situated in the world, perceives it, and interacts with it. This article explores the current standing of research in embodied language learning, highlighting opportunities and discussing common challenges. Lastly, it identifies existing gaps from the perspective of language understanding research within the embodied world and suggests potential future directions.",
}
|
While large language and vision-language models showcase impressive capabilities, they face a notable limitation: the inability to connect language with the physical world. To bridge this gap, research has focused on embodied language learning, where the language learner is situated in the world, perceives it, and interacts with it. This article explores the current standing of research in embodied language learning, highlighting opportunities and discussing common challenges. Lastly, it identifies existing gaps from the perspective of language understanding research within the embodied world and suggests potential future directions.
|
[
"Amin, Nadine",
"Rayz, Julia"
] |
Embodied Language Learning: Opportunities, Challenges, and Future Directions
|
findings-acl.908
|
Poster
|
2402.14304v2
|
https://aclanthology.org/2024.findings-acl.909.bib
|
@inproceedings{porada-etal-2024-challenges,
title = "Challenges to Evaluating the Generalization of Coreference Resolution Models: A Measurement Modeling Perspective",
author = "Porada, Ian and
Olteanu, Alexandra and
Suleman, Kaheer and
Trischler, Adam and
Cheung, Jackie",
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.909",
pages = "15380--15395",
abstract = "It is increasingly common to evaluate the same coreference resolution (CR) model on multiple datasets. Do these multi-dataset evaluations allow us to draw meaningful conclusions about model generalization? Or, do they rather reflect the idiosyncrasies of a particular experimental setup (e.g., the specific datasets used)? To study this, we view evaluation through the lens of measurement modeling, a framework commonly used in the social sciences for analyzing the validity of measurements. By taking this perspective, we show how multi-dataset evaluations risk conflating different factors concerning what, precisely, is being measured. This in turn makes it difficult to draw more generalizable conclusions from these evaluations. For instance, we show that across seven datasets, measurements intended to reflect CR model generalization are often correlated with differences in both how coreference is defined and how it is operationalized; this limits our ability to draw conclusions regarding the ability of CR models to generalize across any singular dimension. We believe the measurement modeling framework provides the needed vocabulary for discussing challenges surrounding what is actually being measured by CR evaluations.",
}
|
It is increasingly common to evaluate the same coreference resolution (CR) model on multiple datasets. Do these multi-dataset evaluations allow us to draw meaningful conclusions about model generalization? Or, do they rather reflect the idiosyncrasies of a particular experimental setup (e.g., the specific datasets used)? To study this, we view evaluation through the lens of measurement modeling, a framework commonly used in the social sciences for analyzing the validity of measurements. By taking this perspective, we show how multi-dataset evaluations risk conflating different factors concerning what, precisely, is being measured. This in turn makes it difficult to draw more generalizable conclusions from these evaluations. For instance, we show that across seven datasets, measurements intended to reflect CR model generalization are often correlated with differences in both how coreference is defined and how it is operationalized; this limits our ability to draw conclusions regarding the ability of CR models to generalize across any singular dimension. We believe the measurement modeling framework provides the needed vocabulary for discussing challenges surrounding what is actually being measured by CR evaluations.
|
[
"Porada, Ian",
"Olteanu, Alex",
"ra",
"Suleman, Kaheer",
"Trischler, Adam",
"Cheung, Jackie"
] |
Challenges to Evaluating the Generalization of Coreference Resolution Models: A Measurement Modeling Perspective
|
findings-acl.909
|
Poster
|
2303.09092v2
|
https://aclanthology.org/2024.findings-acl.910.bib
|
@inproceedings{vallurupalli-etal-2024-saga,
title = "{SAGA}: A Participant-specific Examination of Story Alternatives and Goal Applicability for a Deeper Understanding of Complex Events",
author = "Vallurupalli, Sai and
Erk, Katrin and
Ferraro, Francis",
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.910",
pages = "15396--15420",
abstract = "Interpreting and assessing goal driven actions is vital to understanding and reasoning over complex events. It is important to be able to acquire the knowledge needed for this understanding, though doing so is challenging. We argue that such knowledge can be elicited through a participant achievement lens. We analyze a complex event in a narrative according to the intended achievements of the participants in that narrative, the likely future actions of the participants, and the likelihood of goal success. We collect 6.3K high quality goal and action annotations reflecting our proposed participant achievement lens, with an average weighted Fleiss-Kappa IAA of 80{\%}. Our collection contains annotated alternate versions of each narrative. These alternate versions vary minimally from the {``}original{''} story, but can license drastically different inferences. Our findings suggest that while modern large language models can reflect some of the goal-based knowledge we study, they find it challenging to fully capture the design and intent behind concerted actions, even when the model pretraining included the data from which we extracted the goal knowledge. We show that smaller models fine-tuned on our dataset can achieve performance surpassing larger models.",
}
|
Interpreting and assessing goal driven actions is vital to understanding and reasoning over complex events. It is important to be able to acquire the knowledge needed for this understanding, though doing so is challenging. We argue that such knowledge can be elicited through a participant achievement lens. We analyze a complex event in a narrative according to the intended achievements of the participants in that narrative, the likely future actions of the participants, and the likelihood of goal success. We collect 6.3K high quality goal and action annotations reflecting our proposed participant achievement lens, with an average weighted Fleiss-Kappa IAA of 80{\%}. Our collection contains annotated alternate versions of each narrative. These alternate versions vary minimally from the {``}original{''} story, but can license drastically different inferences. Our findings suggest that while modern large language models can reflect some of the goal-based knowledge we study, they find it challenging to fully capture the design and intent behind concerted actions, even when the model pretraining included the data from which we extracted the goal knowledge. We show that smaller models fine-tuned on our dataset can achieve performance surpassing larger models.
|
[
"Vallurupalli, Sai",
"Erk, Katrin",
"Ferraro, Francis"
] |
{SAGA}: A Participant-specific Examination of Story Alternatives and Goal Applicability for a Deeper Understanding of Complex Events
|
findings-acl.910
|
Poster
|
2408.05793v1
|
https://aclanthology.org/2024.findings-acl.911.bib
|
@inproceedings{zhao-etal-2024-slide,
title = "{SLIDE}: A Framework Integrating Small and Large Language Models for Open-Domain Dialogues Evaluation",
author = "Zhao, Kun and
Yang, Bohao and
Tang, Chen and
Lin, Chenghua and
Zhan, Liang",
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.911",
pages = "15421--15435",
abstract = "The long-standing one-to-many problem of gold standard responses in open-domain dialogue systems presents challenges for automatic evaluation metrics. Though prior works have demonstrated some success by applying powerful Large Language Models (LLMs), existing approaches still struggle with the one-to-many problem, and exhibit subpar performance in domain-specific scenarios. We assume the commonsense reasoning biases within LLMs may hinder their performance in domain-specific evaluations. To address both issues, we propose a novel framework SLIDE (Small and Large Integrated for Dialogue Evaluation), that leverages both a small, specialised model (SLM), and LLMs for the evaluation of open domain dialogues. Our approach introduces several techniques: (1) Contrastive learning to differentiate between robust and non-robust response embeddings; (2) A novel metric for semantic sensitivity that combines embedding cosine distances with similarity learned through neural networks, and (3) A strategy for incorporating the evaluation results from both the SLM and LLMs. Our empirical results demonstrate that our approach achieves state-of-the-art performance in both the classification and evaluation tasks, and additionally the SLIDE evaluator exhibits better correlation with human judgements. Our code is available at https://github.com/hegehongcha/SLIDE-ACL2024.",
}
|
The long-standing one-to-many problem of gold standard responses in open-domain dialogue systems presents challenges for automatic evaluation metrics. Though prior works have demonstrated some success by applying powerful Large Language Models (LLMs), existing approaches still struggle with the one-to-many problem, and exhibit subpar performance in domain-specific scenarios. We assume the commonsense reasoning biases within LLMs may hinder their performance in domain-specific evaluations. To address both issues, we propose a novel framework SLIDE (Small and Large Integrated for Dialogue Evaluation), that leverages both a small, specialised model (SLM), and LLMs for the evaluation of open domain dialogues. Our approach introduces several techniques: (1) Contrastive learning to differentiate between robust and non-robust response embeddings; (2) A novel metric for semantic sensitivity that combines embedding cosine distances with similarity learned through neural networks, and (3) A strategy for incorporating the evaluation results from both the SLM and LLMs. Our empirical results demonstrate that our approach achieves state-of-the-art performance in both the classification and evaluation tasks, and additionally the SLIDE evaluator exhibits better correlation with human judgements. Our code is available at https://github.com/hegehongcha/SLIDE-ACL2024.
|
[
"Zhao, Kun",
"Yang, Bohao",
"Tang, Chen",
"Lin, Chenghua",
"Zhan, Liang"
] |
{SLIDE}: A Framework Integrating Small and Large Language Models for Open-Domain Dialogues Evaluation
|
findings-acl.911
|
Poster
|
2405.15924v3
|
https://aclanthology.org/2024.findings-acl.912.bib
|
@inproceedings{han-etal-2024-deep,
title = "Deep Exploration of Cross-Lingual Zero-Shot Generalization in Instruction Tuning",
author = "Han, Janghoon and
Lee, Changho and
Shin, Joongbo and
Choi, Stanley Jungkyu and
Lee, Honglak and
Bae, Kyunghoon",
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.912",
pages = "15436--15452",
abstract = "Instruction tuning has emerged as a powerful technique, significantly boosting zero-shot performance on unseen tasks. While recent work has explored cross-lingual generalization by applying instruction tuning to multilingual models, previous studies have primarily focused on English, with a limited exploration of non-English tasks. For in-depth exploration of cross-lingual generalization in instruction tuning, we perform instruction tuning individually for two distinct language meta-datasets. Subsequently, we assess the performance on unseen tasks in the language different from the one used for training. To facilitate this investigation, we introduce a novel non-English meta-dataset named {``}KORANI{''} (Korean Natural Instruction), comprising 51 Korean benchmarks. Moreover, we design cross-lingual templates to mitigate discrepancies in language and instruction-format of the template between training and inference within the cross-lingual setting. Our experiments reveal consistent improvements through cross-lingual generalization in both English and Korean, outperforming baseline by average scores of 20.7{\%} and 13.6{\%}, respectively. Remarkably, these enhancements are comparable to those achieved by mono-lingual instruction tuning and even surpass them in some tasks. The result underscores the significance of relevant data acquisition across languages over linguistic congruence with unseen tasks during instruction tuning.",
}
|
Instruction tuning has emerged as a powerful technique, significantly boosting zero-shot performance on unseen tasks. While recent work has explored cross-lingual generalization by applying instruction tuning to multilingual models, previous studies have primarily focused on English, with a limited exploration of non-English tasks. For in-depth exploration of cross-lingual generalization in instruction tuning, we perform instruction tuning individually for two distinct language meta-datasets. Subsequently, we assess the performance on unseen tasks in the language different from the one used for training. To facilitate this investigation, we introduce a novel non-English meta-dataset named {``}KORANI{''} (Korean Natural Instruction), comprising 51 Korean benchmarks. Moreover, we design cross-lingual templates to mitigate discrepancies in language and instruction-format of the template between training and inference within the cross-lingual setting. Our experiments reveal consistent improvements through cross-lingual generalization in both English and Korean, outperforming baseline by average scores of 20.7{\%} and 13.6{\%}, respectively. Remarkably, these enhancements are comparable to those achieved by mono-lingual instruction tuning and even surpass them in some tasks. The result underscores the significance of relevant data acquisition across languages over linguistic congruence with unseen tasks during instruction tuning.
|
[
"Han, Janghoon",
"Lee, Changho",
"Shin, Joongbo",
"Choi, Stanley Jungkyu",
"Lee, Honglak",
"Bae, Kyunghoon"
] |
Deep Exploration of Cross-Lingual Zero-Shot Generalization in Instruction Tuning
|
findings-acl.912
|
Poster
|
2406.08796v1
|
https://aclanthology.org/2024.findings-acl.913.bib
|
@inproceedings{asami-sugawara-2024-makes,
title = "What Makes Language Models Good-enough?",
author = "Asami, Daiki and
Sugawara, Saku",
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.913",
pages = "15453--15467",
abstract = "Psycholinguistic research suggests that humans may build a representation of linguistic input that is {`}good-enough{'} for the task at hand. This study examines what architectural features make language models learn human-like good-enough language processing. We focus on the number of layers and self-attention heads in Transformers. We create a good-enough language processing (GELP) evaluation dataset (7,680 examples), which is designed to test the effects of two plausibility types, eight construction types, and three degrees of memory cost on language processing. To annotate GELP, we first conduct a crowdsourcing experiment whose design follows prior psycholinguistic studies. Our model evaluation against the annotated GELP then reveals that the full model as well as models with fewer layers and/or self-attention heads exhibit a good-enough performance. This result suggests that models with shallower depth and fewer heads can learn good-enough language processing.",
}
|
Psycholinguistic research suggests that humans may build a representation of linguistic input that is {`}good-enough{'} for the task at hand. This study examines what architectural features make language models learn human-like good-enough language processing. We focus on the number of layers and self-attention heads in Transformers. We create a good-enough language processing (GELP) evaluation dataset (7,680 examples), which is designed to test the effects of two plausibility types, eight construction types, and three degrees of memory cost on language processing. To annotate GELP, we first conduct a crowdsourcing experiment whose design follows prior psycholinguistic studies. Our model evaluation against the annotated GELP then reveals that the full model as well as models with fewer layers and/or self-attention heads exhibit a good-enough performance. This result suggests that models with shallower depth and fewer heads can learn good-enough language processing.
|
[
"Asami, Daiki",
"Sugawara, Saku"
] |
What Makes Language Models Good-enough?
|
findings-acl.913
|
Poster
|
1908.10747v1
|
https://aclanthology.org/2024.findings-acl.914.bib
|
@inproceedings{yu-etal-2024-refining,
title = "Refining Corpora from a Model Calibration Perspective for {C}hinese Spelling Correction",
author = "Yu, Dingyao and
An, Yang and
Ye, Wei and
Xiao, Xiongfeng and
Mao, Shaoguang and
Ge, Tao and
Zhang, Shikun",
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.914",
pages = "15468--15480",
abstract = "Chinese Spelling Correction (CSC) commonly lacks large-scale high-quality corpora, due to the labor-intensive labeling of spelling errors in real-life human writing or typing scenarios. Two data augmentation methods are widely adopted: (1) *Random Replacement* with the guidance of confusion sets and (2) *OCR/ASR-based Generation* that simulates character misusing. However, both methods inevitably introduce noisy data (e.g., false spelling errors), potentially leading to over-correction. By carefully analyzing the two types of corpora, we find that though the latter achieves more robust generalization performance, the former yields better-calibrated CSC models. We then provide a theoretical analysis of this empirical observation, based on which a corpus refining strategy is proposed. Specifically, OCR/ASR-based data samples are fed into a well-calibrated CSC model trained on random replacement-based corpora and then filtered based on prediction confidence. By learning a simple BERT-based model on the refined OCR/ASR-based corpus, we set up impressive state-of-the-art performance on three widely-used benchmarks, while significantly alleviating over-correction (e.g., lowering false positive predictions).",
}
|
Chinese Spelling Correction (CSC) commonly lacks large-scale high-quality corpora, due to the labor-intensive labeling of spelling errors in real-life human writing or typing scenarios. Two data augmentation methods are widely adopted: (1) *Random Replacement* with the guidance of confusion sets and (2) *OCR/ASR-based Generation* that simulates character misusing. However, both methods inevitably introduce noisy data (e.g., false spelling errors), potentially leading to over-correction. By carefully analyzing the two types of corpora, we find that though the latter achieves more robust generalization performance, the former yields better-calibrated CSC models. We then provide a theoretical analysis of this empirical observation, based on which a corpus refining strategy is proposed. Specifically, OCR/ASR-based data samples are fed into a well-calibrated CSC model trained on random replacement-based corpora and then filtered based on prediction confidence. By learning a simple BERT-based model on the refined OCR/ASR-based corpus, we set up impressive state-of-the-art performance on three widely-used benchmarks, while significantly alleviating over-correction (e.g., lowering false positive predictions).
|
[
"Yu, Dingyao",
"An, Yang",
"Ye, Wei",
"Xiao, Xiongfeng",
"Mao, Shaoguang",
"Ge, Tao",
"Zhang, Shikun"
] |
Refining Corpora from a Model Calibration Perspective for {C}hinese Spelling Correction
|
findings-acl.914
|
Poster
|
2407.15498v1
|
https://aclanthology.org/2024.findings-acl.915.bib
|
@inproceedings{zhang-etal-2024-countercurate,
title = "{C}ounter{C}urate: Enhancing Physical and Semantic Visio-Linguistic Compositional Reasoning via Counterfactual Examples",
author = "Zhang, Jianrui and
Cai, Mu and
Xie, Tengyang and
Lee, Yong Jae",
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.915",
pages = "15481--15495",
abstract = "We propose CounterCurate, a framework to comprehensively improve the visio-linguistic compositional reasoning capability for both contrastive and generative multimodal models. In particular, we identify two critical under- explored problems: the neglect of physically grounded reasoning (counting and position understanding) and the potential of using highly capable text and image generation models for semantic counterfactual fine-tuning. Our work pioneers an approach in addressing these gaps.We first spotlight the near-chance performance of multimodal models like CLIP and LLaVA in physically grounded compositional reasoning. We then apply simple data augmentation using the grounded image generation model GLIGEN to generate fine-tuning data, resulting in significant performance improvements: +33{\%} and +37{\%} for CLIP and LLaVA, respectively, on our newly curated Flickr30k-Positions benchmark. Moreover, we exploit the capabilities of high-performing text generation and image generation models, specifically GPT-4V and DALLE-3, to curate challenging semantic counterfactuals, thereby further enhancing compositional reasoning capabilities on benchmarks such as SugarCrepe, where CounterCurate outperforms GPT-4V.To facilitate future research, we release ourcode, dataset, benchmark, and checkpoints at https://countercurate.github.io/",
}
|
We propose CounterCurate, a framework to comprehensively improve the visio-linguistic compositional reasoning capability for both contrastive and generative multimodal models. In particular, we identify two critical under- explored problems: the neglect of physically grounded reasoning (counting and position understanding) and the potential of using highly capable text and image generation models for semantic counterfactual fine-tuning. Our work pioneers an approach in addressing these gaps.We first spotlight the near-chance performance of multimodal models like CLIP and LLaVA in physically grounded compositional reasoning. We then apply simple data augmentation using the grounded image generation model GLIGEN to generate fine-tuning data, resulting in significant performance improvements: +33{\%} and +37{\%} for CLIP and LLaVA, respectively, on our newly curated Flickr30k-Positions benchmark. Moreover, we exploit the capabilities of high-performing text generation and image generation models, specifically GPT-4V and DALLE-3, to curate challenging semantic counterfactuals, thereby further enhancing compositional reasoning capabilities on benchmarks such as SugarCrepe, where CounterCurate outperforms GPT-4V.To facilitate future research, we release ourcode, dataset, benchmark, and checkpoints at https://countercurate.github.io/
|
[
"Zhang, Jianrui",
"Cai, Mu",
"Xie, Tengyang",
"Lee, Yong Jae"
] |
{C}ounter{C}urate: Enhancing Physical and Semantic Visio-Linguistic Compositional Reasoning via Counterfactual Examples
|
findings-acl.915
|
Poster
|
2402.13254v4
|
https://aclanthology.org/2024.findings-acl.916.bib
|
@inproceedings{xu-etal-2024-knowledge,
title = "Knowledge-Infused Prompting: Assessing and Advancing Clinical Text Data Generation with Large Language Models",
author = "Xu, Ran and
Cui, Hejie and
Yu, Yue and
Kan, Xuan and
Shi, Wenqi and
Zhuang, Yuchen and
Wang, May Dongmei and
Jin, Wei and
Ho, Joyce and
Yang, Carl",
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.916",
pages = "15496--15523",
abstract = "Clinical natural language processing faces challenges like complex medical terminology and clinical contexts. Recently, large language models (LLMs) have shown promise in this domain. Yet, their direct deployment can lead to privacy issues and are constrained by resources. To address this challenge, we delve into synthetic clinical text generation with LLMs for clinical NLP tasks. We propose an innovative, resource-efficient approach, ClinGen, which infuses knowledge into the process. Our model involves clinical knowledge extraction and context-informed LLM prompting. Both clinical topics and writing styles are drawn from external domain-specific knowledge graphs and LLMs to guide data generation. Our extensive empirical study across 8 clinical NLP tasks and 18 datasets reveals that ClinGen consistently enhances performance across various tasks by 7.7{\%}-8.7{\%} on average, effectively aligning the distribution of real datasets and enriching the diversity of generated training instances.",
}
|
Clinical natural language processing faces challenges like complex medical terminology and clinical contexts. Recently, large language models (LLMs) have shown promise in this domain. Yet, their direct deployment can lead to privacy issues and are constrained by resources. To address this challenge, we delve into synthetic clinical text generation with LLMs for clinical NLP tasks. We propose an innovative, resource-efficient approach, ClinGen, which infuses knowledge into the process. Our model involves clinical knowledge extraction and context-informed LLM prompting. Both clinical topics and writing styles are drawn from external domain-specific knowledge graphs and LLMs to guide data generation. Our extensive empirical study across 8 clinical NLP tasks and 18 datasets reveals that ClinGen consistently enhances performance across various tasks by 7.7{\%}-8.7{\%} on average, effectively aligning the distribution of real datasets and enriching the diversity of generated training instances.
|
[
"Xu, Ran",
"Cui, Hejie",
"Yu, Yue",
"Kan, Xuan",
"Shi, Wenqi",
"Zhuang, Yuchen",
"Wang, May Dongmei",
"Jin, Wei",
"Ho, Joyce",
"Yang, Carl"
] |
Knowledge-Infused Prompting: Assessing and Advancing Clinical Text Data Generation with Large Language Models
|
findings-acl.916
|
Poster
|
2311.00287v1
|
https://aclanthology.org/2024.findings-acl.917.bib
|
@inproceedings{hwang-etal-2024-textless,
title = "Textless Acoustic Model with Self-Supervised Distillation for Noise-Robust Expressive Speech-to-Speech Translation",
author = "Hwang, Min-Jae and
Kulikov, Ilia and
Peloquin, Benjamin and
Gong, Hongyu and
Chen, Peng-Jen and
Lee, Ann",
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.917",
pages = "15524--15541",
abstract = "In this paper, we propose a textless acoustic model with a self-supervised distillation strategy for noise-robust expressive speech-to-speech translation (S2ST).Recently proposed expressive S2ST systems have achieved impressive expressivity preservation performances by cascading unit-to-speech (U2S) generator to the speech-to-unit translation model. However, these systems are vulnerable to the presence of noise in input speech, which is an assumption in real-world translation scenarios. To address this limitation, we propose a U2S generator that incorporates a distillation with no label (DINO) self-supervised training strategy into it{'}s pretraining process.Because the proposed method captures noise-agnostic expressivity representation, it can generate qualified speech even in noisy environment.Objective and subjective evaluation results verified that the proposed method significantly improved the performance of the expressive S2ST system in noisy environments while maintaining competitive performance in clean environments.",
}
|
In this paper, we propose a textless acoustic model with a self-supervised distillation strategy for noise-robust expressive speech-to-speech translation (S2ST).Recently proposed expressive S2ST systems have achieved impressive expressivity preservation performances by cascading unit-to-speech (U2S) generator to the speech-to-unit translation model. However, these systems are vulnerable to the presence of noise in input speech, which is an assumption in real-world translation scenarios. To address this limitation, we propose a U2S generator that incorporates a distillation with no label (DINO) self-supervised training strategy into it{'}s pretraining process.Because the proposed method captures noise-agnostic expressivity representation, it can generate qualified speech even in noisy environment.Objective and subjective evaluation results verified that the proposed method significantly improved the performance of the expressive S2ST system in noisy environments while maintaining competitive performance in clean environments.
|
[
"Hwang, Min-Jae",
"Kulikov, Ilia",
"Peloquin, Benjamin",
"Gong, Hongyu",
"Chen, Peng-Jen",
"Lee, Ann"
] |
Textless Acoustic Model with Self-Supervised Distillation for Noise-Robust Expressive Speech-to-Speech Translation
|
findings-acl.917
|
Poster
|
2406.02733v1
|
https://aclanthology.org/2024.findings-acl.918.bib
|
@inproceedings{wu-etal-2024-knowledge,
title = "Knowledge-Infused Legal Wisdom: Navigating {LLM} Consultation through the Lens of Diagnostics and Positive-Unlabeled Reinforcement Learning",
author = "Wu, Yang and
Wang, Chenghao and
Gumusel, Ece and
Liu, Xiaozhong",
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.918",
pages = "15542--15555",
abstract = "The integration of generative Large Language Models (LLMs) into various applications, including the legal domain, has been accelerated by their expansive and versatile nature. However, when facing a legal case, users without a legal background often struggle to formulate professional queries and may inadvertently overlook critical legal factors when presenting their case narrative to LLMs. To address this issue, we propose the Diagnostic Legal Large Language Model (D3LM), which utilizes adaptive lawyer-like diagnostic questions to collect additional case information and then provides high-quality feedback. D3LM incorporates an innovative graph-based Positive-Unlabeled Reinforcement Learning (PURL) algorithm, enabling the generation of critical questions and enhancing user-LLM interactions. Moreover, an integrated LLM-based stopping criterion facilitates precise Court Views Generation (CVG). Our research also introduces a new English-language CVG dataset based on the US case law database, enriching the realm of LLM research and deployment with a vital dimension. D3LM surpasses classical LLMs by delivering outstanding performance and a remarkable user experience in the legal domain.",
}
|
The integration of generative Large Language Models (LLMs) into various applications, including the legal domain, has been accelerated by their expansive and versatile nature. However, when facing a legal case, users without a legal background often struggle to formulate professional queries and may inadvertently overlook critical legal factors when presenting their case narrative to LLMs. To address this issue, we propose the Diagnostic Legal Large Language Model (D3LM), which utilizes adaptive lawyer-like diagnostic questions to collect additional case information and then provides high-quality feedback. D3LM incorporates an innovative graph-based Positive-Unlabeled Reinforcement Learning (PURL) algorithm, enabling the generation of critical questions and enhancing user-LLM interactions. Moreover, an integrated LLM-based stopping criterion facilitates precise Court Views Generation (CVG). Our research also introduces a new English-language CVG dataset based on the US case law database, enriching the realm of LLM research and deployment with a vital dimension. D3LM surpasses classical LLMs by delivering outstanding performance and a remarkable user experience in the legal domain.
|
[
"Wu, Yang",
"Wang, Chenghao",
"Gumusel, Ece",
"Liu, Xiaozhong"
] |
Knowledge-Infused Legal Wisdom: Navigating {LLM} Consultation through the Lens of Diagnostics and Positive-Unlabeled Reinforcement Learning
|
findings-acl.918
|
Poster
|
2406.03600v1
|
https://aclanthology.org/2024.findings-acl.919.bib
|
@inproceedings{liu-etal-2024-teller,
title = "{TELLER}: A Trustworthy Framework for Explainable, Generalizable and Controllable Fake News Detection",
author = "Liu, Hui and
Wang, Wenya and
Li, Haoru and
Li, Haoliang",
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.919",
pages = "15556--15583",
abstract = "The proliferation of fake news has emerged as a severe societal problem, raising significant interest from industry and academia. While existing deep-learning based methods have made progress in detecting fake news accurately, their reliability may be compromised caused by the non-transparent reasoning processes, poor generalization abilities and inherent risks of integration with large language models (LLMs). To address this challenge, we propose TELLER, a novel framework for trustworthy fake news detection that prioritizes explainability, generalizability and controllability of models. This is achieved via a dual-system framework that integrates cognition and decision systems, adhering to the principles above. The cognition system harnesses human expertise to generate logical predicates, which guide LLMs in generating human-readable logic atoms. Meanwhile, the decision system deduces generalizable logic rules to aggregate these atoms, enabling the identification of the truthfulness of the input news across diverse domains and enhancing transparency in the decision-making process. Finally, we present comprehensive evaluation results on four datasets, demonstrating the feasibility and trustworthiness of our proposed framework.",
}
|
The proliferation of fake news has emerged as a severe societal problem, raising significant interest from industry and academia. While existing deep-learning based methods have made progress in detecting fake news accurately, their reliability may be compromised caused by the non-transparent reasoning processes, poor generalization abilities and inherent risks of integration with large language models (LLMs). To address this challenge, we propose TELLER, a novel framework for trustworthy fake news detection that prioritizes explainability, generalizability and controllability of models. This is achieved via a dual-system framework that integrates cognition and decision systems, adhering to the principles above. The cognition system harnesses human expertise to generate logical predicates, which guide LLMs in generating human-readable logic atoms. Meanwhile, the decision system deduces generalizable logic rules to aggregate these atoms, enabling the identification of the truthfulness of the input news across diverse domains and enhancing transparency in the decision-making process. Finally, we present comprehensive evaluation results on four datasets, demonstrating the feasibility and trustworthiness of our proposed framework.
|
[
"Liu, Hui",
"Wang, Wenya",
"Li, Haoru",
"Li, Haoliang"
] |
{TELLER}: A Trustworthy Framework for Explainable, Generalizable and Controllable Fake News Detection
|
findings-acl.919
|
Poster
|
2402.07776v2
|
https://aclanthology.org/2024.findings-acl.920.bib
|
@inproceedings{cao-wang-2024-verifiable,
title = "Verifiable Generation with Subsentence-Level Fine-Grained Citations",
author = "Cao, Shuyang and
Wang, Lu",
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.920",
pages = "15584--15596",
abstract = "Verifiable generation requires large language models (LLMs) to cite source documents supporting their outputs, thereby improve output transparency and trustworthiness. Yet, previous work mainly targets the generation of sentence-level citations, lacking specificity about which parts of a sentence are backed by the cited sources. This work studies verifiable generation with subsentence-level fine-grained citations for more precise location of generated content supported by the cited sources. We first present a dataset, SCiFi, comprising 10K Wikipedia paragraphs with subsentence-level citations. Each paragraph is paired with a set of candidate source documents for citation and a query that triggers the generation of the paragraph content. On SCiFi, we evaluate the performance of state-of-the-art LLMs and strategies for processing long documents designed for these models. Our experiment results reveals key factors that could enhance the quality of citations, including the expansion of the source documents{'} context accessible to the models and the implementation of specialized model tuning.",
}
|
Verifiable generation requires large language models (LLMs) to cite source documents supporting their outputs, thereby improve output transparency and trustworthiness. Yet, previous work mainly targets the generation of sentence-level citations, lacking specificity about which parts of a sentence are backed by the cited sources. This work studies verifiable generation with subsentence-level fine-grained citations for more precise location of generated content supported by the cited sources. We first present a dataset, SCiFi, comprising 10K Wikipedia paragraphs with subsentence-level citations. Each paragraph is paired with a set of candidate source documents for citation and a query that triggers the generation of the paragraph content. On SCiFi, we evaluate the performance of state-of-the-art LLMs and strategies for processing long documents designed for these models. Our experiment results reveals key factors that could enhance the quality of citations, including the expansion of the source documents{'} context accessible to the models and the implementation of specialized model tuning.
|
[
"Cao, Shuyang",
"Wang, Lu"
] |
Verifiable Generation with Subsentence-Level Fine-Grained Citations
|
findings-acl.920
|
Poster
|
2406.13375v1
|
https://aclanthology.org/2024.findings-acl.921.bib
|
@inproceedings{lal-etal-2024-tailoring,
title = "Tailoring with Targeted Precision: Edit-Based Agents for Open-Domain Procedure Customization",
author = "Lal, Yash Kumar and
Zhang, Li and
Brahman, Faeze and
Majumder, Bodhisattwa Prasad and
Clark, Peter and
Tandon, Niket",
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.921",
pages = "15597--15611",
abstract = "How-to procedures, such as how to plant a garden, are now used by millions of users, but sometimes need customizing to meet a user{'}s specific needs, e.g., planting a garden without pesticides. Our goal is to measure and improve an LLM{'}s ability to perform such customization. Our approach is to test several simple multi-LLM-agent architectures for customization, as well as an end-to-end LLM, using a new evaluation set, called CustomPlans, of over 200 WikiHow procedures each with a customization need. We find that a simple architecture with two LLM agents used sequentially performs best, one that edits a generic how-to procedure and one that verifies its executability, significantly outperforming (10.5{\%} absolute) an end-to-end prompted LLM. This suggests that LLMs can be configured reasonably effectively for procedure customization. This also suggests that multi-agent editing architectures may be worth exploring further for other customization applications (e.g. coding, creative writing) in the future.",
}
|
How-to procedures, such as how to plant a garden, are now used by millions of users, but sometimes need customizing to meet a user{'}s specific needs, e.g., planting a garden without pesticides. Our goal is to measure and improve an LLM{'}s ability to perform such customization. Our approach is to test several simple multi-LLM-agent architectures for customization, as well as an end-to-end LLM, using a new evaluation set, called CustomPlans, of over 200 WikiHow procedures each with a customization need. We find that a simple architecture with two LLM agents used sequentially performs best, one that edits a generic how-to procedure and one that verifies its executability, significantly outperforming (10.5{\%} absolute) an end-to-end prompted LLM. This suggests that LLMs can be configured reasonably effectively for procedure customization. This also suggests that multi-agent editing architectures may be worth exploring further for other customization applications (e.g. coding, creative writing) in the future.
|
[
"Lal, Yash Kumar",
"Zhang, Li",
"Brahman, Faeze",
"Majumder, Bodhisattwa Prasad",
"Clark, Peter",
"T",
"on, Niket"
] |
Tailoring with Targeted Precision: Edit-Based Agents for Open-Domain Procedure Customization
|
findings-acl.921
|
Poster
|
2311.09510v3
|
https://aclanthology.org/2024.findings-acl.922.bib
|
@inproceedings{wu-varshney-2024-meta,
title = "A Meta-Learning Perspective on Transformers for Causal Language Modeling",
author = "Wu, Xinbo and
Varshney, Lav",
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.922",
pages = "15612--15622",
abstract = "The Transformer architecture has become prominent in developing large causal language models. However, mechanisms to explain its capabilities are not well understood. Focused on the training process, here we establish a meta-learning view of the Transformer architecture when trained for the causal language modeling task, by explicating an inner optimization process that may happen within the Transformer. Further, from within the inner optimization, we discover and theoretically analyze a special characteristic of the norms of learned token representations within Transformer-based causal language models. Our analysis is supported by experiments conducted on pre-trained large language models and real-world data.",
}
|
The Transformer architecture has become prominent in developing large causal language models. However, mechanisms to explain its capabilities are not well understood. Focused on the training process, here we establish a meta-learning view of the Transformer architecture when trained for the causal language modeling task, by explicating an inner optimization process that may happen within the Transformer. Further, from within the inner optimization, we discover and theoretically analyze a special characteristic of the norms of learned token representations within Transformer-based causal language models. Our analysis is supported by experiments conducted on pre-trained large language models and real-world data.
|
[
"Wu, Xinbo",
"Varshney, Lav"
] |
A Meta-Learning Perspective on Transformers for Causal Language Modeling
|
findings-acl.922
|
Poster
|
2310.05884v2
|
https://aclanthology.org/2024.findings-acl.923.bib
|
@inproceedings{zhang-etal-2024-plad,
title = "{PL}a{D}: Preference-based Large Language Model Distillation with Pseudo-Preference Pairs",
author = "Zhang, Rongzhi and
Shen, Jiaming and
Liu, Tianqi and
Wang, Haorui and
Qin, Zhen and
Han, Feng and
Liu, Jialu and
Baumgartner, Simon and
Bendersky, Michael and
Zhang, Chao",
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.923",
pages = "15623--15636",
abstract = "Large Language Models (LLMs) have exhibited impressive capabilities in various tasks, yet their vast parameter sizes restrict their applicability in resource-constrained settings. Knowledge distillation (KD) offers a viable solution by transferring expertise from large teacher models to compact student models. However, traditional KD techniques face specific challenges when applied to LLMs, including restricted access to LLM outputs, significant teacher-student capacity gaps, and the inherited mis-calibration issue. In this work, we present PLaD, a novel preference-based LLM distillation framework. PLaD exploits the teacher-student capacity discrepancy to generate pseudo-preference pairs where teacher outputs are preferred over student outputs. Then, PLaD leverages a ranking loss to re-calibrate the student{'}s estimation of sequence likelihood, which steers the student{'}s focus towards understanding the relative quality of outputs instead of simply imitating the teacher. PLaD bypasses the need for access to teacher LLM{'}s internal states, tackles the student{'}s expressivity limitations, and mitigates the student mis-calibration issue. Through extensive experiments on two sequence generation tasks and with various LLMs, we demonstrate the effectiveness of our proposed PLaD framework.",
}
|
Large Language Models (LLMs) have exhibited impressive capabilities in various tasks, yet their vast parameter sizes restrict their applicability in resource-constrained settings. Knowledge distillation (KD) offers a viable solution by transferring expertise from large teacher models to compact student models. However, traditional KD techniques face specific challenges when applied to LLMs, including restricted access to LLM outputs, significant teacher-student capacity gaps, and the inherited mis-calibration issue. In this work, we present PLaD, a novel preference-based LLM distillation framework. PLaD exploits the teacher-student capacity discrepancy to generate pseudo-preference pairs where teacher outputs are preferred over student outputs. Then, PLaD leverages a ranking loss to re-calibrate the student{'}s estimation of sequence likelihood, which steers the student{'}s focus towards understanding the relative quality of outputs instead of simply imitating the teacher. PLaD bypasses the need for access to teacher LLM{'}s internal states, tackles the student{'}s expressivity limitations, and mitigates the student mis-calibration issue. Through extensive experiments on two sequence generation tasks and with various LLMs, we demonstrate the effectiveness of our proposed PLaD framework.
|
[
"Zhang, Rongzhi",
"Shen, Jiaming",
"Liu, Tianqi",
"Wang, Haorui",
"Qin, Zhen",
"Han, Feng",
"Liu, Jialu",
"Baumgartner, Simon",
"Bendersky, Michael",
"Zhang, Chao"
] |
{PL}a{D}: Preference-based Large Language Model Distillation with Pseudo-Preference Pairs
|
findings-acl.923
|
Poster
|
2003.02877v3
|
https://aclanthology.org/2024.findings-acl.924.bib
|
@inproceedings{zhang-etal-2024-small,
title = "Small Language Models Need Strong Verifiers to Self-Correct Reasoning",
author = "Zhang, Yunxiang and
Khalifa, Muhammad and
Logeswaran, Lajanugen and
Kim, Jaekyeom and
Lee, Moontae and
Lee, Honglak and
Wang, Lu",
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.924",
pages = "15637--15653",
abstract = "Self-correction has emerged as a promising solution to boost the reasoning performance of large language models (LLMs), where LLMs refine their solutions using self-generated critiques that pinpoint the errors. This work explores whether small ($\leq 13$B) language models (LMs) have the ability of self-correction on reasoning tasks with minimal inputs from stronger LMs. We propose a novel pipeline that prompts smaller LMs to collect self-correction data that supports the training of self-refinement abilities. First, we leverage correct solutions to guide the model in critiquing their incorrect responses. Second, the generated critiques, after filtering, are used for supervised fine-tuning of the self-correcting reasoner through solution refinement. Our experimental results show improved self-correction abilities of two models on five datasets spanning math and commonsense reasoning, with notable performance gains when paired with a strong GPT-4-based verifier, though limitations are identified when using a weak self-verifier for determining when to correct.",
}
|
Self-correction has emerged as a promising solution to boost the reasoning performance of large language models (LLMs), where LLMs refine their solutions using self-generated critiques that pinpoint the errors. This work explores whether small ($\leq 13$B) language models (LMs) have the ability of self-correction on reasoning tasks with minimal inputs from stronger LMs. We propose a novel pipeline that prompts smaller LMs to collect self-correction data that supports the training of self-refinement abilities. First, we leverage correct solutions to guide the model in critiquing their incorrect responses. Second, the generated critiques, after filtering, are used for supervised fine-tuning of the self-correcting reasoner through solution refinement. Our experimental results show improved self-correction abilities of two models on five datasets spanning math and commonsense reasoning, with notable performance gains when paired with a strong GPT-4-based verifier, though limitations are identified when using a weak self-verifier for determining when to correct.
|
[
"Zhang, Yunxiang",
"Khalifa, Muhammad",
"Logeswaran, Lajanugen",
"Kim, Jaekyeom",
"Lee, Moontae",
"Lee, Honglak",
"Wang, Lu"
] |
Small Language Models Need Strong Verifiers to Self-Correct Reasoning
|
findings-acl.924
|
Poster
|
2404.17140v2
|
https://aclanthology.org/2024.findings-acl.925.bib
|
@inproceedings{zhang-etal-2024-hire,
title = "Hire a Linguist!: Learning Endangered Languages in {LLM}s with In-Context Linguistic Descriptions",
author = "Zhang, Kexun and
Choi, Yee and
Song, Zhenqiao and
He, Taiqi and
Wang, William Yang and
Li, Lei",
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.925",
pages = "15654--15669",
abstract = "How can large language models (LLMs) process and translate endangered languages? Many languages lack a large corpus to train a decent LLM; therefore existing LLMs rarely perform well in unseen, endangered languages. On the contrary, we observe that 2000 endangered languages, though without a large corpus, have a grammar book or a dictionary. We propose LingoLLM, a training-free approach to enable an LLM to process unseen languages that hardly occur in its pre-training. Our key insight is to demonstrate linguistic knowledge of an unseen language in an LLM{'}s prompt, including a dictionary, a grammar book, and morphologically analyzed input text. We implement LingoLLM on top of two models, GPT-4 and Mixtral, and evaluate their performance on 5 tasks across 8 endangered or low-resource languages. Our results show that LingoLLM elevates translation capability from GPT-4{'}s 0 to 10.5 BLEU for 10 language directions. Our findings demonstrate the tremendous value of linguistic knowledge in the age of LLMs for endangered languages. Our data, code, and model generations will be released to the public. Our data, code, and model generations can be found at \url{https://github.com/LLiLab/llm4endangeredlang}.",
}
|
How can large language models (LLMs) process and translate endangered languages? Many languages lack a large corpus to train a decent LLM; therefore existing LLMs rarely perform well in unseen, endangered languages. On the contrary, we observe that 2000 endangered languages, though without a large corpus, have a grammar book or a dictionary. We propose LingoLLM, a training-free approach to enable an LLM to process unseen languages that hardly occur in its pre-training. Our key insight is to demonstrate linguistic knowledge of an unseen language in an LLM{'}s prompt, including a dictionary, a grammar book, and morphologically analyzed input text. We implement LingoLLM on top of two models, GPT-4 and Mixtral, and evaluate their performance on 5 tasks across 8 endangered or low-resource languages. Our results show that LingoLLM elevates translation capability from GPT-4{'}s 0 to 10.5 BLEU for 10 language directions. Our findings demonstrate the tremendous value of linguistic knowledge in the age of LLMs for endangered languages. Our data, code, and model generations will be released to the public. Our data, code, and model generations can be found at \url{https://github.com/LLiLab/llm4endangeredlang}.
|
[
"Zhang, Kexun",
"Choi, Yee",
"Song, Zhenqiao",
"He, Taiqi",
"Wang, William Yang",
"Li, Lei"
] |
Hire a Linguist!: Learning Endangered Languages in {LLM}s with In-Context Linguistic Descriptions
|
findings-acl.925
|
Poster
|
2402.18025
|
https://aclanthology.org/2024.findings-acl.926.bib
|
@inproceedings{malik-etal-2024-tarzan,
title = "From Tarzan to {T}olkien: Controlling the Language Proficiency Level of {LLM}s for Content Generation",
author = "Malik, Ali and
Mayhew, Stephen and
Piech, Christopher and
Bicknell, Klinton",
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.926",
pages = "15670--15693",
abstract = "We study the problem of controlling the difficulty level of text generated by Large Language Models (LLMs) for contexts where end-users are not fully proficient, such as language learners. Using a novel framework, we evaluate the effectiveness of several key approaches for this task, including few-shot prompting, supervised finetuning, and reinforcement learning (RL), utilising both GPT-4 and open source alternatives like LLama2-7B and Mistral-7B.Our findings reveal a large performance gap between GPT-4 and the open source models when using prompt-based strategies. However, we show how to bridge this gap with a careful combination of finetuning and RL alignment. Our best model, CALM (CEFR-Aligned Language Model), surpasses the performance of GPT-4 and other strategies, at only a fraction of the cost. We further validate the quality of our results through a small-scale human study.",
}
|
We study the problem of controlling the difficulty level of text generated by Large Language Models (LLMs) for contexts where end-users are not fully proficient, such as language learners. Using a novel framework, we evaluate the effectiveness of several key approaches for this task, including few-shot prompting, supervised finetuning, and reinforcement learning (RL), utilising both GPT-4 and open source alternatives like LLama2-7B and Mistral-7B.Our findings reveal a large performance gap between GPT-4 and the open source models when using prompt-based strategies. However, we show how to bridge this gap with a careful combination of finetuning and RL alignment. Our best model, CALM (CEFR-Aligned Language Model), surpasses the performance of GPT-4 and other strategies, at only a fraction of the cost. We further validate the quality of our results through a small-scale human study.
|
[
"Malik, Ali",
"Mayhew, Stephen",
"Piech, Christopher",
"Bicknell, Klinton"
] |
From Tarzan to {T}olkien: Controlling the Language Proficiency Level of {LLM}s for Content Generation
|
findings-acl.926
|
Poster
|
2406.03030v1
|
https://aclanthology.org/2024.findings-acl.927.bib
|
@inproceedings{chehbouni-etal-2024-representational,
title = "From Representational Harms to Quality-of-Service Harms: A Case Study on Llama 2 Safety Safeguards",
author = "Chehbouni, Khaoula and
Roshan, Megha and
Ma, Emmanuel and
Wei, Futian and
Taik, Afaf and
Cheung, Jackie and
Farnadi, Golnoosh",
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.927",
pages = "15694--15710",
abstract = "Recent progress in large language models (LLMs) has led to their widespread adoption in various domains. However, these advancements have also introduced additional safety risks and raised concerns regarding their detrimental impact on already marginalized populations.Despite growing mitigation efforts to develop safety safeguards, such as supervised safety-oriented fine-tuning and leveraging safe reinforcement learning from human feedback, multiple concerns regarding the safety and ingrained biases in these models remain. Furthermore, previous work has demonstrated that models optimized for safety often display exaggerated safety behaviors, such as a tendency to refrain from responding to certain requests as a precautionary measure. As such, a clear trade-off between the helpfulness and safety of these models has been documented in the literature. In this paper, we further investigate the effectiveness of safety measures by evaluating models on already mitigated biases. Using the case of Llama 2 as an example, we illustrate how LLMs{'} safety responses can still encode harmful assumptions. To do so, we create a set of non-toxic prompts, which we then use to evaluate Llama models. Through our new taxonomy of LLMs responses to users, we observe that the safety/helpfulness trade-offs are more pronounced for certain demographic groups which can lead to different kinds of harms such as quality-of-service harms for marginalized populations.",
}
|
Recent progress in large language models (LLMs) has led to their widespread adoption in various domains. However, these advancements have also introduced additional safety risks and raised concerns regarding their detrimental impact on already marginalized populations.Despite growing mitigation efforts to develop safety safeguards, such as supervised safety-oriented fine-tuning and leveraging safe reinforcement learning from human feedback, multiple concerns regarding the safety and ingrained biases in these models remain. Furthermore, previous work has demonstrated that models optimized for safety often display exaggerated safety behaviors, such as a tendency to refrain from responding to certain requests as a precautionary measure. As such, a clear trade-off between the helpfulness and safety of these models has been documented in the literature. In this paper, we further investigate the effectiveness of safety measures by evaluating models on already mitigated biases. Using the case of Llama 2 as an example, we illustrate how LLMs{'} safety responses can still encode harmful assumptions. To do so, we create a set of non-toxic prompts, which we then use to evaluate Llama models. Through our new taxonomy of LLMs responses to users, we observe that the safety/helpfulness trade-offs are more pronounced for certain demographic groups which can lead to different kinds of harms such as quality-of-service harms for marginalized populations.
|
[
"Chehbouni, Khaoula",
"Roshan, Megha",
"Ma, Emmanuel",
"Wei, Futian",
"Taik, Afaf",
"Cheung, Jackie",
"Farnadi, Golnoosh"
] |
From Representational Harms to Quality-of-Service Harms: A Case Study on Llama 2 Safety Safeguards
|
findings-acl.927
|
Poster
|
2403.13213v4
|
https://aclanthology.org/2024.findings-acl.928.bib
|
@inproceedings{guo-etal-2024-ctooleval,
title = "{CT}ool{E}val: A {C}hinese Benchmark for {LLM}-Powered Agent Evaluation in Real-World {API} Interactions",
author = "Guo, Zishan and
Huang, Yufei and
Xiong, Deyi",
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.928",
pages = "15711--15724",
abstract = "Assessing the capabilities of large language models (LLMs) as agents in decision making and operational tasks is crucial for the development of LLM-as-agent service. We propose CToolEval, a benchmark designed to evaluate LLMs in the context of Chinese societal applications, featuring 398 APIs across 27 widely-used Apps (e.g., Apps for shopping, map, music, travel, etc.) that cover 14 domains. We further present an evaluation framework that simulates real-life scenarios, to facilitate the assessment of tool invocation ability of LLMs for tool learning and task completion ability for user interation. Our extensive experiments with CToolEval evaluate 11 LLMs, revealing that while GPT-3.5-turbo excels in tool invocation, Chinese LLMs usually struggle with issues like hallucination and a lack of comprehensive tool understanding. Our findings highlight the need for further refinement in decision-making capabilities of LLMs, offering insights into bridging the gap between current functionalities and agent-level performance. To promote further research for LLMs to fully act as reliable agents in complex, real-world situations, we release our data and codes at https://github.com/tjunlp-lab/CToolEval.",
}
|
Assessing the capabilities of large language models (LLMs) as agents in decision making and operational tasks is crucial for the development of LLM-as-agent service. We propose CToolEval, a benchmark designed to evaluate LLMs in the context of Chinese societal applications, featuring 398 APIs across 27 widely-used Apps (e.g., Apps for shopping, map, music, travel, etc.) that cover 14 domains. We further present an evaluation framework that simulates real-life scenarios, to facilitate the assessment of tool invocation ability of LLMs for tool learning and task completion ability for user interation. Our extensive experiments with CToolEval evaluate 11 LLMs, revealing that while GPT-3.5-turbo excels in tool invocation, Chinese LLMs usually struggle with issues like hallucination and a lack of comprehensive tool understanding. Our findings highlight the need for further refinement in decision-making capabilities of LLMs, offering insights into bridging the gap between current functionalities and agent-level performance. To promote further research for LLMs to fully act as reliable agents in complex, real-world situations, we release our data and codes at https://github.com/tjunlp-lab/CToolEval.
|
[
"Guo, Zishan",
"Huang, Yufei",
"Xiong, Deyi"
] |
{CT}ool{E}val: A {C}hinese Benchmark for {LLM}-Powered Agent Evaluation in Real-World {API} Interactions
|
findings-acl.928
|
Poster
|
2407.00132v2
|
https://aclanthology.org/2024.findings-acl.929.bib
|
@inproceedings{athiwaratkun-etal-2024-token,
title = "Token Alignment via Character Matching for Subword Completion",
author = "Athiwaratkun, Ben and
Wang, Shiqi and
Shang, Mingyue and
Tian, Yuchen and
Wang, Zijian and
Gonugondla, Sujan Kumar and
Gouda, Sanjay Krishna and
Kwiatkowski, Robert and
Nallapati, Ramesh and
Bhatia, Parminder and
Xiang, Bing",
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.929",
pages = "15725--15738",
abstract = "Generative models, widely utilized in various applications, can often struggle with prompts corresponding to partial tokens. This struggle stems from tokenization, where partial tokens fall out of distribution during inference, leading to incorrect or nonsensical outputs. This paper examines a technique to alleviate the tokenization artifact on text completion in generative models, maintaining performance even in regular non-subword cases. The method, termed token alignment, involves backtracking to the last complete tokens and ensuring the model{'}s generation aligns with the prompt. This approach showcases marked improvement across many partial token scenarios, including nuanced cases like space-prefix and partial indentation, with only a minor time increase. The technique and analysis detailed in this paper contribute to the continuous advancement of generative models in handling partial inputs, bearing relevance for applications like code completion and text.",
}
|
Generative models, widely utilized in various applications, can often struggle with prompts corresponding to partial tokens. This struggle stems from tokenization, where partial tokens fall out of distribution during inference, leading to incorrect or nonsensical outputs. This paper examines a technique to alleviate the tokenization artifact on text completion in generative models, maintaining performance even in regular non-subword cases. The method, termed token alignment, involves backtracking to the last complete tokens and ensuring the model{'}s generation aligns with the prompt. This approach showcases marked improvement across many partial token scenarios, including nuanced cases like space-prefix and partial indentation, with only a minor time increase. The technique and analysis detailed in this paper contribute to the continuous advancement of generative models in handling partial inputs, bearing relevance for applications like code completion and text.
|
[
"Athiwaratkun, Ben",
"Wang, Shiqi",
"Shang, Mingyue",
"Tian, Yuchen",
"Wang, Zijian",
"Gonugondla, Sujan Kumar",
"Gouda, Sanjay Krishna",
"Kwiatkowski, Robert",
"Nallapati, Ramesh",
"Bhatia, Parminder",
"Xiang, Bing"
] |
Token Alignment via Character Matching for Subword Completion
|
findings-acl.929
|
Poster
|
2403.08688v1
|
https://aclanthology.org/2024.findings-acl.930.bib
|
@inproceedings{han-etal-2024-rethinking,
title = "Rethinking Efficient Multilingual Text Summarization Meta-Evaluation",
author = "Han, Rilyn and
Chen, Jiawen and
Liu, Yixin and
Cohan, Arman",
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.930",
pages = "15739--15746",
abstract = "Evaluating multilingual summarization evaluation metrics, i.e., meta-evaluation, is challenging because of the difficulty of human annotation collection. Therefore, we investigate an efficient multilingual meta-evaluation framework that uses machine translation systems to transform a monolingual meta-evaluation dataset into multilingual versions. To this end, we introduce a statistical test to verify the transformed dataset quality by checking the meta-evaluation result consistency on the original dataset and back-translated dataset. With this quality verification method, we transform an existing English summarization meta-evaluation dataset, RoSE, into 30 languages, and conduct a multilingual meta-evaluation of several representative automatic evaluation metrics. In our meta-evaluation, we find that metric performance varies in different languages and neural metrics generally outperform classical text-matching-based metrics in non-English languages. Moreover, we identify a two-stage evaluation method with superior performance, which first translates multilingual texts into English and then performs evaluation. We make the transformed datasets publicly available to facilitate future research.",
}
|
Evaluating multilingual summarization evaluation metrics, i.e., meta-evaluation, is challenging because of the difficulty of human annotation collection. Therefore, we investigate an efficient multilingual meta-evaluation framework that uses machine translation systems to transform a monolingual meta-evaluation dataset into multilingual versions. To this end, we introduce a statistical test to verify the transformed dataset quality by checking the meta-evaluation result consistency on the original dataset and back-translated dataset. With this quality verification method, we transform an existing English summarization meta-evaluation dataset, RoSE, into 30 languages, and conduct a multilingual meta-evaluation of several representative automatic evaluation metrics. In our meta-evaluation, we find that metric performance varies in different languages and neural metrics generally outperform classical text-matching-based metrics in non-English languages. Moreover, we identify a two-stage evaluation method with superior performance, which first translates multilingual texts into English and then performs evaluation. We make the transformed datasets publicly available to facilitate future research.
|
[
"Han, Rilyn",
"Chen, Jiawen",
"Liu, Yixin",
"Cohan, Arman"
] |
Rethinking Efficient Multilingual Text Summarization Meta-Evaluation
|
findings-acl.930
|
Poster
|
2305.11129v2
|
https://aclanthology.org/2024.findings-acl.931.bib
|
@inproceedings{ma-etal-2024-emotion2vec,
title = "emotion2vec: Self-Supervised Pre-Training for Speech Emotion Representation",
author = "Ma, Ziyang and
Zheng, Zhisheng and
Ye, Jiaxin and
Li, Jinchao and
Gao, Zhifu and
Zhang, ShiLiang and
Chen, Xie",
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.931",
pages = "15747--15760",
abstract = "We propose emotion2vec, a universal speech emotion representation model. emotion2vec is pre-trained on open-source unlabeled emotion data through self-supervised online distillation, combining utterance-level loss and frame-level loss during pre-training. emotion2vec outperforms state-of-the-art pre-trained universal models and emotion specialist models by only training linear layers for the speech emotion recognition task on the mainstream IEMOCAP dataset. In addition, emotion2vec shows consistent improvements among 10 different languages of speech emotion recognition datasets. emotion2vec also shows excellent results on other emotion tasks, such as song emotion recognition, emotion prediction in conversation, and sentiment analysis. Comparison experiments, ablation experiments, and visualization comprehensively demonstrate the universal capability of the proposed emotion2vec. To the best of our knowledge, emotion2vec is the first universal representation model in various emotion-related tasks, filling a gap in the field.",
}
|
We propose emotion2vec, a universal speech emotion representation model. emotion2vec is pre-trained on open-source unlabeled emotion data through self-supervised online distillation, combining utterance-level loss and frame-level loss during pre-training. emotion2vec outperforms state-of-the-art pre-trained universal models and emotion specialist models by only training linear layers for the speech emotion recognition task on the mainstream IEMOCAP dataset. In addition, emotion2vec shows consistent improvements among 10 different languages of speech emotion recognition datasets. emotion2vec also shows excellent results on other emotion tasks, such as song emotion recognition, emotion prediction in conversation, and sentiment analysis. Comparison experiments, ablation experiments, and visualization comprehensively demonstrate the universal capability of the proposed emotion2vec. To the best of our knowledge, emotion2vec is the first universal representation model in various emotion-related tasks, filling a gap in the field.
|
[
"Ma, Ziyang",
"Zheng, Zhisheng",
"Ye, Jiaxin",
"Li, Jinchao",
"Gao, Zhifu",
"Zhang, ShiLiang",
"Chen, Xie"
] |
emotion2vec: Self-Supervised Pre-Training for Speech Emotion Representation
|
findings-acl.931
|
Poster
|
2011.08477v1
|
https://aclanthology.org/2024.findings-acl.932.bib
|
@inproceedings{yang-etal-2024-language-informed,
title = "Language-Informed Beam Search Decoding for Multilingual Machine Translation",
author = "Yang, Yilin and
Lee, Stefan and
Tadepalli, Prasad",
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.932",
pages = "15761--15772",
abstract = "Beam search decoding is the de-facto method for decoding auto-regressive Neural Machine Translation (NMT) models, including multilingual NMT where the target language is specified as an input. However, decoding multilingual NMT models commonly produces off-target translations {--} yielding translation outputs not in the intended language.In this paper, we first conduct an error analysis of off-target translations for a strong multilingual NMT model and identify how these decodings are produced during beam search. We then propose Language-informed Beam Search (LiBS), a general decoding algorithm incorporating an off-the-shelf Language Identification (LiD) model into beam search decoding to reduce off-target translations. LiBS is an inference-time procedure that is NMT-model agnostic and does not require any additional parallel data. Results show that our proposed LiBS algorithm on average improves +1.1 BLEU and +0.9 BLEU on WMT and OPUS datasets, and reduces off-target rates from 22.9{\%} to 7.7{\%} and 65.8{\%} to 25.3{\%} respectively.",
}
|
Beam search decoding is the de-facto method for decoding auto-regressive Neural Machine Translation (NMT) models, including multilingual NMT where the target language is specified as an input. However, decoding multilingual NMT models commonly produces off-target translations {--} yielding translation outputs not in the intended language.In this paper, we first conduct an error analysis of off-target translations for a strong multilingual NMT model and identify how these decodings are produced during beam search. We then propose Language-informed Beam Search (LiBS), a general decoding algorithm incorporating an off-the-shelf Language Identification (LiD) model into beam search decoding to reduce off-target translations. LiBS is an inference-time procedure that is NMT-model agnostic and does not require any additional parallel data. Results show that our proposed LiBS algorithm on average improves +1.1 BLEU and +0.9 BLEU on WMT and OPUS datasets, and reduces off-target rates from 22.9{\%} to 7.7{\%} and 65.8{\%} to 25.3{\%} respectively.
|
[
"Yang, Yilin",
"Lee, Stefan",
"Tadepalli, Prasad"
] |
Language-Informed Beam Search Decoding for Multilingual Machine Translation
|
findings-acl.932
|
Poster
|
2408.05738v1
|
https://aclanthology.org/2024.findings-acl.933.bib
|
@inproceedings{kim-etal-2024-ra,
title = "{RA}-{L}o{RA}: Rank-Adaptive Parameter-Efficient Fine-Tuning for Accurate 2-bit Quantized Large Language Models",
author = "Kim, Minsoo and
Lee, Sihwa and
Sung, Wonyong and
Choi, Jungwook",
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.933",
pages = "15773--15786",
abstract = "Deploying large language models (LLMs) with their extensive parameters and high memory demands challenges computational efficiency, particularly in fine-tuning for specific applications with limited resources. Techniques like Low-Rank Adaptation (LoRA) help by training a smaller, modifiable extension of the base model to reduce memory usage. However, combining quantization with LoRA, especially in low-bit scenarios, can lead to performance losses due to quantization errors. Our innovative Rank-Adaptive LoRA (RA-LoRA) addresses this by dynamically adjusting the adapter{'}s rank using rank-subspace analysis, optimizing performance with fewer parameters. We tested RA-LoRA on state-of-the-art LLMs for 2-bit efficient fine-tuning, showing it can improve model accuracy with minimal trainable parameters, marking a leap forward in quantization-aware fine-tuning methods and highlighting the significance of rank dynamics in optimizing quantized LLMs.",
}
|
Deploying large language models (LLMs) with their extensive parameters and high memory demands challenges computational efficiency, particularly in fine-tuning for specific applications with limited resources. Techniques like Low-Rank Adaptation (LoRA) help by training a smaller, modifiable extension of the base model to reduce memory usage. However, combining quantization with LoRA, especially in low-bit scenarios, can lead to performance losses due to quantization errors. Our innovative Rank-Adaptive LoRA (RA-LoRA) addresses this by dynamically adjusting the adapter{'}s rank using rank-subspace analysis, optimizing performance with fewer parameters. We tested RA-LoRA on state-of-the-art LLMs for 2-bit efficient fine-tuning, showing it can improve model accuracy with minimal trainable parameters, marking a leap forward in quantization-aware fine-tuning methods and highlighting the significance of rank dynamics in optimizing quantized LLMs.
|
[
"Kim, Minsoo",
"Lee, Sihwa",
"Sung, Wonyong",
"Choi, Jungwook"
] |
{RA}-{L}o{RA}: Rank-Adaptive Parameter-Efficient Fine-Tuning for Accurate 2-bit Quantized Large Language Models
|
findings-acl.933
|
Poster
|
2407.10032v1
|
https://aclanthology.org/2024.findings-acl.934.bib
|
@inproceedings{taylor-wang-2024-pgnsc,
title = "The {PGNSC} Benchmark: How Do We Predict Where Information Spreads?",
author = "Taylor, Alexander and
Wang, 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.934",
pages = "15787--15803",
abstract = "Social networks have become ideal vehicles for news dissemination because posted content is easily able to reach users beyond a news outlet{'}s direct audience. Understanding how information is transmitted among communities of users is a critical step towards understanding the impact social networks have on real-world events. Two significant barriers in this vein of work are identifying user clusters and meaningfully characterizing these communities. Thus, we propose the PGNSC benchmark, which builds information pathways based on the audiences of influential news sources and uses their content to characterize the communities. We present methods of aggregating these news-source-centric communities and for constructing the community feature representations that are used sequentially to construct information pathway prediction pipelines. Lastly, we perform extensive experiments to demonstrate the performance of baseline pipeline constructions and to highlight the possibilities for future work.",
}
|
Social networks have become ideal vehicles for news dissemination because posted content is easily able to reach users beyond a news outlet{'}s direct audience. Understanding how information is transmitted among communities of users is a critical step towards understanding the impact social networks have on real-world events. Two significant barriers in this vein of work are identifying user clusters and meaningfully characterizing these communities. Thus, we propose the PGNSC benchmark, which builds information pathways based on the audiences of influential news sources and uses their content to characterize the communities. We present methods of aggregating these news-source-centric communities and for constructing the community feature representations that are used sequentially to construct information pathway prediction pipelines. Lastly, we perform extensive experiments to demonstrate the performance of baseline pipeline constructions and to highlight the possibilities for future work.
|
[
"Taylor, Alex",
"er",
"Wang, Wei"
] |
The {PGNSC} Benchmark: How Do We Predict Where Information Spreads?
|
findings-acl.934
|
Poster
|
2103.02214v3
|
https://aclanthology.org/2024.findings-acl.935.bib
|
@inproceedings{basavatia-etal-2024-starling,
title = "{STARLING}: Self-supervised Training of Text-based Reinforcement Learning Agent with Large Language Models",
author = "Basavatia, Shreyas and
Murugesan, Keerthiram and
Ratnakar, Shivam",
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.935",
pages = "15804--15819",
abstract = "Interactive fiction games have emerged as an important application to improve the generalization capabilities of language-based reinforcement learning (RL) agents. Existing environments for interactive fiction games are domain-specific or time-consuming to generate and do not train the RL agents to master a specific set of skills. In this work, we introduce an interactive environment for self-supervised RL, STARLING, for text-based games that bootstraps the text-based RL agents with automatically generated games (based on the seed set of game ideas) to boost the performance and generalization capabilities to reach a goal of the target environment. These games let the agent hone their skills on a predefined set of tasks. We create and test an environment with 100 games, generated using this automated framework that uses large language models (GPT3) and an interactive fiction game engine (based on Inform7) to provide the user with the ability to generate more games under minimal human supervision. Experimental results based on both the human participants and baseline text-based RL agents reveal that current state-of-the-art text-based RL agents cannot use previously learned skills in new situations at the level humans can. These results enforce STARLING{'}s potential to serve as a sandbox environment for further research in self-supervised text-based RL.",
}
|
Interactive fiction games have emerged as an important application to improve the generalization capabilities of language-based reinforcement learning (RL) agents. Existing environments for interactive fiction games are domain-specific or time-consuming to generate and do not train the RL agents to master a specific set of skills. In this work, we introduce an interactive environment for self-supervised RL, STARLING, for text-based games that bootstraps the text-based RL agents with automatically generated games (based on the seed set of game ideas) to boost the performance and generalization capabilities to reach a goal of the target environment. These games let the agent hone their skills on a predefined set of tasks. We create and test an environment with 100 games, generated using this automated framework that uses large language models (GPT3) and an interactive fiction game engine (based on Inform7) to provide the user with the ability to generate more games under minimal human supervision. Experimental results based on both the human participants and baseline text-based RL agents reveal that current state-of-the-art text-based RL agents cannot use previously learned skills in new situations at the level humans can. These results enforce STARLING{'}s potential to serve as a sandbox environment for further research in self-supervised text-based RL.
|
[
"Basavatia, Shreyas",
"Murugesan, Keerthiram",
"Ratnakar, Shivam"
] |
{STARLING}: Self-supervised Training of Text-based Reinforcement Learning Agent with Large Language Models
|
findings-acl.935
|
Poster
|
2403.03141v1
|
https://aclanthology.org/2024.findings-acl.936.bib
|
@inproceedings{lee-etal-2024-protecting,
title = "Protecting Privacy Through Approximating Optimal Parameters for Sequence Unlearning in Language Models",
author = "Lee, Dohyun and
Rim, Daniel and
Choi, Minseok and
Choo, Jaegul",
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.936",
pages = "15820--15839",
abstract = "Although language models (LMs) demonstrate exceptional capabilities on various tasks, they are potentially vulnerable to extraction attacks, which represent a significant privacy risk.To mitigate the privacy concerns of LMs, machine unlearning has emerged as an important research area, which is utilized to induce the LM to selectively forget about some of its training data.While completely retraining the model will guarantee successful unlearning and privacy assurance, it is impractical for LMs, as it would be time-consuming and resource-intensive.Prior works efficiently unlearn the target token sequences, but upon subsequent iterations, the LM displays significant degradation in performance.In this work, we propose $\textbf{P}$rivacy Protection via $\textbf{O}$ptimal $\textbf{P}$arameters (POP), a novel unlearning method that effectively forgets the target token sequences from the pretrained LM by applying optimal gradient updates to the parameters.Inspired by the gradient derivation of complete retraining, we approximate the optimal training objective that successfully unlearns the target sequence while retaining the knowledge from the rest of the training data.Experimental results demonstrate that POP exhibits remarkable retention performance post-unlearning across 9 classification and 4 dialogue benchmarks, outperforming the state-of-the-art by a large margin.Furthermore, we introduce Remnant Memorization Accuracy that quantifies privacy risks based on token likelihood and validate its effectiveness through both qualitative and quantitative analyses.",
}
|
Although language models (LMs) demonstrate exceptional capabilities on various tasks, they are potentially vulnerable to extraction attacks, which represent a significant privacy risk.To mitigate the privacy concerns of LMs, machine unlearning has emerged as an important research area, which is utilized to induce the LM to selectively forget about some of its training data.While completely retraining the model will guarantee successful unlearning and privacy assurance, it is impractical for LMs, as it would be time-consuming and resource-intensive.Prior works efficiently unlearn the target token sequences, but upon subsequent iterations, the LM displays significant degradation in performance.In this work, we propose $\textbf{P}$rivacy Protection via $\textbf{O}$ptimal $\textbf{P}$arameters (POP), a novel unlearning method that effectively forgets the target token sequences from the pretrained LM by applying optimal gradient updates to the parameters.Inspired by the gradient derivation of complete retraining, we approximate the optimal training objective that successfully unlearns the target sequence while retaining the knowledge from the rest of the training data.Experimental results demonstrate that POP exhibits remarkable retention performance post-unlearning across 9 classification and 4 dialogue benchmarks, outperforming the state-of-the-art by a large margin.Furthermore, we introduce Remnant Memorization Accuracy that quantifies privacy risks based on token likelihood and validate its effectiveness through both qualitative and quantitative analyses.
|
[
"Lee, Dohyun",
"Rim, Daniel",
"Choi, Minseok",
"Choo, Jaegul"
] |
Protecting Privacy Through Approximating Optimal Parameters for Sequence Unlearning in Language Models
|
findings-acl.936
|
Poster
|
2406.14091v1
|
https://aclanthology.org/2024.findings-acl.937.bib
|
@inproceedings{wang-etal-2024-mitigating,
title = "Mitigating Hallucinations in Large Vision-Language Models with Instruction Contrastive Decoding",
author = "Wang, Xintong and
Pan, Jingheng and
Ding, Liang and
Biemann, Chris",
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.937",
pages = "15840--15853",
abstract = "Large Vision-Language Models (LVLMs) are increasingly adept at generating contextually detailed and coherent responses from visual inputs. However, their application in multimodal decision-making and open-ended generation is hindered by a notable rate of hallucinations, where generated text inaccurately represents the visual contents. To address this issue, this paper introduces the Instruction Contrastive Decoding (ICD) method, a novel approach designed to reduce hallucinations during LVLM inference. Our method is inspired by our observation that what we call disturbance instructions significantly exacerbate hallucinations in multimodal fusion modules. ICD contrasts distributions from standard and instruction disturbance, thereby increasing alignment uncertainty and effectively subtracting hallucinated concepts from the original distribution. Through comprehensive experiments on discriminative benchmarks (POPE and MME) and a generative benchmark (LLaVa-Bench), we demonstrate that ICD significantly mitigates both object-level and attribute-level hallucinations. Moreover, our method not only addresses hallucinations but also significantly enhances the general perception and recognition capabilities of LVLMs.",
}
|
Large Vision-Language Models (LVLMs) are increasingly adept at generating contextually detailed and coherent responses from visual inputs. However, their application in multimodal decision-making and open-ended generation is hindered by a notable rate of hallucinations, where generated text inaccurately represents the visual contents. To address this issue, this paper introduces the Instruction Contrastive Decoding (ICD) method, a novel approach designed to reduce hallucinations during LVLM inference. Our method is inspired by our observation that what we call disturbance instructions significantly exacerbate hallucinations in multimodal fusion modules. ICD contrasts distributions from standard and instruction disturbance, thereby increasing alignment uncertainty and effectively subtracting hallucinated concepts from the original distribution. Through comprehensive experiments on discriminative benchmarks (POPE and MME) and a generative benchmark (LLaVa-Bench), we demonstrate that ICD significantly mitigates both object-level and attribute-level hallucinations. Moreover, our method not only addresses hallucinations but also significantly enhances the general perception and recognition capabilities of LVLMs.
|
[
"Wang, Xintong",
"Pan, Jingheng",
"Ding, Liang",
"Biemann, Chris"
] |
Mitigating Hallucinations in Large Vision-Language Models with Instruction Contrastive Decoding
|
findings-acl.937
|
Poster
|
2403.18715v2
|
https://aclanthology.org/2024.findings-acl.938.bib
|
@inproceedings{wang-etal-2024-fine-tuning,
title = "Fine-tuning Language Models for Joint Rewriting and Completion of Code with Potential Bugs",
author = "Wang, Dingmin and
Zhao, Jinman and
Pei, Hengzhi and
Tan, Samson and
Zha, Sheng",
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.938",
pages = "15854--15868",
abstract = "Handling drafty partial code remains a notable challenge in real-time code suggestion applications. Previous work has demonstrated shortcomings of large language models of code (CodeLLMs) in completing partial code with potential bugs. In this study, we view partial code as implementation hints and fine-tune CodeLLMs to jointly rewrite and complete partial code into functional full programs. We explore two strategies: one-pass generation and multi-pass iterative refinement. We construct new training and testing datasets using semantic-altering code transformations and iterative self-generations.We conduct comprehensive experiments over three representative open-sourced CodeLLMs {--} InCoder, CodeGen, and StarCoder.Results show that CodeLLMs fine-tuned using our approach achieve superior pass rates compared to the previous baselines across existing and newly-created benchmarks, effectively handle both potentially buggy and clean code, and largely preserve the integrity of the original partial implementations. We further present findings on the properties of the potential bugs we tested and on the design choices of our methods.",
}
|
Handling drafty partial code remains a notable challenge in real-time code suggestion applications. Previous work has demonstrated shortcomings of large language models of code (CodeLLMs) in completing partial code with potential bugs. In this study, we view partial code as implementation hints and fine-tune CodeLLMs to jointly rewrite and complete partial code into functional full programs. We explore two strategies: one-pass generation and multi-pass iterative refinement. We construct new training and testing datasets using semantic-altering code transformations and iterative self-generations.We conduct comprehensive experiments over three representative open-sourced CodeLLMs {--} InCoder, CodeGen, and StarCoder.Results show that CodeLLMs fine-tuned using our approach achieve superior pass rates compared to the previous baselines across existing and newly-created benchmarks, effectively handle both potentially buggy and clean code, and largely preserve the integrity of the original partial implementations. We further present findings on the properties of the potential bugs we tested and on the design choices of our methods.
|
[
"Wang, Dingmin",
"Zhao, Jinman",
"Pei, Hengzhi",
"Tan, Samson",
"Zha, Sheng"
] |
Fine-tuning Language Models for Joint Rewriting and Completion of Code with Potential Bugs
|
findings-acl.938
|
Poster
|
2306.03438v2
|
https://aclanthology.org/2024.findings-acl.939.bib
|
@inproceedings{anand-etal-2024-critical,
title = "A Critical Study of What Code-{LLM}s (Do Not) Learn",
author = "Anand, Abhinav and
Verma, Shweta and
Narasimhan, Krishna and
Mezini, Mira",
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.939",
pages = "15869--15889",
abstract = "Large Language Models trained on code corpora (code-LLMs) have demonstrated impressive performance in various coding assistance tasks. However, despite their increased size and training dataset, code-LLMs still have limitations such as suggesting codes with syntactic errors, variable misuse etc. Some studies argue that code-LLMs perform well on coding tasks because they use self-attention and hidden representations to encode relations among input tokens. However, previous works have not studied what code properties are not encoded by code-LLMs. In this paper, we conduct a fine-grained analysis of attention maps and hidden representations of code-LLMs. Our study indicates that code-LLMs only encode relations among specific subsets of input tokens. Specifically, by categorizing input tokens into syntactic tokens and identifiers, we found that models encode relations among syntactic tokens and among identifiers, but they fail to encode relations between syntactic tokens and identifiers. We also found that fine-tuned models encode these relations poorly compared to their pre-trained counterparts. Additionally, larger models with billions of parameters encode significantly less information about code than models with only a few hundred million parameters.",
}
|
Large Language Models trained on code corpora (code-LLMs) have demonstrated impressive performance in various coding assistance tasks. However, despite their increased size and training dataset, code-LLMs still have limitations such as suggesting codes with syntactic errors, variable misuse etc. Some studies argue that code-LLMs perform well on coding tasks because they use self-attention and hidden representations to encode relations among input tokens. However, previous works have not studied what code properties are not encoded by code-LLMs. In this paper, we conduct a fine-grained analysis of attention maps and hidden representations of code-LLMs. Our study indicates that code-LLMs only encode relations among specific subsets of input tokens. Specifically, by categorizing input tokens into syntactic tokens and identifiers, we found that models encode relations among syntactic tokens and among identifiers, but they fail to encode relations between syntactic tokens and identifiers. We also found that fine-tuned models encode these relations poorly compared to their pre-trained counterparts. Additionally, larger models with billions of parameters encode significantly less information about code than models with only a few hundred million parameters.
|
[
"An",
", Abhinav",
"Verma, Shweta",
"Narasimhan, Krishna",
"Mezini, Mira"
] |
A Critical Study of What Code-{LLM}s (Do Not) Learn
|
findings-acl.939
|
Poster
|
2304.09406v1
|
https://aclanthology.org/2024.findings-acl.940.bib
|
@inproceedings{zhou-etal-2024-visual,
title = "Visual In-Context Learning for Large Vision-Language Models",
author = "Zhou, Yucheng and
Li, Xiang and
Wang, Qianning and
Shen, Jianbing",
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.940",
pages = "15890--15902",
abstract = "In Large Visual Language Models (LVLMs), the efficacy of In-Context Learning (ICL) remains limited by challenges in cross-modal interactions and representation disparities. To overcome these challenges, we introduce a novel Visual In-Context Learning (VICL) method comprising Visual Demonstration Retrieval, Intent-Oriented Image Summarization, and Intent-Oriented Demonstration Composition. Our approach retrieves images via {''}Retrieval {\&} Rerank{''} paradigm, summarises images with task intent and task-specific visual parsing, and composes language-based demonstrations that reduce token count and alleviate cross-modal interaction problem. Experimental evaluations on five visual reasoning datasets demonstrate the effectiveness of our method. Moreover, our extensive experiments leverage information flow analysis to elucidate the effectiveness of our method, and investigate the impact of length and position of demonstrations for LVLM. The use of in-context unlearning further shows promise in resetting specific model knowledge without retraining.",
}
|
In Large Visual Language Models (LVLMs), the efficacy of In-Context Learning (ICL) remains limited by challenges in cross-modal interactions and representation disparities. To overcome these challenges, we introduce a novel Visual In-Context Learning (VICL) method comprising Visual Demonstration Retrieval, Intent-Oriented Image Summarization, and Intent-Oriented Demonstration Composition. Our approach retrieves images via {''}Retrieval {\&} Rerank{''} paradigm, summarises images with task intent and task-specific visual parsing, and composes language-based demonstrations that reduce token count and alleviate cross-modal interaction problem. Experimental evaluations on five visual reasoning datasets demonstrate the effectiveness of our method. Moreover, our extensive experiments leverage information flow analysis to elucidate the effectiveness of our method, and investigate the impact of length and position of demonstrations for LVLM. The use of in-context unlearning further shows promise in resetting specific model knowledge without retraining.
|
[
"Zhou, Yucheng",
"Li, Xiang",
"Wang, Qianning",
"Shen, Jianbing"
] |
Visual In-Context Learning for Large Vision-Language Models
|
findings-acl.940
|
Poster
|
2305.09731v1
|
https://aclanthology.org/2024.findings-acl.941.bib
|
@inproceedings{cheng-etal-2024-scale,
title = "{SCALE}: Synergized Collaboration of Asymmetric Language Translation Engines",
author = "Cheng, Xin and
Wang, Xun and
Ge, Tao and
Chen, Si-Qing and
Wei, Furu and
Zhao, Dongyan 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.941",
pages = "15903--15918",
abstract = "In this paper, we introduce SCALE, a collaborative framework that connects a compact Specialized Translation Model (STM) and a general-purpose Large Language Model (LLM) as one unified translation engine. By introducing translation from STM into the triplet in-context demonstrations, SCALE unlocks refinement and pivoting ability of LLM, thus 1) mitigating language bias of LLMs and parallel data bias of STMs, 2) enhancing LLM speciality without sacrificing generality, and 3) facilitating continual learning in a LLM-tuning-free way.Our comprehensive experiments show that SCALE significantly outperforms both LLMs (GPT-4, GPT-3.5) and supervised models (NLLB, M2M) in either high-resource or challenging low-resource settings. Moreover SCALE shows great scalability by only updating the lightweight STM and witness consistent system improvement, an averaged 4 BLEURT score across 4 languages without tuning LLM. Interestingly, SCALE could also effectively exploit the existing language bias of LLMs by using an English-centric STM as a pivot to conduct translation between any language pairs, outperforming GPT-4 by an average of 6 COMET points across eight translation directions. Furthermore we provide an in-depth analysis of SCALE{'}s robustness, translation characteristics, latency costs and inherent language bias, providing solid foundation for future studies exploring the potential synergy between LLMs and more specialized models.",
}
|
In this paper, we introduce SCALE, a collaborative framework that connects a compact Specialized Translation Model (STM) and a general-purpose Large Language Model (LLM) as one unified translation engine. By introducing translation from STM into the triplet in-context demonstrations, SCALE unlocks refinement and pivoting ability of LLM, thus 1) mitigating language bias of LLMs and parallel data bias of STMs, 2) enhancing LLM speciality without sacrificing generality, and 3) facilitating continual learning in a LLM-tuning-free way.Our comprehensive experiments show that SCALE significantly outperforms both LLMs (GPT-4, GPT-3.5) and supervised models (NLLB, M2M) in either high-resource or challenging low-resource settings. Moreover SCALE shows great scalability by only updating the lightweight STM and witness consistent system improvement, an averaged 4 BLEURT score across 4 languages without tuning LLM. Interestingly, SCALE could also effectively exploit the existing language bias of LLMs by using an English-centric STM as a pivot to conduct translation between any language pairs, outperforming GPT-4 by an average of 6 COMET points across eight translation directions. Furthermore we provide an in-depth analysis of SCALE{'}s robustness, translation characteristics, latency costs and inherent language bias, providing solid foundation for future studies exploring the potential synergy between LLMs and more specialized models.
|
[
"Cheng, Xin",
"Wang, Xun",
"Ge, Tao",
"Chen, Si-Qing",
"Wei, Furu",
"Zhao, Dongyan",
"Yan, Rui"
] |
{SCALE}: Synergized Collaboration of Asymmetric Language Translation Engines
|
findings-acl.941
|
Poster
|
2309.17061v1
|
https://aclanthology.org/2024.findings-acl.942.bib
|
@inproceedings{naik-etal-2024-perspective,
title = "No perspective, no perception!! Perspective-aware Healthcare Answer Summarization",
author = "Naik, Gauri and
Chandakacherla, Sharad and
Yadav, Shweta and
Akhtar, Md Shad",
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.942",
pages = "15919--15932",
abstract = "Healthcare Community Question Answering (CQA) forums offer an accessible platform for individuals seeking information on various healthcare-related topics. People find such platforms suitable for self-disclosure, seeking medical opinions, finding simplified explanations for their medical conditions, and answering others{'} questions. However, answers on these forums are typically diverse and prone to off-topic discussions. It can be challenging for readers to sift through numerous answers and extract meaningful insights, making answer summarization a crucial task for CQA forums. While several efforts have been made to summarize the community answers, most of them are limited to the open domain and overlook the different perspectives offered by these answers. To address this problem, this paper proposes a novel task of perspective-specific answer summarization. We identify various perspectives, within healthcare-related responses and frame a perspective-driven abstractive summary covering all responses. To achieve this, we annotate 3167 CQA threads with 6193 perspective-aware summaries in our PUMA dataset. Further, we propose PLASMA, a prompt-driven controllable summarization model. To encapsulate the perspective-specific conditions, we design an energy-controlled loss function for the optimization. We also leverage the prefix tuner to learn the intricacies of the healthcare perspective summarization. Our evaluation against five baselines suggests the superior performance of PLASMA by a margin of {\textasciitilde}1.5 - 21{\%} improvement. We supplement our experiments with ablation and qualitative analysis.",
}
|
Healthcare Community Question Answering (CQA) forums offer an accessible platform for individuals seeking information on various healthcare-related topics. People find such platforms suitable for self-disclosure, seeking medical opinions, finding simplified explanations for their medical conditions, and answering others{'} questions. However, answers on these forums are typically diverse and prone to off-topic discussions. It can be challenging for readers to sift through numerous answers and extract meaningful insights, making answer summarization a crucial task for CQA forums. While several efforts have been made to summarize the community answers, most of them are limited to the open domain and overlook the different perspectives offered by these answers. To address this problem, this paper proposes a novel task of perspective-specific answer summarization. We identify various perspectives, within healthcare-related responses and frame a perspective-driven abstractive summary covering all responses. To achieve this, we annotate 3167 CQA threads with 6193 perspective-aware summaries in our PUMA dataset. Further, we propose PLASMA, a prompt-driven controllable summarization model. To encapsulate the perspective-specific conditions, we design an energy-controlled loss function for the optimization. We also leverage the prefix tuner to learn the intricacies of the healthcare perspective summarization. Our evaluation against five baselines suggests the superior performance of PLASMA by a margin of {\textasciitilde}1.5 - 21{\%} improvement. We supplement our experiments with ablation and qualitative analysis.
|
[
"Naik, Gauri",
"Ch",
"akacherla, Sharad",
"Yadav, Shweta",
"Akhtar, Md Shad"
] |
No perspective, no perception!! Perspective-aware Healthcare Answer Summarization
|
findings-acl.942
|
Poster
|
2406.08881
|
https://aclanthology.org/2024.findings-acl.943.bib
|
@inproceedings{shen-etal-2024-retrieval,
title = "Retrieval-Augmented Retrieval: Large Language Models are Strong Zero-Shot Retriever",
author = "Shen, Tao and
Long, Guodong and
Geng, Xiubo and
Tao, Chongyang and
Lei, Yibin and
Zhou, Tianyi and
Blumenstein, Michael and
Jiang, Daxin",
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.943",
pages = "15933--15946",
abstract = "We propose a simple method that applies a large language model (LLM) to large-scale retrieval in zero-shot scenarios. Our method, the Large language model as Retriever (LameR), is built upon no other neural models but an LLM in a retrieval-augmented retrieval fashion, while breaking brute-force combinations of retrievers with LLMs and lifting the performance of zero-shot retrieval to be very competitive on benchmark datasets. Essentially, we propose to augment a query with its potential answers by prompting LLMs with a composition of the query and the query{'}s in-domain candidates. The candidates, regardless of correct or wrong, are obtained by a vanilla retrieval procedure on the target collection. As a part of the prompts, they are likely to help LLM generate more precise answers by pattern imitation or candidate summarization. Even if all the candidates are wrong, the prompts at least make LLM aware of in-collection patterns and genres. Moreover, due to the low performance of a self-supervised retriever, the LLM-based query augmentation becomes less effective as the retriever bottlenecks the whole pipeline. Therefore, we propose to leverage a non-parametric lexicon-based method (e.g., BM25) as the retrieval module to capture query-document overlap in a literal fashion. As such, LameR makes the retrieval procedure transparent to the LLM, thus circumventing the bottleneck.",
}
|
We propose a simple method that applies a large language model (LLM) to large-scale retrieval in zero-shot scenarios. Our method, the Large language model as Retriever (LameR), is built upon no other neural models but an LLM in a retrieval-augmented retrieval fashion, while breaking brute-force combinations of retrievers with LLMs and lifting the performance of zero-shot retrieval to be very competitive on benchmark datasets. Essentially, we propose to augment a query with its potential answers by prompting LLMs with a composition of the query and the query{'}s in-domain candidates. The candidates, regardless of correct or wrong, are obtained by a vanilla retrieval procedure on the target collection. As a part of the prompts, they are likely to help LLM generate more precise answers by pattern imitation or candidate summarization. Even if all the candidates are wrong, the prompts at least make LLM aware of in-collection patterns and genres. Moreover, due to the low performance of a self-supervised retriever, the LLM-based query augmentation becomes less effective as the retriever bottlenecks the whole pipeline. Therefore, we propose to leverage a non-parametric lexicon-based method (e.g., BM25) as the retrieval module to capture query-document overlap in a literal fashion. As such, LameR makes the retrieval procedure transparent to the LLM, thus circumventing the bottleneck.
|
[
"Shen, Tao",
"Long, Guodong",
"Geng, Xiubo",
"Tao, Chongyang",
"Lei, Yibin",
"Zhou, Tianyi",
"Blumenstein, Michael",
"Jiang, Daxin"
] |
Retrieval-Augmented Retrieval: Large Language Models are Strong Zero-Shot Retriever
|
findings-acl.943
|
Poster
|
2306.13421v2
|
https://aclanthology.org/2024.findings-acl.944.bib
|
@inproceedings{nakov-etal-2024-survey,
title = "A Survey on Predicting the Factuality and the Bias of News Media",
author = "Nakov, Preslav and
An, Jisun and
Kwak, Haewoon and
Manzoor, Muhammad Arslan and
Mujahid, Zain and
Sencar, Husrev",
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.944",
pages = "15947--15962",
abstract = "The present level of proliferation of fake, biased, and propagandistic content online has made it impossible to fact-check every single suspicious claim or article, either manually or automatically. An increasing number of scholars are focusing on a coarser granularity, aiming to profile entire news outlets, which allows fast identification of potential {``}fake news{''} by checking the reliability of their source. Source factuality is also an important element of systems for automatic fact-checking and {``}fake news{''} detection, as they need to assess the reliability of the evidence they retrieve online. Political bias detection, which in the Western political landscape is about predicting left-center-right bias, is an equally important topic, which has experienced a similar shift toward profiling entire news outlets. Moreover, there is a clear connection between the two, as highly biased media are less likely to be factual; yet, the two problems have been addressed separately. In this survey, we review the state of the art on media profiling for factuality and bias, arguing for the need to model them jointly. We also shed light on some of the major challenges for modeling bias and factuality jointly. We further discuss interesting recent advances in using different information sources and modalities, which go beyond the text of the articles the target news outlet has published. Finally, we discuss current challenges and outline future research directions.",
}
|
The present level of proliferation of fake, biased, and propagandistic content online has made it impossible to fact-check every single suspicious claim or article, either manually or automatically. An increasing number of scholars are focusing on a coarser granularity, aiming to profile entire news outlets, which allows fast identification of potential {``}fake news{''} by checking the reliability of their source. Source factuality is also an important element of systems for automatic fact-checking and {``}fake news{''} detection, as they need to assess the reliability of the evidence they retrieve online. Political bias detection, which in the Western political landscape is about predicting left-center-right bias, is an equally important topic, which has experienced a similar shift toward profiling entire news outlets. Moreover, there is a clear connection between the two, as highly biased media are less likely to be factual; yet, the two problems have been addressed separately. In this survey, we review the state of the art on media profiling for factuality and bias, arguing for the need to model them jointly. We also shed light on some of the major challenges for modeling bias and factuality jointly. We further discuss interesting recent advances in using different information sources and modalities, which go beyond the text of the articles the target news outlet has published. Finally, we discuss current challenges and outline future research directions.
|
[
"Nakov, Preslav",
"An, Jisun",
"Kwak, Haewoon",
"Manzoor, Muhammad Arslan",
"Mujahid, Zain",
"Sencar, Husrev"
] |
A Survey on Predicting the Factuality and the Bias of News Media
|
findings-acl.944
|
Poster
|
2103.12506v1
|
https://aclanthology.org/2024.findings-acl.945.bib
|
@inproceedings{salama-etal-2024-semantic,
title = "Semantic Compression for Word and Sentence Embeddings using Discrete Wavelet Transform",
author = "Salama, Rana and
Youssef, Abdou and
Diab, Mona",
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.945",
pages = "15963--15977",
abstract = "Wavelet transforms, a powerful mathematical tool, have been widely used in different domains, including Signal and Image processing, to unravel intricate patterns, enhance data representation, and extract meaningful features from data. Tangible results from their application suggest that Wavelet transforms can be applied to NLP capturing a variety of linguistic and semantic properties.In this paper, we empirically leverage the application of Discrete Wavelet Transforms (DWT) to word and sentence embeddings. We aim to showcase the capabilities of DWT in analyzing embedding representations at different levels of resolution and compressing them while maintaining their overall quality.We assess the effectiveness of DWT embeddings on semantic similarity tasks to show how DWT can be used to consolidate important semantic information in an embedding vector. We show the efficacy of the proposed paradigm using different embedding models, including large language models, on downstream tasks. Our results show that DWT can reduce the dimensionality of embeddings by 50-93{\%} with almost no change in performance for semantic similarity tasks, while achieving superior accuracy in most downstream tasks. Our findings pave the way for applying DWT to improve NLP applications.",
}
|
Wavelet transforms, a powerful mathematical tool, have been widely used in different domains, including Signal and Image processing, to unravel intricate patterns, enhance data representation, and extract meaningful features from data. Tangible results from their application suggest that Wavelet transforms can be applied to NLP capturing a variety of linguistic and semantic properties.In this paper, we empirically leverage the application of Discrete Wavelet Transforms (DWT) to word and sentence embeddings. We aim to showcase the capabilities of DWT in analyzing embedding representations at different levels of resolution and compressing them while maintaining their overall quality.We assess the effectiveness of DWT embeddings on semantic similarity tasks to show how DWT can be used to consolidate important semantic information in an embedding vector. We show the efficacy of the proposed paradigm using different embedding models, including large language models, on downstream tasks. Our results show that DWT can reduce the dimensionality of embeddings by 50-93{\%} with almost no change in performance for semantic similarity tasks, while achieving superior accuracy in most downstream tasks. Our findings pave the way for applying DWT to improve NLP applications.
|
[
"Salama, Rana",
"Youssef, Abdou",
"Diab, Mona"
] |
Semantic Compression for Word and Sentence Embeddings using Discrete Wavelet Transform
|
findings-acl.945
|
Poster
|
1909.03104v2
|
https://aclanthology.org/2024.findings-acl.946.bib
|
@inproceedings{kim-etal-2024-improving-multi,
title = "Improving Multi-hop Logical Reasoning in Knowledge Graphs with Context-Aware Query Representation Learning",
author = "Kim, Jeonghoon and
Jung, Heesoo and
Jang, Hyeju and
Park, Hogun",
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.946",
pages = "15978--15991",
abstract = "Multi-hop logical reasoning on knowledge graphs is a pivotal task in natural language processing, with numerous approaches aiming to answer First-Order Logic (FOL) queries. Recent geometry (e.g., box, cone) and probability (e.g., beta distribution)-based methodologies have effectively addressed complex FOL queries. However, a common challenge across these methods lies in determining accurate geometric bounds or probability parameters for these queries. The challenge arises because existing methods rely on linear sequential operations within their computation graphs, overlooking the logical structure of the query and the relation-induced information that can be gleaned from the relations of the query, which we call the context of the query. To address the problem, we propose a model-agnostic methodology that enhances the effectiveness of existing multi-hop logical reasoning approaches by fully integrating the context of the FOL query graph. Our approach distinctively discerns (1) the structural context inherent to the query structure and (2) the relation-induced context unique to each node in the query graph as delineated in the corresponding knowledge graph. This dual-context paradigm helps nodes within a query graph attain refined internal representations throughout the multi-hop reasoning steps. Through experiments on two datasets, our method consistently enhances the three multi-hop reasoning foundation models, achieving performance improvements of up to 19.5{\%}. Our codes are available at https://github.com/kjh9503/caqr.",
}
|
Multi-hop logical reasoning on knowledge graphs is a pivotal task in natural language processing, with numerous approaches aiming to answer First-Order Logic (FOL) queries. Recent geometry (e.g., box, cone) and probability (e.g., beta distribution)-based methodologies have effectively addressed complex FOL queries. However, a common challenge across these methods lies in determining accurate geometric bounds or probability parameters for these queries. The challenge arises because existing methods rely on linear sequential operations within their computation graphs, overlooking the logical structure of the query and the relation-induced information that can be gleaned from the relations of the query, which we call the context of the query. To address the problem, we propose a model-agnostic methodology that enhances the effectiveness of existing multi-hop logical reasoning approaches by fully integrating the context of the FOL query graph. Our approach distinctively discerns (1) the structural context inherent to the query structure and (2) the relation-induced context unique to each node in the query graph as delineated in the corresponding knowledge graph. This dual-context paradigm helps nodes within a query graph attain refined internal representations throughout the multi-hop reasoning steps. Through experiments on two datasets, our method consistently enhances the three multi-hop reasoning foundation models, achieving performance improvements of up to 19.5{\%}. Our codes are available at https://github.com/kjh9503/caqr.
|
[
"Kim, Jeonghoon",
"Jung, Heesoo",
"Jang, Hyeju",
"Park, Hogun"
] |
Improving Multi-hop Logical Reasoning in Knowledge Graphs with Context-Aware Query Representation Learning
|
findings-acl.946
|
Poster
|
2406.07034v1
|
https://aclanthology.org/2024.findings-acl.947.bib
|
@inproceedings{heng-etal-2024-proggen,
title = "{P}rog{G}en: Generating Named Entity Recognition Datasets Step-by-step with Self-Reflexive Large Language Models",
author = "Heng, Yuzhao and
Deng, Chunyuan and
Li, Yitong and
Yu, Yue and
Li, Yinghao and
Zhang, Rongzhi and
Zhang, Chao",
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.947",
pages = "15992--16030",
abstract = "Although Large Language Models (LLMs) exhibit remarkable adaptability across domains, these models often fall short in structured knowledge extraction tasks such as named entity recognition (NER). This paper explores an innovative, cost-efficient strategy to harness LLMs with modest NER capabilities for producing superior NER datasets. Our approach diverges from the basic class-conditional prompts by instructing LLMs to self-reflect on the specific domain, thereby generating domain-relevant attributes (such as category and emotions for movie reviews), which are utilized for creating attribute-rich training data. Furthermore, we preemptively generate entity terms and then develop NER context data around these entities, effectively bypassing the LLMs{'} challenges with complex structures. Our experiments across both general and niche domains reveal significant performance enhancements over conventional data generation methods while being more cost-effective than existing alternatives.",
}
|
Although Large Language Models (LLMs) exhibit remarkable adaptability across domains, these models often fall short in structured knowledge extraction tasks such as named entity recognition (NER). This paper explores an innovative, cost-efficient strategy to harness LLMs with modest NER capabilities for producing superior NER datasets. Our approach diverges from the basic class-conditional prompts by instructing LLMs to self-reflect on the specific domain, thereby generating domain-relevant attributes (such as category and emotions for movie reviews), which are utilized for creating attribute-rich training data. Furthermore, we preemptively generate entity terms and then develop NER context data around these entities, effectively bypassing the LLMs{'} challenges with complex structures. Our experiments across both general and niche domains reveal significant performance enhancements over conventional data generation methods while being more cost-effective than existing alternatives.
|
[
"Heng, Yuzhao",
"Deng, Chunyuan",
"Li, Yitong",
"Yu, Yue",
"Li, Yinghao",
"Zhang, Rongzhi",
"Zhang, Chao"
] |
{P}rog{G}en: Generating Named Entity Recognition Datasets Step-by-step with Self-Reflexive Large Language Models
|
findings-acl.947
|
Poster
|
2308.07791v1
|
https://aclanthology.org/2024.findings-acl.948.bib
|
@inproceedings{wang-etal-2024-defending,
title = "Defending {LLM}s against Jailbreaking Attacks via Backtranslation",
author = "Wang, Yihan and
Shi, Zhouxing and
Bai, Andrew and
Hsieh, Cho-Jui",
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.948",
pages = "16031--16046",
abstract = "Although many large language models (LLMs) have been trained to refuse harmful requests, they are still vulnerable to jailbreaking attacks which rewrite the original prompt to conceal its harmful intent. In this paper, we propose a new method for defending LLMs against jailbreaking attacks by {``}backtranslation{''}. Specifically, given an initial response generated by the target LLM from an input prompt, our backtranslation prompts a language model to infer an input prompt that can lead to the response. The inferred prompt is called the backtranslated prompt which tends to reveal the actual intent of the original prompt, since it is generated based on the LLM{'}s response and not directly manipulated by the attacker. We then run the target LLM again on the backtranslated prompt, and we refuse the original prompt if the model refuses the backtranslated prompt. We explain that the proposed defense provides several benefits on its effectiveness and efficiency. We empirically demonstrate that our defense significantly outperforms the baselines, in the cases that are hard for the baselines, and our defense also has little impact on the generation quality for benign input prompts. Our implementation is based on our library for LLM jailbreaking defense algorithms at \url{https://github.com/YihanWang617/llm-jailbreaking-defense}, and the code for reproducing our experiments is available at \url{https://github.com/YihanWang617/LLM-Jailbreaking-Defense-Backtranslation}.",
}
|
Although many large language models (LLMs) have been trained to refuse harmful requests, they are still vulnerable to jailbreaking attacks which rewrite the original prompt to conceal its harmful intent. In this paper, we propose a new method for defending LLMs against jailbreaking attacks by {``}backtranslation{''}. Specifically, given an initial response generated by the target LLM from an input prompt, our backtranslation prompts a language model to infer an input prompt that can lead to the response. The inferred prompt is called the backtranslated prompt which tends to reveal the actual intent of the original prompt, since it is generated based on the LLM{'}s response and not directly manipulated by the attacker. We then run the target LLM again on the backtranslated prompt, and we refuse the original prompt if the model refuses the backtranslated prompt. We explain that the proposed defense provides several benefits on its effectiveness and efficiency. We empirically demonstrate that our defense significantly outperforms the baselines, in the cases that are hard for the baselines, and our defense also has little impact on the generation quality for benign input prompts. Our implementation is based on our library for LLM jailbreaking defense algorithms at \url{https://github.com/YihanWang617/llm-jailbreaking-defense}, and the code for reproducing our experiments is available at \url{https://github.com/YihanWang617/LLM-Jailbreaking-Defense-Backtranslation}.
|
[
"Wang, Yihan",
"Shi, Zhouxing",
"Bai, Andrew",
"Hsieh, Cho-Jui"
] |
Defending {LLM}s against Jailbreaking Attacks via Backtranslation
|
findings-acl.948
|
Poster
|
2402.16459v3
|
https://aclanthology.org/2024.findings-acl.949.bib
|
@inproceedings{sato-etal-2024-large,
title = "A Large Collection of Model-generated Contradictory Responses for Consistency-aware Dialogue Systems",
author = "Sato, Shiki and
Akama, Reina and
Suzuki, Jun and
Inui, Kentaro",
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.949",
pages = "16047--16062",
abstract = "Mitigating the generation of contradictory responses poses a substantial challenge in dialogue response generation. The quality and quantity of available contradictory response data play a vital role in suppressing these contradictions, offering two significant benefits. First, having access to large contradiction data enables a comprehensive examination of their characteristics. Second, data-driven methods to mitigate contradictions may be enhanced with large-scale contradiction data for training. Nevertheless, no attempt has been made to build an extensive collection of model-generated contradictory responses. In this paper, we build a large dataset of response generation models{'} contradictions for the first time. Then, we acquire valuable insights into the characteristics of model-generated contradictions through an extensive analysis of the collected responses. Lastly, we also demonstrate how this dataset substantially enhances the performance of data-driven contradiction suppression methods.",
}
|
Mitigating the generation of contradictory responses poses a substantial challenge in dialogue response generation. The quality and quantity of available contradictory response data play a vital role in suppressing these contradictions, offering two significant benefits. First, having access to large contradiction data enables a comprehensive examination of their characteristics. Second, data-driven methods to mitigate contradictions may be enhanced with large-scale contradiction data for training. Nevertheless, no attempt has been made to build an extensive collection of model-generated contradictory responses. In this paper, we build a large dataset of response generation models{'} contradictions for the first time. Then, we acquire valuable insights into the characteristics of model-generated contradictions through an extensive analysis of the collected responses. Lastly, we also demonstrate how this dataset substantially enhances the performance of data-driven contradiction suppression methods.
|
[
"Sato, Shiki",
"Akama, Reina",
"Suzuki, Jun",
"Inui, Kentaro"
] |
A Large Collection of Model-generated Contradictory Responses for Consistency-aware Dialogue Systems
|
findings-acl.949
|
Poster
|
2403.12500v1
|
https://aclanthology.org/2024.findings-acl.950.bib
|
@inproceedings{ozeki-etal-2024-exploring,
title = "Exploring Reasoning Biases in Large Language Models Through Syllogism: Insights from the {N}eu{BAROCO} Dataset",
author = "Ozeki, Kentaro and
Ando, Risako and
Morishita, Takanobu and
Abe, Hirohiko and
Mineshima, Koji and
Okada, Mitsuhiro",
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.950",
pages = "16063--16077",
abstract = "This paper explores the question of how accurately current large language models can perform logical reasoning in natural language, with an emphasis on whether these models exhibit reasoning biases similar to humans. Specifically, our study focuses on syllogistic reasoning, a form of deductive reasoning extensively studied in cognitive science as a natural form of human reasoning. We present a syllogism dataset called NeuBAROCO, which consists of syllogistic reasoning problems in English and Japanese. This dataset was originally designed for psychological experiments to assess human reasoning capabilities using various forms of syllogisms. Our experiments with leading large language models indicate that these models exhibit reasoning biases similar to humans, along with other error tendencies. Notably, there is significant room for improvement in reasoning problems where the relationship between premises and hypotheses is neither entailment nor contradiction. We also present experimental results and in-depth analysis using a new Chain-of-Thought prompting method, which asks LLMs to translate syllogisms into abstract logical expressions and then explain their reasoning process. Our analysis using this method suggests that the primary limitations of LLMs lie in the reasoning process itself rather than the interpretation of syllogisms.",
}
|
This paper explores the question of how accurately current large language models can perform logical reasoning in natural language, with an emphasis on whether these models exhibit reasoning biases similar to humans. Specifically, our study focuses on syllogistic reasoning, a form of deductive reasoning extensively studied in cognitive science as a natural form of human reasoning. We present a syllogism dataset called NeuBAROCO, which consists of syllogistic reasoning problems in English and Japanese. This dataset was originally designed for psychological experiments to assess human reasoning capabilities using various forms of syllogisms. Our experiments with leading large language models indicate that these models exhibit reasoning biases similar to humans, along with other error tendencies. Notably, there is significant room for improvement in reasoning problems where the relationship between premises and hypotheses is neither entailment nor contradiction. We also present experimental results and in-depth analysis using a new Chain-of-Thought prompting method, which asks LLMs to translate syllogisms into abstract logical expressions and then explain their reasoning process. Our analysis using this method suggests that the primary limitations of LLMs lie in the reasoning process itself rather than the interpretation of syllogisms.
|
[
"Ozeki, Kentaro",
"Ando, Risako",
"Morishita, Takanobu",
"Abe, Hirohiko",
"Mineshima, Koji",
"Okada, Mitsuhiro"
] |
Exploring Reasoning Biases in Large Language Models Through Syllogism: Insights from the {N}eu{BAROCO} Dataset
|
findings-acl.950
|
Poster
|
2408.04403v1
|
https://aclanthology.org/2024.findings-acl.951.bib
|
@inproceedings{deng-etal-2024-unveiling,
title = "Unveiling the Spectrum of Data Contamination in Language Model: A Survey from Detection to Remediation",
author = "Deng, Chunyuan and
Zhao, Yilun and
Heng, Yuzhao and
Li, Yitong and
Cao, Jiannan and
Tang, Xiangru and
Cohan, Arman",
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.951",
pages = "16078--16092",
abstract = "Data contamination has garnered increased attention in the era of Large language models (LLMs) due to the reliance on extensive internet-derived training corpora. The issue of training corpus overlap with evaluation benchmarks{---}referred to as contamination{---}has been the focus of significant recent research. This body of work aims to identify contamination, understand its impacts, and explore mitigation strategies from diverse perspectives. However, comprehensive studies that provide a clear pathway from foundational concepts to advanced insights are lacking in this nascent field. Therefore, we present the first survey in the field of data contamination. We begin by examining the effects of data contamination across various stages and forms. We then provide a detailed analysis of current contamination detection methods, categorizing them to highlight their focus, assumptions, strengths, and limitations. We also discuss mitigation strategies, offering a clear guide for future research. This survey serves as a succinct overview of the most recent advancements in data contamination research, providing a straightforward guide for the benefit of future research endeavors.",
}
|
Data contamination has garnered increased attention in the era of Large language models (LLMs) due to the reliance on extensive internet-derived training corpora. The issue of training corpus overlap with evaluation benchmarks{---}referred to as contamination{---}has been the focus of significant recent research. This body of work aims to identify contamination, understand its impacts, and explore mitigation strategies from diverse perspectives. However, comprehensive studies that provide a clear pathway from foundational concepts to advanced insights are lacking in this nascent field. Therefore, we present the first survey in the field of data contamination. We begin by examining the effects of data contamination across various stages and forms. We then provide a detailed analysis of current contamination detection methods, categorizing them to highlight their focus, assumptions, strengths, and limitations. We also discuss mitigation strategies, offering a clear guide for future research. This survey serves as a succinct overview of the most recent advancements in data contamination research, providing a straightforward guide for the benefit of future research endeavors.
|
[
"Deng, Chunyuan",
"Zhao, Yilun",
"Heng, Yuzhao",
"Li, Yitong",
"Cao, Jiannan",
"Tang, Xiangru",
"Cohan, Arman"
] |
Unveiling the Spectrum of Data Contamination in Language Model: A Survey from Detection to Remediation
|
findings-acl.951
|
Poster
|
2406.14644v1
|
https://aclanthology.org/2024.findings-acl.952.bib
|
@inproceedings{mondal-etal-2024-dimsim,
title = "{DIMSIM}: Distilled Multilingual Critics for {I}ndic Text Simplification",
author = "Mondal, Sneha and
Ritika, Ritika and
Agrawal, Ashish and
Jyothi, Preethi and
Raghuveer, Aravindan",
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.952",
pages = "16093--16109",
abstract = "Self-correction techniques have recently emerged as a promising framework to improve the quality of responses generated by large language models (LLMs). Few-shot prompted LLMs act as critics to produce feedback for an input, which is further fed to a refiner (also an LLM) to produce an output. However, these critique-refine steps require multiple expensive LLM calls. To circumvent this large inference cost, we borrow inspiration from prior work on knowledge distillation and propose the use of critique distillation to train critic models. These are smaller sequence-to-sequence models that are trained on input-critique pairs generated by an LLM. We focus on the problem of text simplification for three Indian languages: Hindi, Bengali and Marathi. This task is a good fit for self-correction style techniques. It also hasn{'}t been systematically explored for Indian languages before. We train two separate critics that focus on lexical and structure complexity, and show that it is surprisingly more effective than using an LLM directly as a critic in both 0-shot and few-shot settings. We also show the benefits of training multilingual critics, as opposed to monolingual critics. Extensive human evaluations show that on average, raters find 80{\%} of DIMSIM{'}s output to be simple and easy to read.",
}
|
Self-correction techniques have recently emerged as a promising framework to improve the quality of responses generated by large language models (LLMs). Few-shot prompted LLMs act as critics to produce feedback for an input, which is further fed to a refiner (also an LLM) to produce an output. However, these critique-refine steps require multiple expensive LLM calls. To circumvent this large inference cost, we borrow inspiration from prior work on knowledge distillation and propose the use of critique distillation to train critic models. These are smaller sequence-to-sequence models that are trained on input-critique pairs generated by an LLM. We focus on the problem of text simplification for three Indian languages: Hindi, Bengali and Marathi. This task is a good fit for self-correction style techniques. It also hasn{'}t been systematically explored for Indian languages before. We train two separate critics that focus on lexical and structure complexity, and show that it is surprisingly more effective than using an LLM directly as a critic in both 0-shot and few-shot settings. We also show the benefits of training multilingual critics, as opposed to monolingual critics. Extensive human evaluations show that on average, raters find 80{\%} of DIMSIM{'}s output to be simple and easy to read.
|
[
"Mondal, Sneha",
"Ritika, Ritika",
"Agrawal, Ashish",
"Jyothi, Preethi",
"Raghuveer, Aravindan"
] |
{DIMSIM}: Distilled Multilingual Critics for {I}ndic Text Simplification
|
findings-acl.952
|
Poster
|
2302.02888v1
|
https://aclanthology.org/2024.findings-acl.953.bib
|
@inproceedings{lee-etal-2024-matter,
title = "{MATTER}: Memory-Augmented Transformer Using Heterogeneous Knowledge Sources",
author = "Lee, Dongkyu and
Satya Prakash, Chandana and
FitzGerald, Jack and
Lehmann, Jens",
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.953",
pages = "16110--16121",
abstract = "Leveraging external knowledge is crucial for achieving high performance in knowledge-intensive tasks, such as question answering. The retrieve-and-read approach is widely adopted for integrating external knowledge into a language model. However, this approach suffers from increased computational cost and latency due to the long context length, which grows proportionally with the number of retrieved knowledge. Furthermore, existing retrieval-augmented models typically retrieve information from a single type of knowledge source, limiting their scalability to diverse knowledge sources with varying structures. In this work, we introduce an efficient memory-augmented transformer called MATTER, designed to retrieve relevant knowledge from multiple heterogeneous knowledge sources. Specifically, our model retrieves and reads from both unstructured sources (paragraphs) and semi-structured sources (QA pairs) in the form of fixed-length neural memories. We demonstrate that our model outperforms existing efficient retrieval-augmented models on popular QA benchmarks in terms of both accuracy and speed. Furthermore, MATTER achieves competitive results compared to conventional read-and-retrieve models while having 100x throughput during inference.",
}
|
Leveraging external knowledge is crucial for achieving high performance in knowledge-intensive tasks, such as question answering. The retrieve-and-read approach is widely adopted for integrating external knowledge into a language model. However, this approach suffers from increased computational cost and latency due to the long context length, which grows proportionally with the number of retrieved knowledge. Furthermore, existing retrieval-augmented models typically retrieve information from a single type of knowledge source, limiting their scalability to diverse knowledge sources with varying structures. In this work, we introduce an efficient memory-augmented transformer called MATTER, designed to retrieve relevant knowledge from multiple heterogeneous knowledge sources. Specifically, our model retrieves and reads from both unstructured sources (paragraphs) and semi-structured sources (QA pairs) in the form of fixed-length neural memories. We demonstrate that our model outperforms existing efficient retrieval-augmented models on popular QA benchmarks in terms of both accuracy and speed. Furthermore, MATTER achieves competitive results compared to conventional read-and-retrieve models while having 100x throughput during inference.
|
[
"Lee, Dongkyu",
"Satya Prakash, Ch",
"ana",
"FitzGerald, Jack",
"Lehmann, Jens"
] |
{MATTER}: Memory-Augmented Transformer Using Heterogeneous Knowledge Sources
|
findings-acl.953
|
Poster
|
2406.04670v1
|
https://aclanthology.org/2024.findings-acl.954.bib
|
@inproceedings{shin-etal-2024-ask,
title = "Ask {LLM}s Directly, {``}What shapes your bias?{''}: Measuring Social Bias in Large Language Models",
author = "Shin, Jisu and
Song, Hoyun and
Lee, Huije and
Jeong, Soyeong and
Park, Jong",
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.954",
pages = "16122--16143",
abstract = "Social bias is shaped by the accumulation of social perceptions towards targets across various demographic identities. To fully understand such social bias in large language models (LLMs), it is essential to consider the composite of social perceptions from diverse perspectives among identities. Previous studies have either evaluated biases in LLMs by indirectly assessing the presence of sentiments towards demographic identities in the generated text or measuring the degree of alignment with given stereotypes. These methods have limitations in directly quantifying social biases at the level of distinct perspectives among identities. In this paper, we aim to investigate how social perceptions from various viewpoints contribute to the development of social bias in LLMs. To this end, we propose a novel strategy to intuitively quantify these social perceptions and suggest metrics that can evaluate the social biases within LLMs by aggregating diverse social perceptions. The experimental results show the quantitative demonstration of the social attitude in LLMs by examining social perception. The analysis we conducted shows that our proposed metrics capture the multi-dimensional aspects of social bias, enabling a fine-grained and comprehensive investigation of bias in LLMs.",
}
|
Social bias is shaped by the accumulation of social perceptions towards targets across various demographic identities. To fully understand such social bias in large language models (LLMs), it is essential to consider the composite of social perceptions from diverse perspectives among identities. Previous studies have either evaluated biases in LLMs by indirectly assessing the presence of sentiments towards demographic identities in the generated text or measuring the degree of alignment with given stereotypes. These methods have limitations in directly quantifying social biases at the level of distinct perspectives among identities. In this paper, we aim to investigate how social perceptions from various viewpoints contribute to the development of social bias in LLMs. To this end, we propose a novel strategy to intuitively quantify these social perceptions and suggest metrics that can evaluate the social biases within LLMs by aggregating diverse social perceptions. The experimental results show the quantitative demonstration of the social attitude in LLMs by examining social perception. The analysis we conducted shows that our proposed metrics capture the multi-dimensional aspects of social bias, enabling a fine-grained and comprehensive investigation of bias in LLMs.
|
[
"Shin, Jisu",
"Song, Hoyun",
"Lee, Huije",
"Jeong, Soyeong",
"Park, Jong"
] |
Ask {LLM}s Directly, {``}What shapes your bias?{''}: Measuring Social Bias in Large Language Models
|
findings-acl.954
|
Poster
|
2406.04064
|
https://aclanthology.org/2024.findings-acl.955.bib
|
@inproceedings{xia-etal-2024-chain,
title = "Chain-of-History Reasoning for Temporal Knowledge Graph Forecasting",
author = "Xia, Yuwei and
Wang, Ding and
Liu, Qiang and
Wang, Liang and
Wu, Shu and
Zhang, Xiao-Yu",
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.955",
pages = "16144--16159",
abstract = "Temporal Knowledge Graph (TKG) forecasting aims to predict future facts based on given histories. Most recent graph-based models excel at capturing structural information within TKGs but lack semantic comprehension abilities. Nowadays, with the surge of LLMs, the LLM-based TKG prediction model has emerged. However, the existing LLM-based model exhibits three shortcomings: (1) It only focuses on the first-order history for prediction while ignoring high-order historical information, resulting in the provided information for LLMs being extremely limited. (2) LLMs struggle with optimal reasoning performance under heavy historical information loads. (3) For TKG prediction, the temporal reasoning capability of LLM alone is limited. To address the first two challenges, we propose Chain-of-History (CoH) reasoning which explores high-order histories step-by-step, achieving effective utilization of high-order historical information for LLMs on TKG prediction. To address the third issue, we design CoH as a plug-and-play module to enhance the performance of graph-based models for TKG prediction. Extensive experiments on three datasets and backbones demonstrate the effectiveness of CoH.",
}
|
Temporal Knowledge Graph (TKG) forecasting aims to predict future facts based on given histories. Most recent graph-based models excel at capturing structural information within TKGs but lack semantic comprehension abilities. Nowadays, with the surge of LLMs, the LLM-based TKG prediction model has emerged. However, the existing LLM-based model exhibits three shortcomings: (1) It only focuses on the first-order history for prediction while ignoring high-order historical information, resulting in the provided information for LLMs being extremely limited. (2) LLMs struggle with optimal reasoning performance under heavy historical information loads. (3) For TKG prediction, the temporal reasoning capability of LLM alone is limited. To address the first two challenges, we propose Chain-of-History (CoH) reasoning which explores high-order histories step-by-step, achieving effective utilization of high-order historical information for LLMs on TKG prediction. To address the third issue, we design CoH as a plug-and-play module to enhance the performance of graph-based models for TKG prediction. Extensive experiments on three datasets and backbones demonstrate the effectiveness of CoH.
|
[
"Xia, Yuwei",
"Wang, Ding",
"Liu, Qiang",
"Wang, Liang",
"Wu, Shu",
"Zhang, Xiao-Yu"
] |
Chain-of-History Reasoning for Temporal Knowledge Graph Forecasting
|
findings-acl.955
|
Poster
|
2012.15537v5
|
https://aclanthology.org/2024.findings-acl.956.bib
|
@inproceedings{li-etal-2024-llms-speak,
title = "Can {LLM}s Speak For Diverse People? Tuning {LLM}s via Debate to Generate Controllable Controversial Statements",
author = "Li, Ming and
Chen, Jiuhai and
Chen, Lichang and
Zhou, Tianyi",
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.956",
pages = "16160--16176",
abstract = "Making LLMs speak for different, especially minority groups of people, and generate statements supporting their diverse or even controversial perspectives is critical to creating an inclusive environment. However, existing LLMs lack sufficient controllability to the stance of their generated content, which often contains inconsistent, neutral, or biased statements. In this paper, we improve the controllability of LLMs in generating statements supporting an argument the user defined in the prompt. We find that multi-round debates between two LLMs with opposite stances generate higher-quality and more salient statements for each, which are important training data to improve the controllability of LLMs. Motivated by this, we develop a novel debate {\&} tuning ({``}DEBATUNE{''}) pipeline finetuning LLMs to generate the statements obtained via debate. To examine DEBATUNE, we curate the largest dataset of debate topics so far, which covers 710 controversial topics and corresponding arguments for each topic. Evaluations by the GPT-4 judge with a novel controversy controllability metric show that LLMs{'} capability of generating diverse perspectives is significantly improved by DEBATUNE. Moreover, such controllability can be generalized to unseen topics, generating high-quality statements supporting controversial arguments.",
}
|
Making LLMs speak for different, especially minority groups of people, and generate statements supporting their diverse or even controversial perspectives is critical to creating an inclusive environment. However, existing LLMs lack sufficient controllability to the stance of their generated content, which often contains inconsistent, neutral, or biased statements. In this paper, we improve the controllability of LLMs in generating statements supporting an argument the user defined in the prompt. We find that multi-round debates between two LLMs with opposite stances generate higher-quality and more salient statements for each, which are important training data to improve the controllability of LLMs. Motivated by this, we develop a novel debate {\&} tuning ({``}DEBATUNE{''}) pipeline finetuning LLMs to generate the statements obtained via debate. To examine DEBATUNE, we curate the largest dataset of debate topics so far, which covers 710 controversial topics and corresponding arguments for each topic. Evaluations by the GPT-4 judge with a novel controversy controllability metric show that LLMs{'} capability of generating diverse perspectives is significantly improved by DEBATUNE. Moreover, such controllability can be generalized to unseen topics, generating high-quality statements supporting controversial arguments.
|
[
"Li, Ming",
"Chen, Jiuhai",
"Chen, Lichang",
"Zhou, Tianyi"
] |
Can {LLM}s Speak For Diverse People? Tuning {LLM}s via Debate to Generate Controllable Controversial Statements
|
findings-acl.956
|
Poster
|
2402.10614v2
|
https://aclanthology.org/2024.findings-acl.957.bib
|
@inproceedings{kim-etal-2024-label,
title = "Label-aware Hard Negative Sampling Strategies with Momentum Contrastive Learning for Implicit Hate Speech Detection",
author = "Kim, Jaehoon and
Jin, Seungwan and
Park, Sohyun and
Park, Someen and
Han, Kyungsik",
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.957",
pages = "16177--16188",
abstract = "Detecting implicit hate speech that is not directly hateful remains a challenge. Recent research has attempted to detect implicit hate speech by applying contrastive learning to pre-trained language models such as BERT and RoBERTa, but the proposed models still do not have a significant advantage over cross-entropy loss-based learning. We found that contrastive learning based on randomly sampled batch data does not encourage the model to learn hard negative samples. In this work, we propose Label-aware Hard Negative sampling strategies (LAHN) that encourage the model to learn detailed features from hard negative samples, instead of naive negative samples in random batch, using momentum-integrated contrastive learning. LAHN outperforms the existing models for implicit hate speech detection both in- and cross-datasets. The code is available at https://github.com/Hanyang-HCC-Lab/LAHN",
}
|
Detecting implicit hate speech that is not directly hateful remains a challenge. Recent research has attempted to detect implicit hate speech by applying contrastive learning to pre-trained language models such as BERT and RoBERTa, but the proposed models still do not have a significant advantage over cross-entropy loss-based learning. We found that contrastive learning based on randomly sampled batch data does not encourage the model to learn hard negative samples. In this work, we propose Label-aware Hard Negative sampling strategies (LAHN) that encourage the model to learn detailed features from hard negative samples, instead of naive negative samples in random batch, using momentum-integrated contrastive learning. LAHN outperforms the existing models for implicit hate speech detection both in- and cross-datasets. The code is available at https://github.com/Hanyang-HCC-Lab/LAHN
|
[
"Kim, Jaehoon",
"Jin, Seungwan",
"Park, Sohyun",
"Park, Someen",
"Han, Kyungsik"
] |
Label-aware Hard Negative Sampling Strategies with Momentum Contrastive Learning for Implicit Hate Speech Detection
|
findings-acl.957
|
Poster
|
2406.07886v1
|
https://aclanthology.org/2024.findings-acl.958.bib
|
@inproceedings{li-etal-2024-selective,
title = "Selective Reflection-Tuning: Student-Selected Data Recycling for {LLM} Instruction-Tuning",
author = "Li, Ming and
Chen, Lichang and
Chen, Jiuhai and
He, Shwai and
Gu, Jiuxiang and
Zhou, Tianyi",
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.958",
pages = "16189--16211",
abstract = "Instruction tuning is critical to large language models (LLMs) for achieving better instruction following and task adaptation capabilities but its success heavily relies on the training data quality. Many recent methods focus on improving the data quality but often overlook the compatibility of the data with the student model being finetuned. This paper introduces Selective Reflection-Tuning, a novel paradigm that synergizes a teacher LLM{'}s reflection and introspection for improving existing data quality with the data selection capability of the student LLM, to automatically refine existing instruction-tuning data. This teacher-student collaboration produces high-quality and student-compatible instruction-response pairs, resulting in sample-efficient instruction tuning and LLMs of superior performance. Selective Reflection-Tuning is a data augmentation and synthesis that generally improves LLM finetuning and self-improvement without collecting brand-new data. We apply our method to Alpaca and WizardLM data and achieve much stronger and top-tier 7B and 13B LLMs.",
}
|
Instruction tuning is critical to large language models (LLMs) for achieving better instruction following and task adaptation capabilities but its success heavily relies on the training data quality. Many recent methods focus on improving the data quality but often overlook the compatibility of the data with the student model being finetuned. This paper introduces Selective Reflection-Tuning, a novel paradigm that synergizes a teacher LLM{'}s reflection and introspection for improving existing data quality with the data selection capability of the student LLM, to automatically refine existing instruction-tuning data. This teacher-student collaboration produces high-quality and student-compatible instruction-response pairs, resulting in sample-efficient instruction tuning and LLMs of superior performance. Selective Reflection-Tuning is a data augmentation and synthesis that generally improves LLM finetuning and self-improvement without collecting brand-new data. We apply our method to Alpaca and WizardLM data and achieve much stronger and top-tier 7B and 13B LLMs.
|
[
"Li, Ming",
"Chen, Lichang",
"Chen, Jiuhai",
"He, Shwai",
"Gu, Jiuxiang",
"Zhou, Tianyi"
] |
Selective Reflection-Tuning: Student-Selected Data Recycling for {LLM} Instruction-Tuning
|
findings-acl.958
|
Poster
|
2309.03876v1
|
https://aclanthology.org/2024.findings-acl.959.bib
|
@inproceedings{huang-etal-2024-selective,
title = "Selective Prompting Tuning for Personalized Conversations with {LLM}s",
author = "Huang, Qiushi and
Liu, Xubo and
Ko, Tom and
Wu, Bo and
Wang, Wenwu and
Zhang, Yu and
Tang, Lilian",
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.959",
pages = "16212--16226",
abstract = "In conversational AI, personalizing dialogues with persona profiles and contextual understanding is essential. Despite large language models{'} (LLMs) improved response coherence, effective persona integration remains a challenge. In this work, we first study two common approaches for personalizing LLMs: textual prompting and direct fine-tuning. We observed that textual prompting often struggles to yield responses that are similar to the ground truths in datasets, while direct fine-tuning tends to produce repetitive or overly generic replies. To alleviate those issues, we propose **S**elective **P**rompt **T**uning (SPT), which softly prompts LLMs for personalized conversations in a selective way. Concretely, SPT initializes a set of soft prompts and uses a trainable dense retriever to adaptively select suitable soft prompts for LLMs according to different input contexts, where the prompt retriever is dynamically updated through feedback from the LLMs. Additionally, we propose context-prompt contrastive learning and prompt fusion learning to encourage the SPT to enhance the diversity of personalized conversations. Experiments on the CONVAI2 dataset demonstrate that SPT significantly enhances response diversity by up to 90{\%}, along with improvements in other critical performance indicators. Those results highlight the efficacy of SPT in fostering engaging and personalized dialogue generation. The SPT model code is [publicly available](https://github.com/hqsiswiliam/SPT) for further exploration.",
}
|
In conversational AI, personalizing dialogues with persona profiles and contextual understanding is essential. Despite large language models{'} (LLMs) improved response coherence, effective persona integration remains a challenge. In this work, we first study two common approaches for personalizing LLMs: textual prompting and direct fine-tuning. We observed that textual prompting often struggles to yield responses that are similar to the ground truths in datasets, while direct fine-tuning tends to produce repetitive or overly generic replies. To alleviate those issues, we propose **S**elective **P**rompt **T**uning (SPT), which softly prompts LLMs for personalized conversations in a selective way. Concretely, SPT initializes a set of soft prompts and uses a trainable dense retriever to adaptively select suitable soft prompts for LLMs according to different input contexts, where the prompt retriever is dynamically updated through feedback from the LLMs. Additionally, we propose context-prompt contrastive learning and prompt fusion learning to encourage the SPT to enhance the diversity of personalized conversations. Experiments on the CONVAI2 dataset demonstrate that SPT significantly enhances response diversity by up to 90{\%}, along with improvements in other critical performance indicators. Those results highlight the efficacy of SPT in fostering engaging and personalized dialogue generation. The SPT model code is [publicly available](https://github.com/hqsiswiliam/SPT) for further exploration.
|
[
"Huang, Qiushi",
"Liu, Xubo",
"Ko, Tom",
"Wu, Bo",
"Wang, Wenwu",
"Zhang, Yu",
"Tang, Lilian"
] |
Selective Prompting Tuning for Personalized Conversations with {LLM}s
|
findings-acl.959
|
Poster
|
2406.18187v1
|
https://aclanthology.org/2024.findings-acl.960.bib
|
@inproceedings{hazra-etal-2024-sowing,
title = "Sowing the Wind, Reaping the Whirlwind: The Impact of Editing Language Models",
author = "Hazra, Rima and
Layek, Sayan and
Banerjee, Somnath and
Poria, Soujanya",
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.960",
pages = "16227--16239",
abstract = "In the rapidly advancing field of artificial intelligence, the concept of {`}Red-Teaming{'} or {`}Jailbreaking{'} large language models (LLMs) has emerged as a crucial area of study. This approach is especially significant in terms of assessing and enhancing the safety and robustness of these models. This paper investigates the intricate consequences of such modifications through model editing, uncovering a complex relationship between enhancing model accuracy and preserving its ethical integrity. Our in-depth analysis reveals a striking paradox: while injecting accurate information is crucial for model reliability, it can paradoxically destabilize the model{'}s foundational framework, resulting in unpredictable and potentially unsafe behaviors. Additionally, we propose a benchmark dataset $NicheHazardQA$ to investigate this unsafe behavior both within the same and cross topical domain. This aspect of our research sheds light on how the edits, impact the model{'}s safety metrics and guardrails. Our findings show that model editing serves as a cost-effective tool for topical red-teaming by methodically applying targeted edits and evaluating the resultant model behavior.",
}
|
In the rapidly advancing field of artificial intelligence, the concept of {`}Red-Teaming{'} or {`}Jailbreaking{'} large language models (LLMs) has emerged as a crucial area of study. This approach is especially significant in terms of assessing and enhancing the safety and robustness of these models. This paper investigates the intricate consequences of such modifications through model editing, uncovering a complex relationship between enhancing model accuracy and preserving its ethical integrity. Our in-depth analysis reveals a striking paradox: while injecting accurate information is crucial for model reliability, it can paradoxically destabilize the model{'}s foundational framework, resulting in unpredictable and potentially unsafe behaviors. Additionally, we propose a benchmark dataset $NicheHazardQA$ to investigate this unsafe behavior both within the same and cross topical domain. This aspect of our research sheds light on how the edits, impact the model{'}s safety metrics and guardrails. Our findings show that model editing serves as a cost-effective tool for topical red-teaming by methodically applying targeted edits and evaluating the resultant model behavior.
|
[
"Hazra, Rima",
"Layek, Sayan",
"Banerjee, Somnath",
"Poria, Soujanya"
] |
Sowing the Wind, Reaping the Whirlwind: The Impact of Editing Language Models
|
findings-acl.960
|
Poster
|
2401.10647v5
|
https://aclanthology.org/2024.findings-acl.961.bib
|
@inproceedings{lin-etal-2024-contextblip,
title = "{C}ontext{BLIP}: Doubly Contextual Alignment for Contrastive Image Retrieval from Linguistically Complex Descriptions",
author = "Lin, Honglin and
Li, Siyu and
Nan, Guoshun and
Tang, Chaoyue and
Wang, Xueting and
Xu, Jingxin and
Yankai, Rong and
Zhouzhili, Zhouzhili and
Gao, Yutong and
Cui, Qimei and
Tao, Xiaofeng",
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.961",
pages = "16240--16258",
abstract = "Image retrieval from contextual descriptions (IRCD) aims to identify an image within a set of minimally contrastive candidates based on linguistically complex text. Despite the success of VLMs, they still significantly lag behind human performance in IRCD. The main challenges lie in aligning key contextual cues in two modalities, where these subtle cues are concealed in tiny areas of multiple contrastive images and within the complex linguistics of textual descriptions. This motivates us to propose ContextBLIP, a simple yet effective method that relies on a doubly contextual alignment scheme for challenging IRCD. Specifically, 1) our model comprises a multi-scale adapter, a matching loss, and a text-guided masking loss. The adapter learns to capture fine-grained visual cues. The two losses enable iterative supervision for the adapter, gradually highlighting the focal patches of a single image to the key textual cues. We term such a way as intra-contextual alignment. 2) Then, ContextBLIP further employs an inter-context encoder to learn dependencies among candidates, facilitating alignment between the text to multiple images. We term this step as inter-contextual alignment. Consequently, the nuanced cues concealed in each modality can be effectively aligned. Experiments on two benchmarks show the superiority of our method. We observe that ContextBLIP can yield comparable results with GPT-4V, despite involving about 7,500 times fewer parameters.",
}
|
Image retrieval from contextual descriptions (IRCD) aims to identify an image within a set of minimally contrastive candidates based on linguistically complex text. Despite the success of VLMs, they still significantly lag behind human performance in IRCD. The main challenges lie in aligning key contextual cues in two modalities, where these subtle cues are concealed in tiny areas of multiple contrastive images and within the complex linguistics of textual descriptions. This motivates us to propose ContextBLIP, a simple yet effective method that relies on a doubly contextual alignment scheme for challenging IRCD. Specifically, 1) our model comprises a multi-scale adapter, a matching loss, and a text-guided masking loss. The adapter learns to capture fine-grained visual cues. The two losses enable iterative supervision for the adapter, gradually highlighting the focal patches of a single image to the key textual cues. We term such a way as intra-contextual alignment. 2) Then, ContextBLIP further employs an inter-context encoder to learn dependencies among candidates, facilitating alignment between the text to multiple images. We term this step as inter-contextual alignment. Consequently, the nuanced cues concealed in each modality can be effectively aligned. Experiments on two benchmarks show the superiority of our method. We observe that ContextBLIP can yield comparable results with GPT-4V, despite involving about 7,500 times fewer parameters.
|
[
"Lin, Honglin",
"Li, Siyu",
"Nan, Guoshun",
"Tang, Chaoyue",
"Wang, Xueting",
"Xu, Jingxin",
"Yankai, Rong",
"Zhouzhili, Zhouzhili",
"Gao, Yutong",
"Cui, Qimei",
"Tao, Xiaofeng"
] |
{C}ontext{BLIP}: Doubly Contextual Alignment for Contrastive Image Retrieval from Linguistically Complex Descriptions
|
findings-acl.961
|
Poster
|
2405.19226v1
|
https://aclanthology.org/2024.findings-acl.962.bib
|
@inproceedings{chia-etal-2024-puzzlevqa,
title = "{P}uzzle{VQA}: Diagnosing Multimodal Reasoning Challenges of Language Models with Abstract Visual Patterns",
author = "Chia, Yew Ken and
Toh, Vernon and
Ghosal, Deepanway and
Bing, Lidong and
Poria, Soujanya",
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.962",
pages = "16259--16273",
abstract = "Large multimodal models extend the impressive capabilities of large language models by integrating multimodal understanding abilities. However, it is not clear how they can emulate the general intelligence and reasoning ability of humans. As recognizing patterns and abstracting concepts are key to general intelligence, we introduce PuzzleVQA, a collection of 2000 puzzle instances based on abstract patterns. With this dataset, we evaluate large multimodal models with abstract patterns based on fundamental concepts, including colors, numbers, sizes, and shapes. Through our experiments on state-of-the-art large multimodal models, we find that they are not able to generalize well to simple abstract patterns. Notably, GPT-4V achieves a score of 46.4{\%} on single-concept puzzles, which shows that state-of-the-art models struggle on our dataset. To diagnose the reasoning challenges in large multimodal models, we progressively guide the models with our ground truth reasoning explanations for visual perception, inductive reasoning, and deductive reasoning. Our systematic analysis finds that the main bottlenecks of GPT-4V are weaker visual perception and inductive reasoning abilities. Through this work, we hope to shed light on the limitations of large multimodal models and how they can better emulate human cognitive processes in the future.",
}
|
Large multimodal models extend the impressive capabilities of large language models by integrating multimodal understanding abilities. However, it is not clear how they can emulate the general intelligence and reasoning ability of humans. As recognizing patterns and abstracting concepts are key to general intelligence, we introduce PuzzleVQA, a collection of 2000 puzzle instances based on abstract patterns. With this dataset, we evaluate large multimodal models with abstract patterns based on fundamental concepts, including colors, numbers, sizes, and shapes. Through our experiments on state-of-the-art large multimodal models, we find that they are not able to generalize well to simple abstract patterns. Notably, GPT-4V achieves a score of 46.4{\%} on single-concept puzzles, which shows that state-of-the-art models struggle on our dataset. To diagnose the reasoning challenges in large multimodal models, we progressively guide the models with our ground truth reasoning explanations for visual perception, inductive reasoning, and deductive reasoning. Our systematic analysis finds that the main bottlenecks of GPT-4V are weaker visual perception and inductive reasoning abilities. Through this work, we hope to shed light on the limitations of large multimodal models and how they can better emulate human cognitive processes in the future.
|
[
"Chia, Yew Ken",
"Toh, Vernon",
"Ghosal, Deepanway",
"Bing, Lidong",
"Poria, Soujanya"
] |
{P}uzzle{VQA}: Diagnosing Multimodal Reasoning Challenges of Language Models with Abstract Visual Patterns
|
findings-acl.962
|
Poster
|
2403.13315v2
|
https://aclanthology.org/2024.findings-acl.963.bib
|
@inproceedings{kim-etal-2024-moral,
title = "How Do Moral Emotions Shape Political Participation? A Cross-Cultural Analysis of Online Petitions Using Language Models",
author = "Kim, Jaehong and
Jeong, Chaeyoon and
Park, Seongchan and
Cha, Meeyoung and
Lee, Wonjae",
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.963",
pages = "16274--16289",
abstract = "Understanding the interplay between emotions in language and user behaviors is critical. We study how moral emotions shape the political participation of users based on cross-cultural online petition data. To quantify moral emotions, we employ a context-aware NLP model that is designed to capture the subtle nuances of emotions across cultures. For model training, we construct and share a moral emotion dataset comprising nearly 50,000 petition sentences in Korean and English each, along with emotion labels annotated by a fine-tuned LLM. We examine two distinct types of user participation: general support (i.e., registered signatures of petitions) and active support (i.e., sharing petitions on social media). We discover that moral emotions like other-suffering increase both forms of participation and help petitions go viral, while self-conscious have the opposite effect. The most prominent moral emotion, other-condemning, led to polarizing responses among the audience. In contrast, other-praising was perceived differently by culture; it led to a rise in active support in Korea but a decline in the UK. Our findings suggest that both moral emotions embedded in language and cultural perceptions are critical to shaping the public{'}s political discourse.",
}
|
Understanding the interplay between emotions in language and user behaviors is critical. We study how moral emotions shape the political participation of users based on cross-cultural online petition data. To quantify moral emotions, we employ a context-aware NLP model that is designed to capture the subtle nuances of emotions across cultures. For model training, we construct and share a moral emotion dataset comprising nearly 50,000 petition sentences in Korean and English each, along with emotion labels annotated by a fine-tuned LLM. We examine two distinct types of user participation: general support (i.e., registered signatures of petitions) and active support (i.e., sharing petitions on social media). We discover that moral emotions like other-suffering increase both forms of participation and help petitions go viral, while self-conscious have the opposite effect. The most prominent moral emotion, other-condemning, led to polarizing responses among the audience. In contrast, other-praising was perceived differently by culture; it led to a rise in active support in Korea but a decline in the UK. Our findings suggest that both moral emotions embedded in language and cultural perceptions are critical to shaping the public{'}s political discourse.
|
[
"Kim, Jaehong",
"Jeong, Chaeyoon",
"Park, Seongchan",
"Cha, Meeyoung",
"Lee, Wonjae"
] |
How Do Moral Emotions Shape Political Participation? A Cross-Cultural Analysis of Online Petitions Using Language Models
|
findings-acl.963
|
Poster
|
2404.09329v2
|
https://aclanthology.org/2024.findings-acl.964.bib
|
@inproceedings{dong-etal-2024-villageragent,
title = "{V}illager{A}gent: A Graph-Based Multi-Agent Framework for Coordinating Complex Task Dependencies in {M}inecraft",
author = "Dong, Yubo and
Zhu, Xukun and
Pan, Zhengzhe and
Zhu, Linchao and
Yang, 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.964",
pages = "16290--16314",
abstract = "In this paper, we aim to evaluate multi-agent systems against complex dependencies, including spatial, causal, and temporal constraints. First, we construct a new benchmark, named VillagerBench, within the Minecraft environment. VillagerBench comprises diverse tasks crafted to test various aspects of multi-agent collaboration, from workload distribution to dynamic adaptation and synchronized task execution. Second, we introduce a Directed Acyclic Graph Multi-Agent Framework (VillagerAgent) to resolve complex inter-agent dependencies and enhance collaborative efficiency. This solution incorporates a task decomposer that creates a directed acyclic graph (DAG) for structured task management, an agent controller for task distribution, and a state manager for tracking environmental and agent data.Our empirical evaluation on VillagerBench demonstrates that VillagerAgentoutperforms the existing AgentVerse model, reducing hallucinations and improving task decomposition efficacy. The results underscore VillagerAgent{'}s potential in advancing multi-agent collaboration, offering a scalable and generalizable solution in dynamic environments. Source code is open-source on GitHub.",
}
|
In this paper, we aim to evaluate multi-agent systems against complex dependencies, including spatial, causal, and temporal constraints. First, we construct a new benchmark, named VillagerBench, within the Minecraft environment. VillagerBench comprises diverse tasks crafted to test various aspects of multi-agent collaboration, from workload distribution to dynamic adaptation and synchronized task execution. Second, we introduce a Directed Acyclic Graph Multi-Agent Framework (VillagerAgent) to resolve complex inter-agent dependencies and enhance collaborative efficiency. This solution incorporates a task decomposer that creates a directed acyclic graph (DAG) for structured task management, an agent controller for task distribution, and a state manager for tracking environmental and agent data.Our empirical evaluation on VillagerBench demonstrates that VillagerAgentoutperforms the existing AgentVerse model, reducing hallucinations and improving task decomposition efficacy. The results underscore VillagerAgent{'}s potential in advancing multi-agent collaboration, offering a scalable and generalizable solution in dynamic environments. Source code is open-source on GitHub.
|
[
"Dong, Yubo",
"Zhu, Xukun",
"Pan, Zhengzhe",
"Zhu, Linchao",
"Yang, Yi"
] |
{V}illager{A}gent: A Graph-Based Multi-Agent Framework for Coordinating Complex Task Dependencies in {M}inecraft
|
findings-acl.964
|
Poster
|
2406.05720v1
|
https://aclanthology.org/2024.findings-acl.965.bib
|
@inproceedings{yang-etal-2024-cf,
title = "{CF}-{TCIR}: A Compositor-Free Framework for Hierarchical Text-Conditioned Image Retrieval",
author = "Yang, Yuchen and
Wang, Yu and
Wang, Yanfeng",
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.965",
pages = "16315--16325",
abstract = "In text-conditioned image retrieval (TCIR), the combination of a reference image and modification text forms a query tuple, aiming to locate the most congruent target image within a dataset. The advantages of rich image semantic information and text flexibility are combined in this manner for more accurate retrieval. While traditional techniques often employ attention-driven compositors to craft a unified image-text representation, our paper introduces a compositor-free framework, CF-TCIR, which eschews the standard compositor. Compositor-based methods are designed to learn a joint representation of images and text, but they struggle to directly capture the correlations between attributes across the image and text modalities. Instead, we reformulate the retrieval process as a cross-modal interaction between a synthesized image feature and its corresponding text descriptor. This novel methodology offers advantages in terms of computational efficiency, scalability, and superior performance. To optimize the retrieval performance, we advocate a tiered retrieval mechanism, blending both coarse-grain and fine-grain paradigms. Moreover, to enrich the contextual relationship within the query tuple, we integrate a generative cross-modal alignment technique, ensuring synchronization of sequential attributes between image and text data.",
}
|
In text-conditioned image retrieval (TCIR), the combination of a reference image and modification text forms a query tuple, aiming to locate the most congruent target image within a dataset. The advantages of rich image semantic information and text flexibility are combined in this manner for more accurate retrieval. While traditional techniques often employ attention-driven compositors to craft a unified image-text representation, our paper introduces a compositor-free framework, CF-TCIR, which eschews the standard compositor. Compositor-based methods are designed to learn a joint representation of images and text, but they struggle to directly capture the correlations between attributes across the image and text modalities. Instead, we reformulate the retrieval process as a cross-modal interaction between a synthesized image feature and its corresponding text descriptor. This novel methodology offers advantages in terms of computational efficiency, scalability, and superior performance. To optimize the retrieval performance, we advocate a tiered retrieval mechanism, blending both coarse-grain and fine-grain paradigms. Moreover, to enrich the contextual relationship within the query tuple, we integrate a generative cross-modal alignment technique, ensuring synchronization of sequential attributes between image and text data.
|
[
"Yang, Yuchen",
"Wang, Yu",
"Wang, Yanfeng"
] |
{CF}-{TCIR}: A Compositor-Free Framework for Hierarchical Text-Conditioned Image Retrieval
|
findings-acl.965
|
Poster
|
2309.08250v1
|
https://aclanthology.org/2024.findings-acl.966.bib
|
@inproceedings{huang-etal-2024-dmin,
title = "{DMIN}: A Discourse-specific Multi-granularity Integration Network for Conversational Aspect-based Sentiment Quadruple Analysis",
author = "Huang, Peijie and
Xiao, Xisheng and
Xu, Yuhong and
Chen, Jiawei",
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.966",
pages = "16326--16338",
abstract = "Conversational Aspect-based Sentiment Quadruple Analysis (DiaASQ) aims to extract fine-grained sentiment quadruples from dialogues. Previous research has primarily concentrated on enhancing token-level interactions, still lacking in sufficient modeling of the discourse structure information in dialogue. Firstly, it does not incorporate interactions among different utterances in the encoding stage, resulting in a limited token-level context understanding for subsequent modules. Secondly, it ignores the critical fact that discourse information is naturally organized at the utterance level and learning it solely at the token level is incomplete. In this work, we strengthen the token-level encoder by utilizing a discourse structure called {``}thread{''} and graph convolutional networks to enhance the token interaction among different utterances. Moreover, we propose an utterance-level encoder to learn the structured speaker and reply information, providing a macro understanding of dialogue discourse. Furthermore, we introduce a novel Multi-granularities Integrator to integrate token-level and utterance-level representations, resulting in a comprehensive and cohesive dialogue contextual understanding. Experiments on two datasets demonstrate that our model achieves state-of-the-art performance. Our codes are publicly available at https://github.com/SIGSDSscau/DMIN.",
}
|
Conversational Aspect-based Sentiment Quadruple Analysis (DiaASQ) aims to extract fine-grained sentiment quadruples from dialogues. Previous research has primarily concentrated on enhancing token-level interactions, still lacking in sufficient modeling of the discourse structure information in dialogue. Firstly, it does not incorporate interactions among different utterances in the encoding stage, resulting in a limited token-level context understanding for subsequent modules. Secondly, it ignores the critical fact that discourse information is naturally organized at the utterance level and learning it solely at the token level is incomplete. In this work, we strengthen the token-level encoder by utilizing a discourse structure called {``}thread{''} and graph convolutional networks to enhance the token interaction among different utterances. Moreover, we propose an utterance-level encoder to learn the structured speaker and reply information, providing a macro understanding of dialogue discourse. Furthermore, we introduce a novel Multi-granularities Integrator to integrate token-level and utterance-level representations, resulting in a comprehensive and cohesive dialogue contextual understanding. Experiments on two datasets demonstrate that our model achieves state-of-the-art performance. Our codes are publicly available at https://github.com/SIGSDSscau/DMIN.
|
[
"Huang, Peijie",
"Xiao, Xisheng",
"Xu, Yuhong",
"Chen, Jiawei"
] |
{DMIN}: A Discourse-specific Multi-granularity Integration Network for Conversational Aspect-based Sentiment Quadruple Analysis
|
findings-acl.966
|
Poster
|
2309.15476v1
|
https://aclanthology.org/2024.findings-acl.967.bib
|
@inproceedings{qorib-etal-2024-decoder,
title = "Are Decoder-Only Language Models Better than Encoder-Only Language Models in Understanding Word Meaning?",
author = "Qorib, Muhammad and
Moon, Geonsik and
Ng, Hwee Tou",
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.967",
pages = "16339--16347",
abstract = "The natural language processing field has been evolving around language models for the past few years, from the usage of n-gram language models for re-ranking, to transfer learning with encoder-only (BERT-like) language models, and finally to large language models (LLMs) as general solvers. LLMs are dominated by the decoder-only type, and they are popular for their efficacy in numerous tasks. LLMs are regarded as having strong comprehension abilities and strong capabilities to solve new unseen tasks. As such, people may quickly assume that decoder-only LLMs always perform better than the encoder-only ones, especially for understanding word meaning. In this paper, we demonstrate that decoder-only LLMs perform worse on word meaning comprehension than an encoder-only language model that has vastly fewer parameters.",
}
|
The natural language processing field has been evolving around language models for the past few years, from the usage of n-gram language models for re-ranking, to transfer learning with encoder-only (BERT-like) language models, and finally to large language models (LLMs) as general solvers. LLMs are dominated by the decoder-only type, and they are popular for their efficacy in numerous tasks. LLMs are regarded as having strong comprehension abilities and strong capabilities to solve new unseen tasks. As such, people may quickly assume that decoder-only LLMs always perform better than the encoder-only ones, especially for understanding word meaning. In this paper, we demonstrate that decoder-only LLMs perform worse on word meaning comprehension than an encoder-only language model that has vastly fewer parameters.
|
[
"Qorib, Muhammad",
"Moon, Geonsik",
"Ng, Hwee Tou"
] |
Are Decoder-Only Language Models Better than Encoder-Only Language Models in Understanding Word Meaning?
|
findings-acl.967
|
Poster
|
2305.17268v1
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.