bibtex_url
stringlengths 41
52
| proceedings
stringlengths 38
49
| bibtext
stringlengths 788
3.49k
| abstract
stringlengths 0
2.12k
| authors
sequencelengths 1
58
| title
stringlengths 16
181
| id
stringlengths 7
18
| type
stringclasses 2
values | arxiv_id
stringlengths 0
10
| GitHub
sequencelengths 1
1
| paper_page
stringclasses 170
values | n_linked_authors
int64 -1
9
| upvotes
int64 -1
56
| num_comments
int64 -1
9
| n_authors
int64 -1
57
| paper_page_exists_pre_conf
int64 0
1
| Models
sequencelengths 0
99
| Datasets
sequencelengths 0
5
| Spaces
sequencelengths 0
57
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
https://aclanthology.org/2024.findings-naacl.39.bib | https://aclanthology.org/2024.findings-naacl.39/ | @inproceedings{lyu-etal-2024-llm,
title = "{LLM}-Rec: Personalized Recommendation via Prompting Large Language Models",
author = "Lyu, Hanjia and
Jiang, Song and
Zeng, Hanqing and
Xia, Yinglong and
Wang, Qifan and
Zhang, Si and
Chen, Ren and
Leung, Chris and
Tang, Jiajie and
Luo, Jiebo",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.39",
doi = "10.18653/v1/2024.findings-naacl.39",
pages = "583--612",
abstract = "Text-based recommendation holds a wide range of practical applications due to its versatility, as textual descriptions can represent nearly any type of item. However, directly employing the original item descriptions may not yield optimal recommendation performance due to the lack of comprehensive information to align with user preferences. Recent advances in large language models (LLMs) have showcased their remarkable ability to harness commonsense knowledge and reasoning. In this study, we introduce a novel approach, coined LLM-Rec, which incorporates four distinct prompting strategies of text enrichment for improving personalized text-based recommendations. Our empirical experiments reveal that using LLM-augmented text significantly enhances recommendation quality. Even basic MLP (Multi-Layer Perceptron) models achieve comparable or even better results than complex content-based methods. Notably, the success of LLM-Rec lies in its prompting strategies, which effectively tap into the language model{'}s comprehension of both general and specific item characteristics. This highlights the importance of employing diverse prompts and input augmentation techniques to boost the recommendation effectiveness of LLMs.",
}
| Text-based recommendation holds a wide range of practical applications due to its versatility, as textual descriptions can represent nearly any type of item. However, directly employing the original item descriptions may not yield optimal recommendation performance due to the lack of comprehensive information to align with user preferences. Recent advances in large language models (LLMs) have showcased their remarkable ability to harness commonsense knowledge and reasoning. In this study, we introduce a novel approach, coined LLM-Rec, which incorporates four distinct prompting strategies of text enrichment for improving personalized text-based recommendations. Our empirical experiments reveal that using LLM-augmented text significantly enhances recommendation quality. Even basic MLP (Multi-Layer Perceptron) models achieve comparable or even better results than complex content-based methods. Notably, the success of LLM-Rec lies in its prompting strategies, which effectively tap into the language model{'}s comprehension of both general and specific item characteristics. This highlights the importance of employing diverse prompts and input augmentation techniques to boost the recommendation effectiveness of LLMs. | [
"Lyu, Hanjia",
"Jiang, Song",
"Zeng, Hanqing",
"Xia, Yinglong",
"Wang, Qifan",
"Zhang, Si",
"Chen, Ren",
"Leung, Chris",
"Tang, Jiajie",
"Luo, Jiebo"
] | LLM-Rec: Personalized Recommendation via Prompting Large Language Models | findings-naacl.39 | Poster | 2307.15780 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.40.bib | https://aclanthology.org/2024.findings-naacl.40/ | @inproceedings{ren-etal-2024-robust,
title = "A Robust Semantics-based Watermark for Large Language Model against Paraphrasing",
author = "Ren, Jie and
Xu, Han and
Liu, Yiding and
Cui, Yingqian and
Wang, Shuaiqiang and
Yin, Dawei and
Tang, Jiliang",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.40",
doi = "10.18653/v1/2024.findings-naacl.40",
pages = "613--625",
abstract = "Large language models (LLMs) have show their remarkable ability in various natural language tasks. However, there are concerns that LLMs are possible to be used improperly or even illegally. To prevent the malicious usage of LLMs, detecting LLM-generated text becomes crucial in the deployment of LLM applications. Watermarking is an effective strategy to detect the LLM-generated content by encoding a pre-defined secret watermark to facilitate the detection process. However, the majority of existing watermark methods leverage the simple hashes of precedent tokens to partition vocabulary. Such watermarks can be easily eliminated by paraphrase and, correspondingly, the detection effectiveness will be greatly compromised. Thus, to enhance the robustness against paraphrase, we propose a semantics-based watermark framework, SemaMark. It leverages the semantics as an alternative to simple hashes of tokens since the semantic meaning of the sentences will be likely preserved under paraphrase and the watermark can remain robust. Comprehensive experiments are conducted to demonstrate the effectiveness and robustness of SemaMark under different paraphrases.",
}
| Large language models (LLMs) have show their remarkable ability in various natural language tasks. However, there are concerns that LLMs are possible to be used improperly or even illegally. To prevent the malicious usage of LLMs, detecting LLM-generated text becomes crucial in the deployment of LLM applications. Watermarking is an effective strategy to detect the LLM-generated content by encoding a pre-defined secret watermark to facilitate the detection process. However, the majority of existing watermark methods leverage the simple hashes of precedent tokens to partition vocabulary. Such watermarks can be easily eliminated by paraphrase and, correspondingly, the detection effectiveness will be greatly compromised. Thus, to enhance the robustness against paraphrase, we propose a semantics-based watermark framework, SemaMark. It leverages the semantics as an alternative to simple hashes of tokens since the semantic meaning of the sentences will be likely preserved under paraphrase and the watermark can remain robust. Comprehensive experiments are conducted to demonstrate the effectiveness and robustness of SemaMark under different paraphrases. | [
"Ren, Jie",
"Xu, Han",
"Liu, Yiding",
"Cui, Yingqian",
"Wang, Shuaiqiang",
"Yin, Dawei",
"Tang, Jiliang"
] | A Robust Semantics-based Watermark for Large Language Model against Paraphrasing | findings-naacl.40 | Poster | 2311.08721 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.41.bib | https://aclanthology.org/2024.findings-naacl.41/ | @inproceedings{barke-etal-2024-solving,
title = "Solving Data-centric Tasks using Large Language Models",
author = "Barke, Shraddha and
Poelitz, Christian and
Negreanu, Carina and
Zorn, Benjamin and
Cambronero, Jos{\'e} and
Gordon, Andrew and
Le, Vu and
Nouri, Elnaz and
Polikarpova, Nadia and
Sarkar, Advait and
Slininger, Brian and
Toronto, Neil and
Williams, Jack",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.41",
doi = "10.18653/v1/2024.findings-naacl.41",
pages = "626--638",
abstract = "Large language models are rapidly replacing help forums like StackOverflow, and are especially helpful to non-professional programmers and end users. These users are often interested in \textit{data-centric tasks}, like spreadsheet manipulation and data wrangling, which are hard to solve if the intent is only communicated using a natural-language description, without including data. But how do we decide how much data and which data to include in the prompt?This paper makes two contributions towards answering this question. First, we create a dataset of real-world NL-to-code tasks manipulating tabular data, mined from StackOverflow posts. Second, we introduce a novel \textit{cluster-then-select} prompting technique, which adds the most representative rows from the input data to the LLM prompt. Our experiments show that LLM performance is indeed sensitive to the amount of data passed in the prompt, and that for tasks with a lot of syntactic variation in the input table,our cluster-then-select technique outperforms a random selection baseline.",
}
| Large language models are rapidly replacing help forums like StackOverflow, and are especially helpful to non-professional programmers and end users. These users are often interested in \textit{data-centric tasks}, like spreadsheet manipulation and data wrangling, which are hard to solve if the intent is only communicated using a natural-language description, without including data. But how do we decide how much data and which data to include in the prompt?This paper makes two contributions towards answering this question. First, we create a dataset of real-world NL-to-code tasks manipulating tabular data, mined from StackOverflow posts. Second, we introduce a novel \textit{cluster-then-select} prompting technique, which adds the most representative rows from the input data to the LLM prompt. Our experiments show that LLM performance is indeed sensitive to the amount of data passed in the prompt, and that for tasks with a lot of syntactic variation in the input table,our cluster-then-select technique outperforms a random selection baseline. | [
"Barke, Shraddha",
"Poelitz, Christian",
"Negreanu, Carina",
"Zorn, Benjamin",
"Cambronero, Jos{\\'e}",
"Gordon, Andrew",
"Le, Vu",
"Nouri, Elnaz",
"Polikarpova, Nadia",
"Sarkar, Advait",
"Slininger, Brian",
"Toronto, Neil",
"Williams, Jack"
] | Solving Data-centric Tasks using Large Language Models | findings-naacl.41 | Poster | 2402.11734 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.42.bib | https://aclanthology.org/2024.findings-naacl.42/ | @inproceedings{guo-etal-2024-novel,
title = "A Novel Paradigm Boosting Translation Capabilities of Large Language Models",
author = "Guo, Jiaxin and
Yang, Hao and
Li, Zongyao and
Wei, Daimeng and
Shang, Hengchao and
Chen, Xiaoyu",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.42",
doi = "10.18653/v1/2024.findings-naacl.42",
pages = "639--649",
abstract = "This paper presents a study on strategies to enhance the translation capabilities of large language models (LLMs) in the context of machine translation (MT) tasks. The paper proposes a novel paradigm consisting of three stages: Secondary Pre-training using Extensive Monolingual Data, Continual Pre-training with Interlinear Text Format Documents, and Leveraging Source-Language Consistent Instruction for Supervised Fine-Tuning. Previous research on LLMs focused on various strategies for supervised fine-tuning (SFT), but their effectiveness has been limited. While traditional machine translation approaches rely on vast amounts of parallel bilingual data, our paradigm highlights the importance of using smaller sets of high-quality bilingual data. We argue that the focus should be on augmenting LLMs{'} cross-lingual alignment abilities during pre-training rather than solely relying on extensive bilingual data during SFT. Experimental results conducted using the Llama2(CITATION)model, particularly on Chinese-Llama2(CITATION) after monolingual augmentation, demonstrate the improved translation capabilities of LLMs. A significant contribution of our approach lies in Stage2: Continual Pre-training with Interlinear Text Format Documents, which requires less than 1B training data, making our method highly efficient. Additionally, in Stage3, we observed that setting instructions consistent with the source language benefits the supervised fine-tuning process. Experimental results demonstrate that our approach surpasses previous work and achieves superior performance compared to models such as NLLB-54B(CITATION) and GPT3.5-text-davinci-003, despite having a significantly smaller parameter count of only 7B or 13B. This achievement establishes our method as a pioneering strategy in the field of machine translation.",
}
| This paper presents a study on strategies to enhance the translation capabilities of large language models (LLMs) in the context of machine translation (MT) tasks. The paper proposes a novel paradigm consisting of three stages: Secondary Pre-training using Extensive Monolingual Data, Continual Pre-training with Interlinear Text Format Documents, and Leveraging Source-Language Consistent Instruction for Supervised Fine-Tuning. Previous research on LLMs focused on various strategies for supervised fine-tuning (SFT), but their effectiveness has been limited. While traditional machine translation approaches rely on vast amounts of parallel bilingual data, our paradigm highlights the importance of using smaller sets of high-quality bilingual data. We argue that the focus should be on augmenting LLMs{'} cross-lingual alignment abilities during pre-training rather than solely relying on extensive bilingual data during SFT. Experimental results conducted using the Llama2(CITATION)model, particularly on Chinese-Llama2(CITATION) after monolingual augmentation, demonstrate the improved translation capabilities of LLMs. A significant contribution of our approach lies in Stage2: Continual Pre-training with Interlinear Text Format Documents, which requires less than 1B training data, making our method highly efficient. Additionally, in Stage3, we observed that setting instructions consistent with the source language benefits the supervised fine-tuning process. Experimental results demonstrate that our approach surpasses previous work and achieves superior performance compared to models such as NLLB-54B(CITATION) and GPT3.5-text-davinci-003, despite having a significantly smaller parameter count of only 7B or 13B. This achievement establishes our method as a pioneering strategy in the field of machine translation. | [
"Guo, Jiaxin",
"Yang, Hao",
"Li, Zongyao",
"Wei, Daimeng",
"Shang, Hengchao",
"Chen, Xiaoyu"
] | A Novel Paradigm Boosting Translation Capabilities of Large Language Models | findings-naacl.42 | Poster | 2403.11430 | [
""
] | https://huggingface.co/papers/2403.11430 | 0 | 0 | 0 | 6 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.43.bib | https://aclanthology.org/2024.findings-naacl.43/ | @inproceedings{yuan-etal-2024-measuring,
title = "Measuring Social Norms of Large Language Models",
author = "Yuan, Ye and
Tang, Kexin and
Shen, Jianhao and
Zhang, Ming and
Wang, Chenguang",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.43",
doi = "10.18653/v1/2024.findings-naacl.43",
pages = "650--699",
abstract = "We present a new challenge to examine whether large language models understand social norms. In contrast to existing datasets, our dataset requires a fundamental understanding of social norms to solve. Our dataset features the largest set of social norm skills, consisting of 402 skills and 12,383 questions covering a wide set of social norms ranging from opinions and arguments to culture and laws. We design our dataset according to the K-12 curriculum. This enables the direct comparison of the social understanding of large language models to humans, more specifically, elementary students. While prior work generates nearly random accuracy on our benchmark, recent large language models such as GPT3.5-Turbo and LLaMA2-Chat are able to improve the performance significantly, only slightly below human performance. We then propose a multi-agent framework based on large language models to improve the models{'} ability to understand social norms. This method further improves large language models to be on par with humans. Given the increasing adoption of large language models in real-world applications, our finding is particularly important and presents a unique direction for future improvements.",
}
| We present a new challenge to examine whether large language models understand social norms. In contrast to existing datasets, our dataset requires a fundamental understanding of social norms to solve. Our dataset features the largest set of social norm skills, consisting of 402 skills and 12,383 questions covering a wide set of social norms ranging from opinions and arguments to culture and laws. We design our dataset according to the K-12 curriculum. This enables the direct comparison of the social understanding of large language models to humans, more specifically, elementary students. While prior work generates nearly random accuracy on our benchmark, recent large language models such as GPT3.5-Turbo and LLaMA2-Chat are able to improve the performance significantly, only slightly below human performance. We then propose a multi-agent framework based on large language models to improve the models{'} ability to understand social norms. This method further improves large language models to be on par with humans. Given the increasing adoption of large language models in real-world applications, our finding is particularly important and presents a unique direction for future improvements. | [
"Yuan, Ye",
"Tang, Kexin",
"Shen, Jianhao",
"Zhang, Ming",
"Wang, Chenguang"
] | Measuring Social Norms of Large Language Models | findings-naacl.43 | Poster | 2404.02491 | [
""
] | https://huggingface.co/papers/2404.02491 | 0 | 0 | 0 | 5 | 1 | [] | [
"socialnormdataset/social"
] | [] |
https://aclanthology.org/2024.findings-naacl.44.bib | https://aclanthology.org/2024.findings-naacl.44/ | @inproceedings{yin-etal-2024-source,
title = "Source-Free Unsupervised Domain Adaptation for Question Answering via Prompt-Assisted Self-learning",
author = "Yin, Maxwell and
Wang, Boyu and
Ling, Charles",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.44",
doi = "10.18653/v1/2024.findings-naacl.44",
pages = "700--713",
abstract = "This work addresses source-free domain adaptation (SFDA) for Question Answering (QA), wherein a model trained on a source domain is adapted to unlabeled target domains without additional source data. Existing SFDA methods only focus on the adaptation phase, overlooking the impact of source domain training on model generalizability. In this paper, we argue that source model training itself is also critical for improving the adaptation performance and stability. To this end, we investigate the role of prompt learning as an effective method to internalize domain-agnostic QA knowledge, which can be integrated into source training. After source training, an interactive self-learning strategy is proposed to further fine tune both model and prompt in the model adaptation phase. This leads to the Prompt-Assisted Self-Adaptive Learning (PASAL), an innovative SFDA approach for QA. Empirical evaluation on four benchmark datasets shows that PASAL surpasses existing methods in managing domain gaps and demonstrates greater stability across various target domains, validating the significance of source domain training for effective domain adaptation.",
}
| This work addresses source-free domain adaptation (SFDA) for Question Answering (QA), wherein a model trained on a source domain is adapted to unlabeled target domains without additional source data. Existing SFDA methods only focus on the adaptation phase, overlooking the impact of source domain training on model generalizability. In this paper, we argue that source model training itself is also critical for improving the adaptation performance and stability. To this end, we investigate the role of prompt learning as an effective method to internalize domain-agnostic QA knowledge, which can be integrated into source training. After source training, an interactive self-learning strategy is proposed to further fine tune both model and prompt in the model adaptation phase. This leads to the Prompt-Assisted Self-Adaptive Learning (PASAL), an innovative SFDA approach for QA. Empirical evaluation on four benchmark datasets shows that PASAL surpasses existing methods in managing domain gaps and demonstrates greater stability across various target domains, validating the significance of source domain training for effective domain adaptation. | [
"Yin, Maxwell",
"Wang, Boyu",
"Ling, Charles"
] | Source-Free Unsupervised Domain Adaptation for Question Answering via Prompt-Assisted Self-learning | findings-naacl.44 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.45.bib | https://aclanthology.org/2024.findings-naacl.45/ | @inproceedings{zhao-etal-2024-hierarchical,
title = "Hierarchical Attention Graph for Scientific Document Summarization in Global and Local Level",
author = "Zhao, Chenlong and
Zhou, Xiwen and
Xie, Xiaopeng and
Zhang, Yong",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.45",
doi = "10.18653/v1/2024.findings-naacl.45",
pages = "714--726",
abstract = "Scientific document summarization has been a challenging task due to the long structure of the input text. The long input hinders the simultaneous effective modeling of both global high-order relations between sentences and local intra-sentence relations which is the most critical step in extractive summarization. However, existing methods mostly focus on one type of relation, neglecting the simultaneous effective modeling of both relations, which can lead to insufficient learning of semantic representations. In this paper, we propose HAESum, a novel approach utilizing graph neural networks to locally and globally model documents based on their hierarchical discourse structure. First, intra-sentence relations are learned using a local heterogeneous graph. Subsequently, a novel hypergraph self-attention layer is introduced to further enhance the characterization of high-order inter-sentence relations. We validate our approach on two benchmark datasets, and the experimental results demonstrate the effectiveness of HAESum and the importance of considering hierarchical structures in modeling long scientific documents.",
}
| Scientific document summarization has been a challenging task due to the long structure of the input text. The long input hinders the simultaneous effective modeling of both global high-order relations between sentences and local intra-sentence relations which is the most critical step in extractive summarization. However, existing methods mostly focus on one type of relation, neglecting the simultaneous effective modeling of both relations, which can lead to insufficient learning of semantic representations. In this paper, we propose HAESum, a novel approach utilizing graph neural networks to locally and globally model documents based on their hierarchical discourse structure. First, intra-sentence relations are learned using a local heterogeneous graph. Subsequently, a novel hypergraph self-attention layer is introduced to further enhance the characterization of high-order inter-sentence relations. We validate our approach on two benchmark datasets, and the experimental results demonstrate the effectiveness of HAESum and the importance of considering hierarchical structures in modeling long scientific documents. | [
"Zhao, Chenlong",
"Zhou, Xiwen",
"Xie, Xiaopeng",
"Zhang, Yong"
] | Hierarchical Attention Graph for Scientific Document Summarization in Global and Local Level | findings-naacl.45 | Poster | 2405.10202 | [
"https://github.com/molichenxi/haesum"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.46.bib | https://aclanthology.org/2024.findings-naacl.46/ | @inproceedings{kumar-dusek-2024-leeets,
title = "{LEEET}s-Dial: Linguistic Entrainment in End-to-End Task-oriented Dialogue systems",
author = "Kumar, Nalin and
Dusek, Ondrej",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.46",
doi = "10.18653/v1/2024.findings-naacl.46",
pages = "727--735",
abstract = "Linguistic entrainment, or alignment, represents a phenomenon where linguistic patterns employed by conversational participants converge to one another. While entrainment has been shown to produce a more natural user experience, most dialogue systems do not have any provisions for it. In this work, we introduce methods for achieving dialogue entrainment in a GPT-2-based end-to-end task-oriented dialogue system through the utilization of shared vocabulary. We experiment with training instance weighting, entrainment-specific loss, and additional conditioning to generate responses that align with the user. We demonstrate that all three approaches produce significantly better entrainment than the base, non-entrainment-optimized model, as confirmed by both automated and manual evaluation metrics.",
}
| Linguistic entrainment, or alignment, represents a phenomenon where linguistic patterns employed by conversational participants converge to one another. While entrainment has been shown to produce a more natural user experience, most dialogue systems do not have any provisions for it. In this work, we introduce methods for achieving dialogue entrainment in a GPT-2-based end-to-end task-oriented dialogue system through the utilization of shared vocabulary. We experiment with training instance weighting, entrainment-specific loss, and additional conditioning to generate responses that align with the user. We demonstrate that all three approaches produce significantly better entrainment than the base, non-entrainment-optimized model, as confirmed by both automated and manual evaluation metrics. | [
"Kumar, Nalin",
"Dusek, Ondrej"
] | LEEETs-Dial: Linguistic Entrainment in End-to-End Task-oriented Dialogue systems | findings-naacl.46 | Poster | 2311.09390 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.47.bib | https://aclanthology.org/2024.findings-naacl.47/ | @inproceedings{dobre-2024-efficient,
title = "Efficient Dependency Tree Sampling Without Replacement",
author = "Dobre, Bogdan",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.47",
doi = "10.18653/v1/2024.findings-naacl.47",
pages = "736--741",
abstract = "In the context of computational models of dependency syntax, most dependency treebanks have the restriction that any valid dependency tree must have exactly one edge coming out of the root node in addition to respecting the spanning tree constraints. Many algorithms for dependency tree sampling were recently proposed, both for sampling with and without replacement.In this paper we propose a new algorithm called Wilson Reject SWOR for the case of sampling without replacement by adapting the Wilson Reject algorithm originally created for sampling with replacement and combining it with a Trie data structure. Experimental results indicate the efficiency of our approach in the scenario of sampling without replacement from dependency graphs with random weights.",
}
| In the context of computational models of dependency syntax, most dependency treebanks have the restriction that any valid dependency tree must have exactly one edge coming out of the root node in addition to respecting the spanning tree constraints. Many algorithms for dependency tree sampling were recently proposed, both for sampling with and without replacement.In this paper we propose a new algorithm called Wilson Reject SWOR for the case of sampling without replacement by adapting the Wilson Reject algorithm originally created for sampling with replacement and combining it with a Trie data structure. Experimental results indicate the efficiency of our approach in the scenario of sampling without replacement from dependency graphs with random weights. | [
"Dobre, Bogdan"
] | Efficient Dependency Tree Sampling Without Replacement | findings-naacl.47 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.48.bib | https://aclanthology.org/2024.findings-naacl.48/ | @inproceedings{zhang-etal-2024-towards,
title = "Towards Better Generalization in Open-Domain Question Answering by Mitigating Context Memorization",
author = "Zhang, Zixuan and
Gangi Reddy, Revanth and
Small, Kevin and
Zhang, Tong and
Ji, Heng",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.48",
doi = "10.18653/v1/2024.findings-naacl.48",
pages = "742--753",
abstract = "Open-domain Question Answering (OpenQA) aims at answering factual questions with an external large-scale knowledge corpus. However, real-world knowledge is not static; it updates and evolves continually. Such a dynamic characteristic of knowledge poses a vital challenge for these models, as the trained models need to constantly adapt to the latest information to make sure that the answers remain accurate. In addition, it is still unclear how well an OpenQA model can transfer to completely new knowledge domains. In this paper, we investigate the generalization performance of a retrieval-augmented QA model in two specific scenarios: 1) adapting to updated versions of the same knowledge corpus; 2) switching to completely different knowledge domains. We observe that the generalization challenges of OpenQA models stem from the reader{'}s over-reliance on memorizing the knowledge from the external corpus, which hinders the model from generalizing to a new knowledge corpus. We introduce Corpus-Invariant Tuning (CIT), a simple but effective training strategy, to mitigate the knowledge over-memorization by controlling the likelihood of retrieved contexts during training. Extensive experimental results on multiple OpenQA benchmarks show that CIT achieves significantly better generalizability without compromising the model{'}s performance in its original corpus and domain.",
}
| Open-domain Question Answering (OpenQA) aims at answering factual questions with an external large-scale knowledge corpus. However, real-world knowledge is not static; it updates and evolves continually. Such a dynamic characteristic of knowledge poses a vital challenge for these models, as the trained models need to constantly adapt to the latest information to make sure that the answers remain accurate. In addition, it is still unclear how well an OpenQA model can transfer to completely new knowledge domains. In this paper, we investigate the generalization performance of a retrieval-augmented QA model in two specific scenarios: 1) adapting to updated versions of the same knowledge corpus; 2) switching to completely different knowledge domains. We observe that the generalization challenges of OpenQA models stem from the reader{'}s over-reliance on memorizing the knowledge from the external corpus, which hinders the model from generalizing to a new knowledge corpus. We introduce Corpus-Invariant Tuning (CIT), a simple but effective training strategy, to mitigate the knowledge over-memorization by controlling the likelihood of retrieved contexts during training. Extensive experimental results on multiple OpenQA benchmarks show that CIT achieves significantly better generalizability without compromising the model{'}s performance in its original corpus and domain. | [
"Zhang, Zixuan",
"Gangi Reddy, Revanth",
"Small, Kevin",
"Zhang, Tong",
"Ji, Heng"
] | Towards Better Generalization in Open-Domain Question Answering by Mitigating Context Memorization | findings-naacl.48 | Poster | 2404.01652 | [
""
] | https://huggingface.co/papers/2404.01652 | 0 | 0 | 0 | 5 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.49.bib | https://aclanthology.org/2024.findings-naacl.49/ | @inproceedings{song-etal-2024-gee,
title = "{GEE}! Grammar Error Explanation with Large Language Models",
author = "Song, Yixiao and
Krishna, Kalpesh and
Bhatt, Rajesh and
Gimpel, Kevin and
Iyyer, Mohit",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.49",
doi = "10.18653/v1/2024.findings-naacl.49",
pages = "754--781",
abstract = "Existing grammatical error correction tools do not provide natural language explanations of the errors that they correct in user-written text. However, such explanations are essential for helping users learn the language by gaining a deeper understanding of its grammatical rules (DeKeyser, 2003; Ellis et al., 2006).To address this gap, we propose the task of grammar error explanation, where a system needs to provide one-sentence explanations for each grammatical error in a pair of erroneous and corrected sentences. The task is not easily solved by prompting LLMs: we find that, using one-shot prompting, GPT-4 only explains 40.6{\%} of the errors and does not even attempt to explain 39.8{\%} of the errors.Since LLMs struggle to identify grammar errors, we develop a two-step pipeline that leverages fine-tuned and prompted large language models to perform structured atomic token edit extraction, followed by prompting GPT-4 to explain each edit. We evaluate our pipeline on German, Chinese, and English grammar error correction data. Our atomic edit extraction achieves an F1 of 0.93 on German, 0.91 on Chinese, and 0.891 on English. Human evaluation of generated explanations reveals that 93.9{\%} of German errors, 96.4{\%} of Chinese errors, and 92.20{\%} of English errors are correctly detected and explained. To encourage further research, we open-source our data and code.",
}
| Existing grammatical error correction tools do not provide natural language explanations of the errors that they correct in user-written text. However, such explanations are essential for helping users learn the language by gaining a deeper understanding of its grammatical rules (DeKeyser, 2003; Ellis et al., 2006).To address this gap, we propose the task of grammar error explanation, where a system needs to provide one-sentence explanations for each grammatical error in a pair of erroneous and corrected sentences. The task is not easily solved by prompting LLMs: we find that, using one-shot prompting, GPT-4 only explains 40.6{\%} of the errors and does not even attempt to explain 39.8{\%} of the errors.Since LLMs struggle to identify grammar errors, we develop a two-step pipeline that leverages fine-tuned and prompted large language models to perform structured atomic token edit extraction, followed by prompting GPT-4 to explain each edit. We evaluate our pipeline on German, Chinese, and English grammar error correction data. Our atomic edit extraction achieves an F1 of 0.93 on German, 0.91 on Chinese, and 0.891 on English. Human evaluation of generated explanations reveals that 93.9{\%} of German errors, 96.4{\%} of Chinese errors, and 92.20{\%} of English errors are correctly detected and explained. To encourage further research, we open-source our data and code. | [
"Song, Yixiao",
"Krishna, Kalpesh",
"Bhatt, Rajesh",
"Gimpel, Kevin",
"Iyyer, Mohit"
] | GEE! Grammar Error Explanation with Large Language Models | findings-naacl.49 | Poster | 2311.09517 | [
"https://github.com/yixiao-song/gee-with-llms"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.50.bib | https://aclanthology.org/2024.findings-naacl.50/ | @inproceedings{zhang-lu-2024-adarefiner,
title = "{A}da{R}efiner: Refining Decisions of Language Models with Adaptive Feedback",
author = "Zhang, Wanpeng and
Lu, Zongqing",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.50",
doi = "10.18653/v1/2024.findings-naacl.50",
pages = "782--799",
abstract = "Large Language Models (LLMs) have demonstrated significant success across various domains. However, their application in complex decision-making tasks frequently necessitates intricate prompt engineering or fine-tuning, leading to challenges in unseen downstream tasks and heavy demands on computational resources. Meanwhile, Reinforcement Learning (RL) has been recognized as effective in decision-making problems but struggles in environments with sparse rewards, such as open-world games. To overcome these challenges, we introduce AdaRefiner, a novel framework designed to enhance the synergy between LLMs and RL feedback. The key component of AdaRefiner is a lightweight Adapter Language Model (LM), which automatically refines task comprehension based on feedback from RL agents. This method mitigates the need for intricate prompt engineering and intensive LLM fine-tuning while maintaining the LLMs{'} generalization abilities and enhancing their decision-making capabilities in downstream tasks. Empirical evaluations of AdaRefiner on 22 diverse tasks within the open-world game \textit{Crafter} have demonstrated its superior effectiveness, especially in guiding agents towards higher-level and common-sense skills. Our work makes contributions to the automatic self-refinement of LLMs with RL feedback, offering a more adaptable and efficient solution for complex decision-making problems. The code is available at https://github.com/PKU-RL/AdaRefiner.",
}
| Large Language Models (LLMs) have demonstrated significant success across various domains. However, their application in complex decision-making tasks frequently necessitates intricate prompt engineering or fine-tuning, leading to challenges in unseen downstream tasks and heavy demands on computational resources. Meanwhile, Reinforcement Learning (RL) has been recognized as effective in decision-making problems but struggles in environments with sparse rewards, such as open-world games. To overcome these challenges, we introduce AdaRefiner, a novel framework designed to enhance the synergy between LLMs and RL feedback. The key component of AdaRefiner is a lightweight Adapter Language Model (LM), which automatically refines task comprehension based on feedback from RL agents. This method mitigates the need for intricate prompt engineering and intensive LLM fine-tuning while maintaining the LLMs{'} generalization abilities and enhancing their decision-making capabilities in downstream tasks. Empirical evaluations of AdaRefiner on 22 diverse tasks within the open-world game \textit{Crafter} have demonstrated its superior effectiveness, especially in guiding agents towards higher-level and common-sense skills. Our work makes contributions to the automatic self-refinement of LLMs with RL feedback, offering a more adaptable and efficient solution for complex decision-making problems. The code is available at https://github.com/PKU-RL/AdaRefiner. | [
"Zhang, Wanpeng",
"Lu, Zongqing"
] | AdaRefiner: Refining Decisions of Language Models with Adaptive Feedback | findings-naacl.50 | Poster | 2309.17176 | [
"https://github.com/pku-rl/adarefiner"
] | https://huggingface.co/papers/2309.17176 | 0 | 1 | 0 | 2 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.51.bib | https://aclanthology.org/2024.findings-naacl.51/ | @inproceedings{zeng-etal-2024-divtod,
title = "{D}iv{TOD}: Unleashing the Power of {LLM}s for Diversifying Task-Oriented Dialogue Representations",
author = "Zeng, Weihao and
Fu, Dayuan and
He, Keqing and
Wang, Yejie and
Xu, Yukai and
Xu, Weiran",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.51",
doi = "10.18653/v1/2024.findings-naacl.51",
pages = "800--813",
abstract = "Language models pre-trained on general text have achieved impressive results in diverse fields. Yet, the distinct linguistic characteristics of task-oriented dialogues (TOD) compared to general text limit the practical utility of existing language models. Current task-oriented dialogue pre-training methods overlook the one-to-many property of conversations, where multiple responses can be appropriate given the same conversation context.In this paper, we propose a novel dialogue pre-training model called DivTOD, which collaborates with LLMs to learn diverse task-oriented dialogue representations. DivTOD guides LLMs in transferring diverse knowledge to smaller models while removing domain knowledge that contradicts task-oriented dialogues. Experiments show that our model outperforms strong TOD baselines on various downstream dialogue tasks and learns the intrinsic diversity of task-oriented dialogues.",
}
| Language models pre-trained on general text have achieved impressive results in diverse fields. Yet, the distinct linguistic characteristics of task-oriented dialogues (TOD) compared to general text limit the practical utility of existing language models. Current task-oriented dialogue pre-training methods overlook the one-to-many property of conversations, where multiple responses can be appropriate given the same conversation context.In this paper, we propose a novel dialogue pre-training model called DivTOD, which collaborates with LLMs to learn diverse task-oriented dialogue representations. DivTOD guides LLMs in transferring diverse knowledge to smaller models while removing domain knowledge that contradicts task-oriented dialogues. Experiments show that our model outperforms strong TOD baselines on various downstream dialogue tasks and learns the intrinsic diversity of task-oriented dialogues. | [
"Zeng, Weihao",
"Fu, Dayuan",
"He, Keqing",
"Wang, Yejie",
"Xu, Yukai",
"Xu, Weiran"
] | DivTOD: Unleashing the Power of LLMs for Diversifying Task-Oriented Dialogue Representations | findings-naacl.51 | Poster | 2404.00557 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.52.bib | https://aclanthology.org/2024.findings-naacl.52/ | @inproceedings{denisov-vu-2024-teaching,
title = "Teaching a Multilingual Large Language Model to Understand Multilingual Speech via Multi-Instructional Training",
author = "Denisov, Pavel and
Vu, Thang",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.52",
doi = "10.18653/v1/2024.findings-naacl.52",
pages = "814--834",
abstract = "Recent advancements in language modeling have led to the emergenceof Large Language Models (LLMs) capable ofvarious natural language processing tasks.Despite their success in text-based tasks, applying LLMs to the speech domainremains limited and challenging. This paper presents BLOOMZMMS, a novel modelthat integrates a multilingual LLM with a multilingual speech encoder,aiming to harness the capabilities of LLMs for speech recognition and beyond.Utilizing a multi-instructional training approach, we demonstrate the transferabilityof linguistic knowledge from the text to the speech modality.Our experiments, conducted on 1900 hours of transcribed data from 139 languages,establish that a multilingual speech representation can be effectivelylearned and aligned with a multilingual LLM. While this learned representationinitially shows limitations in task generalization, we address this issue bygenerating synthetic targets in a multi-instructional style.Our zero-shot evaluation results confirm the robustness of our approach acrossmultiple tasks, including speech translation and multilingual spoken languageunderstanding, thereby opening new avenues for applying LLMs in the speech domain.",
}
| Recent advancements in language modeling have led to the emergenceof Large Language Models (LLMs) capable ofvarious natural language processing tasks.Despite their success in text-based tasks, applying LLMs to the speech domainremains limited and challenging. This paper presents BLOOMZMMS, a novel modelthat integrates a multilingual LLM with a multilingual speech encoder,aiming to harness the capabilities of LLMs for speech recognition and beyond.Utilizing a multi-instructional training approach, we demonstrate the transferabilityof linguistic knowledge from the text to the speech modality.Our experiments, conducted on 1900 hours of transcribed data from 139 languages,establish that a multilingual speech representation can be effectivelylearned and aligned with a multilingual LLM. While this learned representationinitially shows limitations in task generalization, we address this issue bygenerating synthetic targets in a multi-instructional style.Our zero-shot evaluation results confirm the robustness of our approach acrossmultiple tasks, including speech translation and multilingual spoken languageunderstanding, thereby opening new avenues for applying LLMs in the speech domain. | [
"Denisov, Pavel",
"Vu, Thang"
] | Teaching a Multilingual Large Language Model to Understand Multilingual Speech via Multi-Instructional Training | findings-naacl.52 | Poster | 2404.10922 | [
"https://github.com/akreal/bloomzmms"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.53.bib | https://aclanthology.org/2024.findings-naacl.53/ | @inproceedings{zhu-etal-2024-clean,
title = "{CLEAN}{--}{EVAL}: Clean Evaluation on Contaminated Large Language Models",
author = "Zhu, Wenhong and
Hao, Hongkun and
He, Zhiwei and
Song, Yun-Ze and
Yueyang, Jiao and
Zhang, Yumeng and
Hu, Hanxu and
Wei, Yiran and
Wang, Rui and
Lu, Hongyuan",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.53",
doi = "10.18653/v1/2024.findings-naacl.53",
pages = "835--847",
abstract = "We are currently in an era of fierce competition among various large language models (LLMs), continuously pushing the boundaries of benchmark performance. However, genuinely assessing the capabilities of these LLMs has become a challenging and critical issue due to potential data contamination. In this paper, we propose a novel and valuable method, Clean-Eval, which mitigates the issue of data contamination and evaluates the LLMs more cleanly. Clean-Eval employs a neural-based model to paraphrase and back-translate the contaminated data into a candidate set, generating expressions with the same meaning but in different surface forms. A semantic detector is then used to filter those generated low-quality samples to narrow down this candidate set. Candidates with moderate BLEURT scores against the original samples are selected as the final evaluation set. According to human assessment, this set is almost semantically equivalent to the original contamination set but expressed differently. We conduct experiments on 20 existing benchmarks across diverse tasks, and results demonstrate that Clean-Eval substantially restores the actual evaluation results on contaminated LLMs under both few-shot learning and fine-tuning scenarios.",
}
| We are currently in an era of fierce competition among various large language models (LLMs), continuously pushing the boundaries of benchmark performance. However, genuinely assessing the capabilities of these LLMs has become a challenging and critical issue due to potential data contamination. In this paper, we propose a novel and valuable method, Clean-Eval, which mitigates the issue of data contamination and evaluates the LLMs more cleanly. Clean-Eval employs a neural-based model to paraphrase and back-translate the contaminated data into a candidate set, generating expressions with the same meaning but in different surface forms. A semantic detector is then used to filter those generated low-quality samples to narrow down this candidate set. Candidates with moderate BLEURT scores against the original samples are selected as the final evaluation set. According to human assessment, this set is almost semantically equivalent to the original contamination set but expressed differently. We conduct experiments on 20 existing benchmarks across diverse tasks, and results demonstrate that Clean-Eval substantially restores the actual evaluation results on contaminated LLMs under both few-shot learning and fine-tuning scenarios. | [
"Zhu, Wenhong",
"Hao, Hongkun",
"He, Zhiwei",
"Song, Yun-Ze",
"Yueyang, Jiao",
"Zhang, Yumeng",
"Hu, Hanxu",
"Wei, Yiran",
"Wang, Rui",
"Lu, Hongyuan"
] | CLEAN–EVAL: Clean Evaluation on Contaminated Large Language Models | findings-naacl.53 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.54.bib | https://aclanthology.org/2024.findings-naacl.54/ | @inproceedings{sharma-etal-2024-r,
title = "{R}-{BASS} : Relevance-aided Block-wise Adaptation for Speech Summarization",
author = "Sharma, Roshan and
Sharma, Ruchira and
Dhamyal, Hira and
Singh, Rita and
Raj, Bhiksha",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.54",
doi = "10.18653/v1/2024.findings-naacl.54",
pages = "848--857",
abstract = "End-to-end speech summarization on long recordings is challenging because of the high computational cost. Block-wise Adaptation for Speech Summarization (BASS) summarizes arbitrarily long sequences by sequentially processing abutting chunks of audio. Despite the benefits of BASS, it has higher compute time due to sequential processing of all blocks, regardless of whether they are relevant to the final summary. In this paper, we propose R-BASS, a new relevance-aware block-wise adaptation method. First, we introduce two approaches to automatically estimate block relevance based on lexical and semantic similarity between the block-level transcript and the summary. Experiments on the How2 dataset show that using ground truth relevance during inference improves efficiency by 63.9 {\%} by dropping irrelevant blocks. Finally, we incorporate relevance scores into training using a novel relevance loss and relevance predictor, and the proposed R-BASS model makes it possible to drop 86.3 {\%} of the blocks while retaining comparable performance, resulting in a 2.2x speedup over BASS.",
}
| End-to-end speech summarization on long recordings is challenging because of the high computational cost. Block-wise Adaptation for Speech Summarization (BASS) summarizes arbitrarily long sequences by sequentially processing abutting chunks of audio. Despite the benefits of BASS, it has higher compute time due to sequential processing of all blocks, regardless of whether they are relevant to the final summary. In this paper, we propose R-BASS, a new relevance-aware block-wise adaptation method. First, we introduce two approaches to automatically estimate block relevance based on lexical and semantic similarity between the block-level transcript and the summary. Experiments on the How2 dataset show that using ground truth relevance during inference improves efficiency by 63.9 {\%} by dropping irrelevant blocks. Finally, we incorporate relevance scores into training using a novel relevance loss and relevance predictor, and the proposed R-BASS model makes it possible to drop 86.3 {\%} of the blocks while retaining comparable performance, resulting in a 2.2x speedup over BASS. | [
"Sharma, Roshan",
"Sharma, Ruchira",
"Dhamyal, Hira",
"Singh, Rita",
"Raj, Bhiksha"
] | R-BASS : Relevance-aided Block-wise Adaptation for Speech Summarization | findings-naacl.54 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.55.bib | https://aclanthology.org/2024.findings-naacl.55/ | @inproceedings{yu-etal-2024-ovm,
title = "{OVM}, Outcome-supervised Value Models for Planning in Mathematical Reasoning",
author = "Yu, Fei and
Gao, Anningzhe and
Wang, Benyou",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.55",
doi = "10.18653/v1/2024.findings-naacl.55",
pages = "858--875",
abstract = "Large language models (LLMs) often struggle with maintaining accuracy throughout multiple multiple reasoning steps, especially in mathematical reasoning where an error in earlier steps can propagate to subsequent ones and it ultimately leading to an incorrect answer.To reduce error propagation, guided decoding is employed to direct the LM decoding on a step-by-step basis. We argue that in guided decoding, assessing the potential of an incomplete reasoning path can be more advantageous than simply ensuring per-step correctness, as the former approach leads towards a correct final answer. This transforms the task into a $\textit{value estimation}$ problem in planning.Inspired by the findings that $\textit{outcome supervision for guided decoding essentially acts as a value model}$, we propose Outcome-supervised Value Model (OVM) that employs outcome supervision for training a value model, which prioritizes steps that lead to accurate conclusions. Furthermore, the OVM eliminates the need for labor-intensive annotations of step-level correctness, thereby significantly enhancing its scalability. Our experiments on two multi-step mathematical reasoning datasets, GSM8K and Game of 24, demonstrate the superior performance of the OVM model. Notably, in GSM8K, our $\textbf{OVM-7B model achieves state-of-the-art results among LLMs up to 13B parameters}$; especially it does not utilize GPT-4 or code execution. These findings offer a novel perspective on the role of outcome supervision in training value models for multi-step reasoning tasks and provide theoretical justification for its advantage in value estimation for guided decoding.",
}
| Large language models (LLMs) often struggle with maintaining accuracy throughout multiple multiple reasoning steps, especially in mathematical reasoning where an error in earlier steps can propagate to subsequent ones and it ultimately leading to an incorrect answer.To reduce error propagation, guided decoding is employed to direct the LM decoding on a step-by-step basis. We argue that in guided decoding, assessing the potential of an incomplete reasoning path can be more advantageous than simply ensuring per-step correctness, as the former approach leads towards a correct final answer. This transforms the task into a $\textit{value estimation}$ problem in planning.Inspired by the findings that $\textit{outcome supervision for guided decoding essentially acts as a value model}$, we propose Outcome-supervised Value Model (OVM) that employs outcome supervision for training a value model, which prioritizes steps that lead to accurate conclusions. Furthermore, the OVM eliminates the need for labor-intensive annotations of step-level correctness, thereby significantly enhancing its scalability. Our experiments on two multi-step mathematical reasoning datasets, GSM8K and Game of 24, demonstrate the superior performance of the OVM model. Notably, in GSM8K, our $\textbf{OVM-7B model achieves state-of-the-art results among LLMs up to 13B parameters}$; especially it does not utilize GPT-4 or code execution. These findings offer a novel perspective on the role of outcome supervision in training value models for multi-step reasoning tasks and provide theoretical justification for its advantage in value estimation for guided decoding. | [
"Yu, Fei",
"Gao, Anningzhe",
"Wang, Benyou"
] | OVM, Outcome-supervised Value Models for Planning in Mathematical Reasoning | findings-naacl.55 | Poster | 2311.09724 | [
"https://github.com/freedomintelligence/ovm"
] | https://huggingface.co/papers/2311.09724 | 1 | 0 | 0 | 3 | 1 | [
"FreedomIntelligence/OVM-Mistral-7b",
"FreedomIntelligence/OVM-llama2-7b",
"longhoang06/OVM-generator"
] | [
"FreedomIntelligence/OVM-dataset",
"FreedomIntelligence/OVM-process"
] | [] |
https://aclanthology.org/2024.findings-naacl.56.bib | https://aclanthology.org/2024.findings-naacl.56/ | @inproceedings{wang-lim-2024-whole,
title = "The Whole is Better than the Sum: Using Aggregated Demonstrations in In-Context Learning for Sequential Recommendation",
author = "Wang, Lei and
Lim, Ee-Peng",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.56",
doi = "10.18653/v1/2024.findings-naacl.56",
pages = "876--895",
abstract = "Large language models (LLMs) have shown excellent performance on various NLP tasks. To use LLMs as strong sequential recommenders, we explore the in-context learning approach to sequential recommendation. We investigate the effects of instruction format, task consistency, demonstration selection, and number of demonstrations. As increasing the number of demonstrations in ICL does not improve accuracy despite using a long prompt, we propose a novel method called LLMSRec-Syn that incorporates multiple demonstration users into one aggregated demonstration. Our experiments on three recommendation datasets show that LLMSRec-Syn outperforms state-of-the-art LLM-based sequential recommendation methods. In some cases, LLMSRec-Syn can perform on par with or even better than supervised learning methods. Our code is publicly available at https://github.com/demoleiwang/LLMSRec{\_}Syn.",
}
| Large language models (LLMs) have shown excellent performance on various NLP tasks. To use LLMs as strong sequential recommenders, we explore the in-context learning approach to sequential recommendation. We investigate the effects of instruction format, task consistency, demonstration selection, and number of demonstrations. As increasing the number of demonstrations in ICL does not improve accuracy despite using a long prompt, we propose a novel method called LLMSRec-Syn that incorporates multiple demonstration users into one aggregated demonstration. Our experiments on three recommendation datasets show that LLMSRec-Syn outperforms state-of-the-art LLM-based sequential recommendation methods. In some cases, LLMSRec-Syn can perform on par with or even better than supervised learning methods. Our code is publicly available at https://github.com/demoleiwang/LLMSRec{\_}Syn. | [
"Wang, Lei",
"Lim, Ee-Peng"
] | The Whole is Better than the Sum: Using Aggregated Demonstrations in In-Context Learning for Sequential Recommendation | findings-naacl.56 | Poster | 2403.10135 | [
"https://github.com/demoleiwang/llmsrec_syn"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.57.bib | https://aclanthology.org/2024.findings-naacl.57/ | @inproceedings{agarwal-etal-2024-bring,
title = "Bring Your Own {KG}: Self-Supervised Program Synthesis for Zero-Shot {KGQA}",
author = "Agarwal, Dhruv and
Das, Rajarshi and
Khosla, Sopan and
Gangadharaiah, Rashmi",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.57",
doi = "10.18653/v1/2024.findings-naacl.57",
pages = "896--919",
abstract = "We present BYOKG, a universal question-answering (QA) system that can operate on any knowledge graph (KG), requires no human-annotated training data, and can be ready to use within a day{---}attributes that are out-of-scope for current KGQA systems. BYOKG draws inspiration from the remarkable ability of humans to comprehend information present in an unseen KG through exploration{---}starting at random nodes, inspecting the labels of adjacent nodes and edges, and combining them with their prior world knowledge. Exploration in BYOKG leverages an LLM-backed symbolic agent that generates a diverse set of query-program exemplars, which are then used to ground a retrieval-augmented reasoning procedure to synthesize programs for arbitrary questions. BYOKG is effective over both small- and large-scale graphs, showing dramatic gains in zero-shot QA accuracy of 27.89 and 59.88 F1 on GrailQA and MetaQA, respectively. We further find that performance of BYOKG reliably improves with continued exploration as well as improvements in the base LLM, notably outperforming a state-of-the-art fine-tuned model by 7.08 F1 on a sub-sampled zero-shot split of GrailQA. Lastly, we verify our universality claim by evaluating BYOKG on a domain-specific materials science KG and show that it improves zero-shot performance by 46.33 F1.",
}
| We present BYOKG, a universal question-answering (QA) system that can operate on any knowledge graph (KG), requires no human-annotated training data, and can be ready to use within a day{---}attributes that are out-of-scope for current KGQA systems. BYOKG draws inspiration from the remarkable ability of humans to comprehend information present in an unseen KG through exploration{---}starting at random nodes, inspecting the labels of adjacent nodes and edges, and combining them with their prior world knowledge. Exploration in BYOKG leverages an LLM-backed symbolic agent that generates a diverse set of query-program exemplars, which are then used to ground a retrieval-augmented reasoning procedure to synthesize programs for arbitrary questions. BYOKG is effective over both small- and large-scale graphs, showing dramatic gains in zero-shot QA accuracy of 27.89 and 59.88 F1 on GrailQA and MetaQA, respectively. We further find that performance of BYOKG reliably improves with continued exploration as well as improvements in the base LLM, notably outperforming a state-of-the-art fine-tuned model by 7.08 F1 on a sub-sampled zero-shot split of GrailQA. Lastly, we verify our universality claim by evaluating BYOKG on a domain-specific materials science KG and show that it improves zero-shot performance by 46.33 F1. | [
"Agarwal, Dhruv",
"Das, Rajarshi",
"Khosla, Sopan",
"Gangadharaiah, Rashmi"
] | Bring Your Own KG: Self-Supervised Program Synthesis for Zero-Shot KGQA | findings-naacl.57 | Poster | 2311.07850 | [
"https://github.com/dhdhagar/byokg"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.58.bib | https://aclanthology.org/2024.findings-naacl.58/ | @inproceedings{yuan-farber-2024-grasame,
title = "{G}ra{SAME}: Injecting Token-Level Structural Information to Pretrained Language Models via Graph-guided Self-Attention Mechanism",
author = {Yuan, Shuzhou and
F{\"a}rber, Michael},
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.58",
doi = "10.18653/v1/2024.findings-naacl.58",
pages = "920--933",
abstract = "Pretrained Language Models (PLMs) benefit from external knowledge stored in graph structures for various downstream tasks. However, bridging the modality gap between graph structures and text remains a significant challenge. Traditional methods like linearizing graphs for PLMs lose vital graph connectivity, whereas Graph Neural Networks (GNNs) require cumbersome processes for integration into PLMs. In this work, we propose a novel graph-guided self-attention mechanism, GraSAME. GraSAME seamlessly incorporates token-level structural information into PLMs without necessitating additional alignment or concatenation efforts. As an end-to-end, lightweight multimodal module, GraSAME follows a multi-task learning strategy and effectively bridges the gap between graph and textual modalities, facilitating dynamic interactions between GNNs and PLMs. Our experiments on the graph-to-text generation task demonstrate that GraSAME outperforms baseline models and achieves results comparable to state-of-the-art (SOTA) models on WebNLG datasets. Furthermore, compared to SOTA models, GraSAME eliminates the need for extra pre-training tasks to adjust graph inputs and reduces the number of trainable parameters by over 100 million.",
}
| Pretrained Language Models (PLMs) benefit from external knowledge stored in graph structures for various downstream tasks. However, bridging the modality gap between graph structures and text remains a significant challenge. Traditional methods like linearizing graphs for PLMs lose vital graph connectivity, whereas Graph Neural Networks (GNNs) require cumbersome processes for integration into PLMs. In this work, we propose a novel graph-guided self-attention mechanism, GraSAME. GraSAME seamlessly incorporates token-level structural information into PLMs without necessitating additional alignment or concatenation efforts. As an end-to-end, lightweight multimodal module, GraSAME follows a multi-task learning strategy and effectively bridges the gap between graph and textual modalities, facilitating dynamic interactions between GNNs and PLMs. Our experiments on the graph-to-text generation task demonstrate that GraSAME outperforms baseline models and achieves results comparable to state-of-the-art (SOTA) models on WebNLG datasets. Furthermore, compared to SOTA models, GraSAME eliminates the need for extra pre-training tasks to adjust graph inputs and reduces the number of trainable parameters by over 100 million. | [
"Yuan, Shuzhou",
"F{\\\"a}rber, Michael"
] | GraSAME: Injecting Token-Level Structural Information to Pretrained Language Models via Graph-guided Self-Attention Mechanism | findings-naacl.58 | Poster | 2404.06911 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.59.bib | https://aclanthology.org/2024.findings-naacl.59/ | @inproceedings{wang-etal-2024-public,
title = "Can Public Large Language Models Help Private Cross-device Federated Learning?",
author = "Wang, Boxin and
Zhang, Yibo and
Cao, Yuan and
Li, Bo and
McMahan, Hugh and
Oh, Sewoong and
Xu, Zheng and
Zaheer, Manzil",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.59",
doi = "10.18653/v1/2024.findings-naacl.59",
pages = "934--949",
abstract = "We study (differentially) private federated learning (FL) of language models. The language models in cross-device FL are relatively small, which can be trained with meaningful formal user-level differential privacy (DP) guarantees when massive parallelism in training is enabled by the participation of a moderate size of users. Recently, public data has been used to improve privacy-utility trade-offs for both large and small language models. In this work, we provide a systematic study of using large-scale public data and LLMs to help differentially private training of on-device FL models, and further improve the privacy-utility tradeoff by techniques of distillation. Moreover, we propose a novel distribution matching algorithm with theoretical grounding to sample public data close to private data distribution, which significantly improves the sample efficiency of (pre-)training on public data. The proposed method is efficient and effective for training private models by taking advantage of public data, especially for customized on-device architectures that do not have ready-touse pre-trained models.",
}
| We study (differentially) private federated learning (FL) of language models. The language models in cross-device FL are relatively small, which can be trained with meaningful formal user-level differential privacy (DP) guarantees when massive parallelism in training is enabled by the participation of a moderate size of users. Recently, public data has been used to improve privacy-utility trade-offs for both large and small language models. In this work, we provide a systematic study of using large-scale public data and LLMs to help differentially private training of on-device FL models, and further improve the privacy-utility tradeoff by techniques of distillation. Moreover, we propose a novel distribution matching algorithm with theoretical grounding to sample public data close to private data distribution, which significantly improves the sample efficiency of (pre-)training on public data. The proposed method is efficient and effective for training private models by taking advantage of public data, especially for customized on-device architectures that do not have ready-touse pre-trained models. | [
"Wang, Boxin",
"Zhang, Yibo",
"Cao, Yuan",
"Li, Bo",
"McMahan, Hugh",
"Oh, Sewoong",
"Xu, Zheng",
"Zaheer, Manzil"
] | Can Public Large Language Models Help Private Cross-device Federated Learning? | findings-naacl.59 | Poster | 2305.12132 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.60.bib | https://aclanthology.org/2024.findings-naacl.60/ | @inproceedings{pan-etal-2024-langnav,
title = "{L}ang{N}av: Language as a Perceptual Representation for Navigation",
author = "Pan, Bowen and
Panda, Rameswar and
Jin, SouYoung and
Feris, Rogerio and
Oliva, Aude and
Isola, Phillip and
Kim, Yoon",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.60",
doi = "10.18653/v1/2024.findings-naacl.60",
pages = "950--974",
abstract = "We explore the use of language as a perceptual representation for vision-and-language navigation (VLN), with a focus on low-data settings. Our approach uses off-the-shelf vision systems for image captioning and object detection to convert an agent{'}s egocentric panoramic view at each time step into natural language descriptions. We then finetune a pretrained language model to select an action, based on the current view and the trajectory history, that would best fulfill the navigation instructions. In contrast to the standard setup which adapts a pretrained language model to work directly with continuous visual features from pretrained vision models, our approach instead uses (discrete) language as the perceptual representation. We explore several use cases of our language-based navigation (LangNav) approach on the R2R VLN benchmark: generating synthetic trajectories from a prompted language model (GPT-4) with which to finetune a smaller language model; domain transfer where we transfer a policy learned on one simulated environment (ALFRED) to another (more realistic) environment (R2R); and combining both vision- and language-based representations for VLN. Our approach is found to improve upon baselines that rely on visual features in settings where only a few expert trajectories (10-100) are available, demonstrating the potential of language as a perceptual representation for navigation.",
}
| We explore the use of language as a perceptual representation for vision-and-language navigation (VLN), with a focus on low-data settings. Our approach uses off-the-shelf vision systems for image captioning and object detection to convert an agent{'}s egocentric panoramic view at each time step into natural language descriptions. We then finetune a pretrained language model to select an action, based on the current view and the trajectory history, that would best fulfill the navigation instructions. In contrast to the standard setup which adapts a pretrained language model to work directly with continuous visual features from pretrained vision models, our approach instead uses (discrete) language as the perceptual representation. We explore several use cases of our language-based navigation (LangNav) approach on the R2R VLN benchmark: generating synthetic trajectories from a prompted language model (GPT-4) with which to finetune a smaller language model; domain transfer where we transfer a policy learned on one simulated environment (ALFRED) to another (more realistic) environment (R2R); and combining both vision- and language-based representations for VLN. Our approach is found to improve upon baselines that rely on visual features in settings where only a few expert trajectories (10-100) are available, demonstrating the potential of language as a perceptual representation for navigation. | [
"Pan, Bowen",
"P",
"a, Rameswar",
"Jin, SouYoung",
"Feris, Rogerio",
"Oliva, Aude",
"Isola, Phillip",
"Kim, Yoon"
] | LangNav: Language as a Perceptual Representation for Navigation | findings-naacl.60 | Poster | 2310.07889 | [
""
] | https://huggingface.co/papers/2310.07889 | 4 | 5 | 1 | 7 | 1 | [
"bpan/LangNav-Sim2k-Llama2"
] | [] | [] |
https://aclanthology.org/2024.findings-naacl.61.bib | https://aclanthology.org/2024.findings-naacl.61/ | @inproceedings{huang-etal-2024-planning,
title = "Planning and Editing What You Retrieve for Enhanced Tool Learning",
author = "Huang, Tenghao and
Jung, Dongwon and
Kumar, Vaibhav and
Kachuee, Mohammad and
Li, Xiang and
Xu, Puyang and
Chen, Muhao",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.61",
doi = "10.18653/v1/2024.findings-naacl.61",
pages = "975--988",
abstract = "Recent advancements in integrating external tools with Large Language Models (LLMs) have opened new frontiers, with applications in mathematical reasoning, code generators, and smart assistants. However, existing methods, relying on simple one-time retrieval strategies, fall short on effectively and accurately shortlisting relevant tools. This paper introduces a novel PLUTO (Planning, Learning, and Understanding for TOols) approach, encompassing {``}Plan-and-Retrieve (P{\&}R){''} and {``}Edit-and-Ground (E{\&}G){''} paradigms. The P{\&}R paradigm consists of a neural retrieval module for shortlisting relevant tools and an LLM-based query planner that decomposes complex queries into actionable tasks, enhancing the effectiveness of tool utilization. The E{\&}G paradigm utilizes LLMs to enrich tool descriptions based on user scenarios, bridging the gap between user queries and tool functionalities. Experiment results demonstrate that these paradigms significantly improve the recall and NDCG in tool retrieval tasks, significantly surpassing current state-of-the-art models.",
}
| Recent advancements in integrating external tools with Large Language Models (LLMs) have opened new frontiers, with applications in mathematical reasoning, code generators, and smart assistants. However, existing methods, relying on simple one-time retrieval strategies, fall short on effectively and accurately shortlisting relevant tools. This paper introduces a novel PLUTO (Planning, Learning, and Understanding for TOols) approach, encompassing {``}Plan-and-Retrieve (P{\&}R){''} and {``}Edit-and-Ground (E{\&}G){''} paradigms. The P{\&}R paradigm consists of a neural retrieval module for shortlisting relevant tools and an LLM-based query planner that decomposes complex queries into actionable tasks, enhancing the effectiveness of tool utilization. The E{\&}G paradigm utilizes LLMs to enrich tool descriptions based on user scenarios, bridging the gap between user queries and tool functionalities. Experiment results demonstrate that these paradigms significantly improve the recall and NDCG in tool retrieval tasks, significantly surpassing current state-of-the-art models. | [
"Huang, Tenghao",
"Jung, Dongwon",
"Kumar, Vaibhav",
"Kachuee, Mohammad",
"Li, Xiang",
"Xu, Puyang",
"Chen, Muhao"
] | Planning and Editing What You Retrieve for Enhanced Tool Learning | findings-naacl.61 | Poster | 2404.00450 | [
"https://github.com/tenghaohuang/pluto"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.62.bib | https://aclanthology.org/2024.findings-naacl.62/ | @inproceedings{carbune-etal-2024-chart,
title = "Chart-based Reasoning: Transferring Capabilities from {LLM}s to {VLM}s",
author = "Carbune, Victor and
Mansoor, Hassan and
Liu, Fangyu and
Aralikatte, Rahul and
Baechler, Gilles and
Chen, Jindong and
Sharma, Abhanshu",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.62",
doi = "10.18653/v1/2024.findings-naacl.62",
pages = "989--1004",
abstract = "Vision-language models (VLMs) are achieving increasingly strong performance on multimodal tasks. However, reasoning capabilities remain limited particularly for smaller VLMs, while those of large-language models (LLMs) have seen numerous improvements. We pro-pose a technique to transfer capabilities from LLMs to VLMs. On the recently introduced ChartQA, our method obtains state-of-the-artperformance when applied on the PaLI3-5B VLM by Chen et al. (2023c), while also enabling much better performance on PlotQA and FigureQA.We first improve the chart representation by continuing the pre-training stage using an improved version of the chart-to-table translation task by Liu et al. (2023a). We then propose constructing a 20x larger dataset than the original training set. To improve general reasoning capabilities and improve numerical operations, we synthesize reasoning traces using the table representation of charts. Lastly, our model is fine-tuned using the multitask loss introduced by Hsieh et al. (2023).Our variant ChartPaLI-5B outperforms even 10x larger models such as PaLIX-55B without using an upstream OCR system, while keeping inference time constant compared to the PaLI3-5B baseline. When rationales are further refined with a simple program-of-thought prompt (Chen et al., 2023a), our model outperforms the recently introduced Gemini Ultra and GPT-4V.",
}
| Vision-language models (VLMs) are achieving increasingly strong performance on multimodal tasks. However, reasoning capabilities remain limited particularly for smaller VLMs, while those of large-language models (LLMs) have seen numerous improvements. We pro-pose a technique to transfer capabilities from LLMs to VLMs. On the recently introduced ChartQA, our method obtains state-of-the-artperformance when applied on the PaLI3-5B VLM by Chen et al. (2023c), while also enabling much better performance on PlotQA and FigureQA.We first improve the chart representation by continuing the pre-training stage using an improved version of the chart-to-table translation task by Liu et al. (2023a). We then propose constructing a 20x larger dataset than the original training set. To improve general reasoning capabilities and improve numerical operations, we synthesize reasoning traces using the table representation of charts. Lastly, our model is fine-tuned using the multitask loss introduced by Hsieh et al. (2023).Our variant ChartPaLI-5B outperforms even 10x larger models such as PaLIX-55B without using an upstream OCR system, while keeping inference time constant compared to the PaLI3-5B baseline. When rationales are further refined with a simple program-of-thought prompt (Chen et al., 2023a), our model outperforms the recently introduced Gemini Ultra and GPT-4V. | [
"Carbune, Victor",
"Mansoor, Hassan",
"Liu, Fangyu",
"Aralikatte, Rahul",
"Baechler, Gilles",
"Chen, Jindong",
"Sharma, Abhanshu"
] | Chart-based Reasoning: Transferring Capabilities from LLMs to VLMs | findings-naacl.62 | Poster | 2403.12596 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.63.bib | https://aclanthology.org/2024.findings-naacl.63/ | @inproceedings{lin-etal-2024-slim,
title = "{SL}i{M}: Speculative Decoding with Hypothesis Reduction",
author = "Lin, Chi-Heng and
Tuli, Shikhar and
Smith, James and
Hsu, Yen-Chang and
Shen, Yilin and
Jin, Hongxia",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.63",
doi = "10.18653/v1/2024.findings-naacl.63",
pages = "1005--1017",
abstract = "Speculative decoding has emerged as a prominent alternative to autoregressive decoding for expediting inference in large language models (LLMs). However, prevailing assumptions often focus solely on latency reduction, neglecting the computational expenses. In this paper, we present \textbf{S}peculate \textbf{L}ess, val\textbf{i}date \textbf{M}ore (SLiM), a speculative decoding enhancement to reduce the speculation set while validating more effective tokens. SLiM is designed to mitigate LLMs{'} computation costs associated with the token verification by introducing hypothesis reduction based on a fast posterior estimation. It consistently surpasses counterparts lacking cost reduction across a spectrum from CPU to GPU. Our evaluation with diverse conversational datasets shows that SLiM can achieve a substantial 70{\%} reduction in FLOPs while generating more effective predictions on top of prior arts.",
}
| Speculative decoding has emerged as a prominent alternative to autoregressive decoding for expediting inference in large language models (LLMs). However, prevailing assumptions often focus solely on latency reduction, neglecting the computational expenses. In this paper, we present \textbf{S}peculate \textbf{L}ess, val\textbf{i}date \textbf{M}ore (SLiM), a speculative decoding enhancement to reduce the speculation set while validating more effective tokens. SLiM is designed to mitigate LLMs{'} computation costs associated with the token verification by introducing hypothesis reduction based on a fast posterior estimation. It consistently surpasses counterparts lacking cost reduction across a spectrum from CPU to GPU. Our evaluation with diverse conversational datasets shows that SLiM can achieve a substantial 70{\%} reduction in FLOPs while generating more effective predictions on top of prior arts. | [
"Lin, Chi-Heng",
"Tuli, Shikhar",
"Smith, James",
"Hsu, Yen-Chang",
"Shen, Yilin",
"Jin, Hongxia"
] | SLiM: Speculative Decoding with Hypothesis Reduction | findings-naacl.63 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.64.bib | https://aclanthology.org/2024.findings-naacl.64/ | @inproceedings{kachwala-etal-2024-rematch,
title = "{REMATCH}: Robust and Efficient Matching of Local Knowledge Graphs to Improve Structural and Semantic Similarity",
author = "Kachwala, Zoher and
An, Jisun and
Kwak, Haewoon and
Menczer, Filippo",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.64",
doi = "10.18653/v1/2024.findings-naacl.64",
pages = "1018--1028",
abstract = "Knowledge graphs play a pivotal role in various applications, such as question-answering and fact-checking. Abstract Meaning Representation (AMR) represents text as knowledge graphs. Evaluating the quality of these graphs involves matching them structurally to each other and semantically to the source text. Existing AMR metrics are inefficient and struggle to capture semantic similarity. We also lack a systematic evaluation benchmark for assessing structural similarity between AMR graphs. To overcome these limitations, we introduce a novel AMR similarity metric, rematch, alongside a new evaluation for structural similarity called RARE. Among state-of-the-art metrics, rematch ranks second in structural similarity; and first in semantic similarity by 1{--}5 percentage points on the STS-B and SICK-R benchmarks. Rematch is also five times faster than the next most efficient metric.",
}
| Knowledge graphs play a pivotal role in various applications, such as question-answering and fact-checking. Abstract Meaning Representation (AMR) represents text as knowledge graphs. Evaluating the quality of these graphs involves matching them structurally to each other and semantically to the source text. Existing AMR metrics are inefficient and struggle to capture semantic similarity. We also lack a systematic evaluation benchmark for assessing structural similarity between AMR graphs. To overcome these limitations, we introduce a novel AMR similarity metric, rematch, alongside a new evaluation for structural similarity called RARE. Among state-of-the-art metrics, rematch ranks second in structural similarity; and first in semantic similarity by 1{--}5 percentage points on the STS-B and SICK-R benchmarks. Rematch is also five times faster than the next most efficient metric. | [
"Kachwala, Zoher",
"An, Jisun",
"Kwak, Haewoon",
"Menczer, Filippo"
] | REMATCH: Robust and Efficient Matching of Local Knowledge Graphs to Improve Structural and Semantic Similarity | findings-naacl.64 | Poster | 2404.02126 | [
"https://github.com/osome-iu/Rematch-RARE"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.65.bib | https://aclanthology.org/2024.findings-naacl.65/ | @inproceedings{hutchinson-2024-modeling,
title = "Modeling the Sacred: Considerations when Using Religious Texts in Natural Language Processing",
author = "Hutchinson, Ben",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.65",
doi = "10.18653/v1/2024.findings-naacl.65",
pages = "1029--1043",
abstract = "This position paper concerns the use of religious texts in Natural Language Processing (NLP), which is of special interest to the Ethics of NLP. Religious texts are expressions of culturally important values, and machine learned models have a propensity to reproduce cultural values encoded in their training data. Furthermore, translations of religious texts are frequently used by NLP researchers when language data is scarce. This repurposes the translations from their original uses and motivations, which often involve attracting new followers. This paper argues that NLP{'}s use of such texts raises considerations that go beyond model biases, including data provenance, cultural contexts, and their use in proselytism. We argue for more consideration of researcher positionality, and of the perspectives of marginalized linguistic and religious communities.",
}
| This position paper concerns the use of religious texts in Natural Language Processing (NLP), which is of special interest to the Ethics of NLP. Religious texts are expressions of culturally important values, and machine learned models have a propensity to reproduce cultural values encoded in their training data. Furthermore, translations of religious texts are frequently used by NLP researchers when language data is scarce. This repurposes the translations from their original uses and motivations, which often involve attracting new followers. This paper argues that NLP{'}s use of such texts raises considerations that go beyond model biases, including data provenance, cultural contexts, and their use in proselytism. We argue for more consideration of researcher positionality, and of the perspectives of marginalized linguistic and religious communities. | [
"Hutchinson, Ben"
] | Modeling the Sacred: Considerations when Using Religious Texts in Natural Language Processing | findings-naacl.65 | Poster | 2404.14740 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.66.bib | https://aclanthology.org/2024.findings-naacl.66/ | @inproceedings{macke-doyle-2024-testing,
title = "Testing the Effect of Code Documentation on Large Language Model Code Understanding",
author = "Macke, William and
Doyle, Michael",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.66",
doi = "10.18653/v1/2024.findings-naacl.66",
pages = "1044--1050",
abstract = "Large Language Models (LLMs) have demonstrated impressive abilities in recent years with regards to code generation and understanding. However, little work has investigated how documentation and other code properties affect an LLM{'}s ability to understand and generate code or documentation. We present an empirical analysis of how underlying properties of code or documentation can affect an LLM{'}s capabilities. We show that providing an LLM with {``}incorrect{''} documentation can greatly hinder code understanding, while incomplete or missing documentation does not seem to significantly affect an LLM{'}s ability to understand code.",
}
| Large Language Models (LLMs) have demonstrated impressive abilities in recent years with regards to code generation and understanding. However, little work has investigated how documentation and other code properties affect an LLM{'}s ability to understand and generate code or documentation. We present an empirical analysis of how underlying properties of code or documentation can affect an LLM{'}s capabilities. We show that providing an LLM with {``}incorrect{''} documentation can greatly hinder code understanding, while incomplete or missing documentation does not seem to significantly affect an LLM{'}s ability to understand code. | [
"Macke, William",
"Doyle, Michael"
] | Testing the Effect of Code Documentation on Large Language Model Code Understanding | findings-naacl.66 | Poster | 2404.03114 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.67.bib | https://aclanthology.org/2024.findings-naacl.67/ | @inproceedings{cao-etal-2024-aligning,
title = "Aligning Large Language Models with Recommendation Knowledge",
author = "Cao, Yuwei and
Mehta, Nikhil and
Yi, Xinyang and
Hulikal Keshavan, Raghunandan and
Heldt, Lukasz and
Hong, Lichan and
Chi, Ed and
Sathiamoorthy, Maheswaran",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.67",
doi = "10.18653/v1/2024.findings-naacl.67",
pages = "1051--1066",
abstract = "Large language models (LLMs) have recently been used as backbones for recommender systems. However, their performance often lags behind conventional methods in standard tasks like retrieval. We attribute this to a mismatch between LLMs{'} knowledge and the knowledge crucial for effective recommendations. While LLMs excel at natural language reasoning, they cannot model complex user-item interactions inherent in recommendation tasks. We propose bridging the knowledge gap and equipping LLMs with recommendation-specific knowledge to address this. Operations such as Masked Item Modeling (MIM) and Bayesian Personalized Ranking (BPR) have found success in conventional recommender systems. Inspired by this, we simulate these operations through natural language to generate auxiliary-task data samples that encode item correlations and user preferences. Fine-tuning LLMs on such auxiliary-task data samples and incorporating more informative recommendation-task data samples facilitates the injection of recommendation-specific knowledge into LLMs. Extensive experiments across retrieval, ranking, and rating prediction tasks on LLMs such as FLAN-T5-Base and FLAN-T5-XL show the effectiveness of our technique in domains such as Amazon Toys {\&} Games, Beauty, and Sports {\&} Outdoors. Notably, our method outperforms conventional and LLM-based baselines, including the current SOTA, by significant margins in retrieval, showcasing its potential for enhancing recommendation quality.",
}
| Large language models (LLMs) have recently been used as backbones for recommender systems. However, their performance often lags behind conventional methods in standard tasks like retrieval. We attribute this to a mismatch between LLMs{'} knowledge and the knowledge crucial for effective recommendations. While LLMs excel at natural language reasoning, they cannot model complex user-item interactions inherent in recommendation tasks. We propose bridging the knowledge gap and equipping LLMs with recommendation-specific knowledge to address this. Operations such as Masked Item Modeling (MIM) and Bayesian Personalized Ranking (BPR) have found success in conventional recommender systems. Inspired by this, we simulate these operations through natural language to generate auxiliary-task data samples that encode item correlations and user preferences. Fine-tuning LLMs on such auxiliary-task data samples and incorporating more informative recommendation-task data samples facilitates the injection of recommendation-specific knowledge into LLMs. Extensive experiments across retrieval, ranking, and rating prediction tasks on LLMs such as FLAN-T5-Base and FLAN-T5-XL show the effectiveness of our technique in domains such as Amazon Toys {\&} Games, Beauty, and Sports {\&} Outdoors. Notably, our method outperforms conventional and LLM-based baselines, including the current SOTA, by significant margins in retrieval, showcasing its potential for enhancing recommendation quality. | [
"Cao, Yuwei",
"Mehta, Nikhil",
"Yi, Xinyang",
"Hulikal Keshavan, Raghun",
"an",
"Heldt, Lukasz",
"Hong, Lichan",
"Chi, Ed",
"Sathiamoorthy, Maheswaran"
] | Aligning Large Language Models with Recommendation Knowledge | findings-naacl.67 | Poster | 2404.00245 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.68.bib | https://aclanthology.org/2024.findings-naacl.68/ | @inproceedings{liu-etal-2024-ofa,
title = "{OFA}: A Framework of Initializing Unseen Subword Embeddings for Efficient Large-scale Multilingual Continued Pretraining",
author = "Liu, Yihong and
Lin, Peiqin and
Wang, Mingyang and
Schuetze, Hinrich",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.68",
doi = "10.18653/v1/2024.findings-naacl.68",
pages = "1067--1097",
abstract = "Instead of pretraining multilingual language models from scratch, a more efficient method is to adapt existing pretrained language models (PLMs) to new languages via vocabulary extension and continued pretraining. However, this method usually randomly initializes the embeddings of new subwords and introduces substantially more embedding parameters to the model, thus weakening the efficiency. To address these issues, we propose a novel framework: $\textbf{O}$ne $\textbf{F}$or $\textbf{A}$ll ($\textbf{OFA}$), which wisely initializes the embeddings of unseen subwords and thus can adapt a PLM to multiple languages efficiently and effectively. OFA takes advantage of external well-aligned multilingual static word vectors and injects the alignment knowledge into the subword embeddings. In addition, OFA applies matrix factorization and replaces the cumbersome embeddings with two lower-dimensional matrices, which largely reduces the number of parameters. We show OFA accelerates the convergence of continued pretraining, which is environmentally friendly as much fewer carbon footprints are generated. Through extensive experiments, we demonstrate OFA can achieve competitive or better performance than default continued pretraining baselines on a wide range of crosslingual downstream tasks. We make our code and models publicly available.",
}
| Instead of pretraining multilingual language models from scratch, a more efficient method is to adapt existing pretrained language models (PLMs) to new languages via vocabulary extension and continued pretraining. However, this method usually randomly initializes the embeddings of new subwords and introduces substantially more embedding parameters to the model, thus weakening the efficiency. To address these issues, we propose a novel framework: $\textbf{O}$ne $\textbf{F}$or $\textbf{A}$ll ($\textbf{OFA}$), which wisely initializes the embeddings of unseen subwords and thus can adapt a PLM to multiple languages efficiently and effectively. OFA takes advantage of external well-aligned multilingual static word vectors and injects the alignment knowledge into the subword embeddings. In addition, OFA applies matrix factorization and replaces the cumbersome embeddings with two lower-dimensional matrices, which largely reduces the number of parameters. We show OFA accelerates the convergence of continued pretraining, which is environmentally friendly as much fewer carbon footprints are generated. Through extensive experiments, we demonstrate OFA can achieve competitive or better performance than default continued pretraining baselines on a wide range of crosslingual downstream tasks. We make our code and models publicly available. | [
"Liu, Yihong",
"Lin, Peiqin",
"Wang, Mingyang",
"Schuetze, Hinrich"
] | OFA: A Framework of Initializing Unseen Subword Embeddings for Efficient Large-scale Multilingual Continued Pretraining | findings-naacl.68 | Poster | 2311.08849 | [
"https://github.com/cisnlp/ofa"
] | https://huggingface.co/papers/2311.08849 | 0 | 5 | 3 | 4 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.69.bib | https://aclanthology.org/2024.findings-naacl.69/ | @inproceedings{kim-etal-2024-self,
title = "{SELF}-{EXPERTISE}: Knowledge-based Instruction Dataset Augmentation for a Legal Expert Language Model",
author = "Kim, Minju and
Jung, Haein and
Koo, Myoung-Wan",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.69",
doi = "10.18653/v1/2024.findings-naacl.69",
pages = "1098--1112",
abstract = "The advent of instruction-tuned large language models (LLMs) has significantly advanced the field of automatic instruction dataset augmentation. However, the method of generating instructions and outputs from inherent knowledge of LLM can unintentionally produce hallucinations {---} instances of generating factually incorrect or misleading information. To overcome this, we propose SELF-EXPERTISE, automatically generating instruction dataset in the legal domain from a seed dataset. SELF-EXPERTISE extracts knowledge from the outputs of the seed dataset, and generates new instructions, inputs, and outputs. In this way, the proposed method reduces hallucination in automatic instruction augmentation. We trained an SELF-EXPERTISE augmented instruction dataset on the LLaMA-2 7B model to construct Korean legal specialized model, called LxPERT. LxPERT has demonstrated performance surpassing GPT-3.5-turbo in both in-domain and out-of-domain datasets. The SELF-EXPERTISE augmentation pipeline is not only applicable to the legal field but is also expected to be extendable to various domains, potentially advancing domain-specialized LLMs.",
}
| The advent of instruction-tuned large language models (LLMs) has significantly advanced the field of automatic instruction dataset augmentation. However, the method of generating instructions and outputs from inherent knowledge of LLM can unintentionally produce hallucinations {---} instances of generating factually incorrect or misleading information. To overcome this, we propose SELF-EXPERTISE, automatically generating instruction dataset in the legal domain from a seed dataset. SELF-EXPERTISE extracts knowledge from the outputs of the seed dataset, and generates new instructions, inputs, and outputs. In this way, the proposed method reduces hallucination in automatic instruction augmentation. We trained an SELF-EXPERTISE augmented instruction dataset on the LLaMA-2 7B model to construct Korean legal specialized model, called LxPERT. LxPERT has demonstrated performance surpassing GPT-3.5-turbo in both in-domain and out-of-domain datasets. The SELF-EXPERTISE augmentation pipeline is not only applicable to the legal field but is also expected to be extendable to various domains, potentially advancing domain-specialized LLMs. | [
"Kim, Minju",
"Jung, Haein",
"Koo, Myoung-Wan"
] | SELF-EXPERTISE: Knowledge-based Instruction Dataset Augmentation for a Legal Expert Language Model | findings-naacl.69 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.70.bib | https://aclanthology.org/2024.findings-naacl.70/ | @inproceedings{li-etal-2024-evaluating,
title = "Re-evaluating the Need for Visual Signals in Unsupervised Grammar Induction",
author = "Li, Boyi and
Corona, Rodolfo and
Mangalam, Karttikeya and
Chen, Catherine and
Flaherty, Daniel and
Belongie, Serge and
Weinberger, Kilian and
Malik, Jitendra and
Darrell, Trevor and
Klein, Dan",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.70",
doi = "10.18653/v1/2024.findings-naacl.70",
pages = "1113--1123",
abstract = "Are multimodal inputs necessary for grammar induction? Recent work has shown that multimodal training inputs can improve grammar induction. However, these improvements are based on comparisons to weak text-only baselines that were trained on relatively little textual data. To determine whether multimodal inputs are needed in regimes with large amounts of textual training data, we design a stronger text-only baseline, which we refer to as LC-PCFG. LC-PCFG is a C-PFCG that incorporates embeddings from text-only large language models (LLMs). We use a fixed grammar family to directly compare LC-PCFG to various multimodal grammar induction methods. We compare performance on four benchmark datasets. LC-PCFG provides an up to 17{\%} relative improvement in Corpus-F1 compared to state-of-the-art multimodal grammar induction methods. LC-PCFG is also more computationally efficient, providing an up to 85{\%} reduction in parameter count and $8.8\times$ reduction in training time compared to multimodal approaches. These results suggest that multimodal inputs may not be necessary for grammar induction, and emphasize the importance of strong vision-free baselines for evaluating the benefit of multimodal approaches.",
}
| Are multimodal inputs necessary for grammar induction? Recent work has shown that multimodal training inputs can improve grammar induction. However, these improvements are based on comparisons to weak text-only baselines that were trained on relatively little textual data. To determine whether multimodal inputs are needed in regimes with large amounts of textual training data, we design a stronger text-only baseline, which we refer to as LC-PCFG. LC-PCFG is a C-PFCG that incorporates embeddings from text-only large language models (LLMs). We use a fixed grammar family to directly compare LC-PCFG to various multimodal grammar induction methods. We compare performance on four benchmark datasets. LC-PCFG provides an up to 17{\%} relative improvement in Corpus-F1 compared to state-of-the-art multimodal grammar induction methods. LC-PCFG is also more computationally efficient, providing an up to 85{\%} reduction in parameter count and $8.8\times$ reduction in training time compared to multimodal approaches. These results suggest that multimodal inputs may not be necessary for grammar induction, and emphasize the importance of strong vision-free baselines for evaluating the benefit of multimodal approaches. | [
"Li, Boyi",
"Corona, Rodolfo",
"Mangalam, Karttikeya",
"Chen, Catherine",
"Flaherty, Daniel",
"Belongie, Serge",
"Weinberger, Kilian",
"Malik, Jitendra",
"Darrell, Trevor",
"Klein, Dan"
] | Re-evaluating the Need for Visual Signals in Unsupervised Grammar Induction | findings-naacl.70 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.71.bib | https://aclanthology.org/2024.findings-naacl.71/ | @inproceedings{zhu-etal-2024-edentail,
title = "{EDE}ntail: An Entailment-based Few-shot Text Classification with Extensional Definition",
author = "Zhu, Zixiao and
Qian, Junlang and
Feng, Zijian and
Zhou, Hanzhang and
Mao, Kezhi",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.71",
doi = "10.18653/v1/2024.findings-naacl.71",
pages = "1124--1137",
abstract = "Few-shot text classification has seen significant advancements, particularly with entailment-based methods, which typically use either class labels or intensional definitions of class labels in hypotheses for label semantics expression. In this paper, we propose EDEntail, a method that employs extensional definition (EDef) of class labels in hypotheses, aiming to express the semantics of class labels more explicitly. To achieve the above goal, we develop an algorithm to gather and select extensional descriptive words of class labels and then order and format them into a sequence to form hypotheses. Our method has been evaluated and compared with state-of-the-art models on five classification datasets. The results demonstrate that our approach surpasses the supervised-learning methods and prompt-based methods under the few-shot setting, which underlines the potential of using an extensional definition of class labels for entailment-based few-shot text classification. Our code is available at https://github.com/MidiyaZhu/EDEntail.",
}
| Few-shot text classification has seen significant advancements, particularly with entailment-based methods, which typically use either class labels or intensional definitions of class labels in hypotheses for label semantics expression. In this paper, we propose EDEntail, a method that employs extensional definition (EDef) of class labels in hypotheses, aiming to express the semantics of class labels more explicitly. To achieve the above goal, we develop an algorithm to gather and select extensional descriptive words of class labels and then order and format them into a sequence to form hypotheses. Our method has been evaluated and compared with state-of-the-art models on five classification datasets. The results demonstrate that our approach surpasses the supervised-learning methods and prompt-based methods under the few-shot setting, which underlines the potential of using an extensional definition of class labels for entailment-based few-shot text classification. Our code is available at https://github.com/MidiyaZhu/EDEntail. | [
"Zhu, Zixiao",
"Qian, Junlang",
"Feng, Zijian",
"Zhou, Hanzhang",
"Mao, Kezhi"
] | EDEntail: An Entailment-based Few-shot Text Classification with Extensional Definition | findings-naacl.71 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.72.bib | https://aclanthology.org/2024.findings-naacl.72/ | @inproceedings{srivatsa-kochmar-2024-makes,
title = "What Makes Math Word Problems Challenging for {LLM}s?",
author = "Srivatsa, Kv Aditya and
Kochmar, Ekaterina",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.72",
doi = "10.18653/v1/2024.findings-naacl.72",
pages = "1138--1148",
abstract = "This paper investigates the question of what makes math word problems (MWPs) in English challenging for large language models (LLMs). We conduct an in-depth analysis of the key linguistic and mathematical characteristics of MWPs. In addition, we train feature-based classifiers to better understand the impact of each feature on the overall difficulty of MWPs for prominent LLMs and investigate whether this helps predict how well LLMs fare against specific categories of MWPs.",
}
| This paper investigates the question of what makes math word problems (MWPs) in English challenging for large language models (LLMs). We conduct an in-depth analysis of the key linguistic and mathematical characteristics of MWPs. In addition, we train feature-based classifiers to better understand the impact of each feature on the overall difficulty of MWPs for prominent LLMs and investigate whether this helps predict how well LLMs fare against specific categories of MWPs. | [
"Srivatsa, Kv Aditya",
"Kochmar, Ekaterina"
] | What Makes Math Word Problems Challenging for LLMs? | findings-naacl.72 | Poster | 2403.11369 | [
"https://github.com/kvadityasrivatsa/analyzing-llms-for-mwps"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.73.bib | https://aclanthology.org/2024.findings-naacl.73/ | @inproceedings{hyun-etal-2024-smile,
title = "{SMILE}: Multimodal Dataset for Understanding Laughter in Video with Language Models",
author = "Hyun, Lee and
Sung-Bin, Kim and
Han, Seungju and
Yu, Youngjae and
Oh, Tae-Hyun",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.73",
doi = "10.18653/v1/2024.findings-naacl.73",
pages = "1149--1167",
abstract = "Despite the recent advances in artificial intelligence, building social intelligence remains a challenge.Among social signals, laughter is one of the distinctive expressions that occurs during social interactions between humans.In this work, we tackle a new challenge for machines to understand the rationale behind laughter in video, Video Laugh Reasoning.We introduce this new task to explain why people laugh in a particular video and a dataset for this task.Our proposed dataset, SMILE, comprises video clips and language descriptions of why people laugh. We propose a baseline by leveraging the reasoning capacity of large language models (LLMs) with textual video representation. Experiments show that our baseline can generate plausible explanations for laughter. We further investigate the scalability of our baseline by probing other video understanding tasks and in-the-wild videos. We release our dataset, code, and model checkpoints on https://github.com/postech-ami/SMILE-Dataset.",
}
| Despite the recent advances in artificial intelligence, building social intelligence remains a challenge.Among social signals, laughter is one of the distinctive expressions that occurs during social interactions between humans.In this work, we tackle a new challenge for machines to understand the rationale behind laughter in video, Video Laugh Reasoning.We introduce this new task to explain why people laugh in a particular video and a dataset for this task.Our proposed dataset, SMILE, comprises video clips and language descriptions of why people laugh. We propose a baseline by leveraging the reasoning capacity of large language models (LLMs) with textual video representation. Experiments show that our baseline can generate plausible explanations for laughter. We further investigate the scalability of our baseline by probing other video understanding tasks and in-the-wild videos. We release our dataset, code, and model checkpoints on https://github.com/postech-ami/SMILE-Dataset. | [
"Hyun, Lee",
"Sung-Bin, Kim",
"Han, Seungju",
"Yu, Youngjae",
"Oh, Tae-Hyun"
] | SMILE: Multimodal Dataset for Understanding Laughter in Video with Language Models | findings-naacl.73 | Poster | 2312.09818 | [
"https://github.com/smile-data/smile"
] | https://huggingface.co/papers/2312.09818 | 0 | 0 | 0 | 5 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.74.bib | https://aclanthology.org/2024.findings-naacl.74/ | @inproceedings{peng-etal-2024-t3m,
title = "{T}3{M}: Text Guided 3{D} Human Motion Synthesis from Speech",
author = "Peng, Wenshuo and
Zhang, Kaipeng and
Zhang, Sai Qian",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.74",
doi = "10.18653/v1/2024.findings-naacl.74",
pages = "1168--1177",
abstract = "Speech-driven 3D motion synthesis seeks to create lifelike animations based on human speech, with potential uses in virtual reality, gaming, and the film production. Existing approaches reply solely on speech audio for motion generation, leading to inaccurate and inflexible synthesis results. To mitigate this problem, we introduce a novel text-guided 3D human motion synthesis method, termed T3M. Unlike traditional approaches, T3M allows precise control over motion synthesis via textual input, enhancing the degree of diversity and user customization. The experiment results demonstrate that T3M can greatly outperform the state-of-the-art methods in both quantitative metrics and qualitative evaluations. We have publicly released our code at https://github.com/Gloria2tt/naacl2024.git",
}
| Speech-driven 3D motion synthesis seeks to create lifelike animations based on human speech, with potential uses in virtual reality, gaming, and the film production. Existing approaches reply solely on speech audio for motion generation, leading to inaccurate and inflexible synthesis results. To mitigate this problem, we introduce a novel text-guided 3D human motion synthesis method, termed T3M. Unlike traditional approaches, T3M allows precise control over motion synthesis via textual input, enhancing the degree of diversity and user customization. The experiment results demonstrate that T3M can greatly outperform the state-of-the-art methods in both quantitative metrics and qualitative evaluations. We have publicly released our code at https://github.com/Gloria2tt/naacl2024.git | [
"Peng, Wenshuo",
"Zhang, Kaipeng",
"Zhang, Sai Qian"
] | T3M: Text Guided 3D Human Motion Synthesis from Speech | findings-naacl.74 | Poster | 2408.12885 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.75.bib | https://aclanthology.org/2024.findings-naacl.75/ | @inproceedings{peng-etal-2024-deja,
title = "Deja vu: Contrastive Historical Modeling with Prefix-tuning for Temporal Knowledge Graph Reasoning",
author = "Peng, Miao and
Liu, Ben and
Xu, Wenjie and
Jiang, Zihao and
Zhu, Jiahui and
Peng, Min",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.75",
doi = "10.18653/v1/2024.findings-naacl.75",
pages = "1178--1191",
abstract = "Temporal Knowledge Graph Reasoning (TKGR) is the task of inferring missing facts for incomplete TKGs in complex scenarios (e.g., transductive and inductive settings), which has been gaining increasing attention. Recently, to mitigate dependence on structured connections in TKGs, text-based methods have been developed to utilize rich linguistic information from entity descriptions. However, suffering from the enormous parameters and inflexibility of pre-trained language models, existing text-based methods struggle to balance the textual knowledge and temporal information with computationally expensive purpose-built training strategies. To tap the potential of text-based models for TKGR in various complex scenarios, we propose ChapTER, a Contrastive historical modeling framework with prefix-tuning for TEmporal Reasoning. ChapTER feeds history-contextualized text into the pseudo-Siamese encoders to strike a textual-temporal balance via contrastive estimation between queries and candidates. By introducing virtual time prefix tokens, it applies a prefix-based tuning method to facilitate the frozen PLM capable for TKGR tasks under different settings. We evaluate ChapTER on four transductive and three few-shot inductive TKGR benchmarks, and experimental results demonstrate that ChapTER achieves superior performance compared to competitive baselines with only 0.17{\%} tuned parameters. We conduct thorough analysis to verify the effectiveness, flexibility and efficiency of ChapTER.",
}
| Temporal Knowledge Graph Reasoning (TKGR) is the task of inferring missing facts for incomplete TKGs in complex scenarios (e.g., transductive and inductive settings), which has been gaining increasing attention. Recently, to mitigate dependence on structured connections in TKGs, text-based methods have been developed to utilize rich linguistic information from entity descriptions. However, suffering from the enormous parameters and inflexibility of pre-trained language models, existing text-based methods struggle to balance the textual knowledge and temporal information with computationally expensive purpose-built training strategies. To tap the potential of text-based models for TKGR in various complex scenarios, we propose ChapTER, a Contrastive historical modeling framework with prefix-tuning for TEmporal Reasoning. ChapTER feeds history-contextualized text into the pseudo-Siamese encoders to strike a textual-temporal balance via contrastive estimation between queries and candidates. By introducing virtual time prefix tokens, it applies a prefix-based tuning method to facilitate the frozen PLM capable for TKGR tasks under different settings. We evaluate ChapTER on four transductive and three few-shot inductive TKGR benchmarks, and experimental results demonstrate that ChapTER achieves superior performance compared to competitive baselines with only 0.17{\%} tuned parameters. We conduct thorough analysis to verify the effectiveness, flexibility and efficiency of ChapTER. | [
"Peng, Miao",
"Liu, Ben",
"Xu, Wenjie",
"Jiang, Zihao",
"Zhu, Jiahui",
"Peng, Min"
] | Deja vu: Contrastive Historical Modeling with Prefix-tuning for Temporal Knowledge Graph Reasoning | findings-naacl.75 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.76.bib | https://aclanthology.org/2024.findings-naacl.76/ | @inproceedings{prasad-etal-2024-explanation,
title = "Explanation Extraction from Hierarchical Classification Frameworks for Long Legal Documents",
author = "Prasad, Nishchal and
Dkaki, Taoufiq and
Boughanem, Mohand",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.76",
doi = "10.18653/v1/2024.findings-naacl.76",
pages = "1192--1201",
abstract = "Hierarchical classification frameworks have been widely used to process long sequences, especially in the legal domain for predictions from long legal documents. But being black-box models they are unable to explain their predictions making them less reliable for practical applications, more so in the legal domain. In this work, we develop an extractive explanation algorithm for hierarchical frameworks for long sequences based on the sensitivity of the trained model to its input perturbations. We perturb using occlusion and develop Ob-HEx; an Occlusion-based Hierarchical Explanation-extractor. We adapt Ob-HEx to Hierarchical Transformer models trained on long Indian legal texts. And use Ob-HEx to analyze them and extract their explanations for the ILDC-Expert dataset, achieving a minimum gain of 1 point over the previous benchmark on most of our performance evaluation metrics.",
}
| Hierarchical classification frameworks have been widely used to process long sequences, especially in the legal domain for predictions from long legal documents. But being black-box models they are unable to explain their predictions making them less reliable for practical applications, more so in the legal domain. In this work, we develop an extractive explanation algorithm for hierarchical frameworks for long sequences based on the sensitivity of the trained model to its input perturbations. We perturb using occlusion and develop Ob-HEx; an Occlusion-based Hierarchical Explanation-extractor. We adapt Ob-HEx to Hierarchical Transformer models trained on long Indian legal texts. And use Ob-HEx to analyze them and extract their explanations for the ILDC-Expert dataset, achieving a minimum gain of 1 point over the previous benchmark on most of our performance evaluation metrics. | [
"Prasad, Nishchal",
"Dkaki, Taoufiq",
"Boughanem, Moh",
""
] | Explanation Extraction from Hierarchical Classification Frameworks for Long Legal Documents | findings-naacl.76 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.77.bib | https://aclanthology.org/2024.findings-naacl.77/ | @inproceedings{whitehouse-etal-2024-low,
title = "Low-Rank Adaptation for Multilingual Summarization: An Empirical Study",
author = "Whitehouse, Chenxi and
Huot, Fantine and
Bastings, Jasmijn and
Dehghani, Mostafa and
Lin, Chu-Cheng and
Lapata, Mirella",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.77",
doi = "10.18653/v1/2024.findings-naacl.77",
pages = "1202--1228",
abstract = "Although the advancements of pre-trained Large Language Models have significantly accelerated recent progress in NLP, their ever-increasing size poses significant challenges for conventional fine-tuning, especially in memory-intensive tasks. We investigate the potential of Parameter-Efficient Fine-Tuning, focusing on Low-Rank Adaptation (LoRA), in the domain of multilingual summarization, a task that is both challenging (due to typically long inputs), and relatively unexplored. We conduct an extensive study across different data availability scenarios, including high- and low-data settings, and cross-lingual transfer, leveraging models of different sizes. Our findings reveal that LoRA is competitive with full fine-tuning when trained with high quantities of data, and excels in low-data scenarios and cross-lingual transfer. We also study different strategies for few-shot cross-lingual transfer, finding that continued LoRA tuning outperforms full fine-tuning and the dynamic composition of language-specific LoRA modules.",
}
| Although the advancements of pre-trained Large Language Models have significantly accelerated recent progress in NLP, their ever-increasing size poses significant challenges for conventional fine-tuning, especially in memory-intensive tasks. We investigate the potential of Parameter-Efficient Fine-Tuning, focusing on Low-Rank Adaptation (LoRA), in the domain of multilingual summarization, a task that is both challenging (due to typically long inputs), and relatively unexplored. We conduct an extensive study across different data availability scenarios, including high- and low-data settings, and cross-lingual transfer, leveraging models of different sizes. Our findings reveal that LoRA is competitive with full fine-tuning when trained with high quantities of data, and excels in low-data scenarios and cross-lingual transfer. We also study different strategies for few-shot cross-lingual transfer, finding that continued LoRA tuning outperforms full fine-tuning and the dynamic composition of language-specific LoRA modules. | [
"Whitehouse, Chenxi",
"Huot, Fantine",
"Bastings, Jasmijn",
"Dehghani, Mostafa",
"Lin, Chu-Cheng",
"Lapata, Mirella"
] | Low-Rank Adaptation for Multilingual Summarization: An Empirical Study | findings-naacl.77 | Poster | 2311.08572 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.78.bib | https://aclanthology.org/2024.findings-naacl.78/ | @inproceedings{ranaldi-etal-2024-tree,
title = "A Tree-of-Thoughts to Broaden Multi-step Reasoning across Languages",
author = "Ranaldi, Leonardo and
Pucci, Giulia and
Ranaldi, Federico and
Ruzzetti, Elena Sofia and
Zanzotto, Fabio Massimo",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.78",
doi = "10.18653/v1/2024.findings-naacl.78",
pages = "1229--1241",
abstract = "Reasoning methods, best exemplified by the well-known Chain-of-Thought (CoT), empower the reasoning abilities of Large Language Models (LLMs) by eliciting them to solve complex tasks in a step-by-step manner. Although they are achieving significant success, the ability to deliver multi-step reasoning remains limited to English because of the imbalance in the distribution of pre-training data, which makes other languages a barrier. In this paper, we propose Cross-lingual Tree-of-Thoughts (Cross-ToT), a method for aligning Cross-lingual CoT reasoning across languages. The proposed method, through a self-consistent cross-lingual prompting mechanism inspired by the Tree-of-Thoughts approach, provides multi-step reasoning paths in different languages that, during the steps, lead to the final solution. Experimental evaluations show that our method significantly outperforms existing prompting methods by reducing the number of interactions and achieving state-of-the-art performance.",
}
| Reasoning methods, best exemplified by the well-known Chain-of-Thought (CoT), empower the reasoning abilities of Large Language Models (LLMs) by eliciting them to solve complex tasks in a step-by-step manner. Although they are achieving significant success, the ability to deliver multi-step reasoning remains limited to English because of the imbalance in the distribution of pre-training data, which makes other languages a barrier. In this paper, we propose Cross-lingual Tree-of-Thoughts (Cross-ToT), a method for aligning Cross-lingual CoT reasoning across languages. The proposed method, through a self-consistent cross-lingual prompting mechanism inspired by the Tree-of-Thoughts approach, provides multi-step reasoning paths in different languages that, during the steps, lead to the final solution. Experimental evaluations show that our method significantly outperforms existing prompting methods by reducing the number of interactions and achieving state-of-the-art performance. | [
"Ranaldi, Leonardo",
"Pucci, Giulia",
"Ranaldi, Federico",
"Ruzzetti, Elena Sofia",
"Zanzotto, Fabio Massimo"
] | A Tree-of-Thoughts to Broaden Multi-step Reasoning across Languages | findings-naacl.78 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.79.bib | https://aclanthology.org/2024.findings-naacl.79/ | @inproceedings{muckatira-etal-2024-emergent,
title = "Emergent Abilities in Reduced-Scale Generative Language Models",
author = "Muckatira, Sherin and
Deshpande, Vijeta and
Lialin, Vladislav and
Rumshisky, Anna",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.79",
doi = "10.18653/v1/2024.findings-naacl.79",
pages = "1242--1257",
abstract = "Large language models can solve new tasks without task-specific fine-tuning. This ability, also known as in-context learning (ICL), is considered an emergent ability and is primarily seen in large language models with billions of parameters. This study investigates if such emergent properties are strictly tied to model size or can be demonstrated by smaller models trained on reduced-scale data. To explore this, we simplify pre-training data and pre-train 36 causal language models with parameters varying from 1 million to 165 million parameters. We show that models trained on this simplified pre-training data demonstrate enhanced zero-shot capabilities across various tasks in simplified language, achieving performance comparable to that of pre-trained models six times larger on unrestricted language. This suggests that downscaling the language allows zero-shot learning capabilities to emerge in models with limited size.Additionally, we find that these smaller models pre-trained on simplified data demonstrate a power law relationship between the evaluation loss and the three scaling factors: compute, dataset size, and model size.",
}
| Large language models can solve new tasks without task-specific fine-tuning. This ability, also known as in-context learning (ICL), is considered an emergent ability and is primarily seen in large language models with billions of parameters. This study investigates if such emergent properties are strictly tied to model size or can be demonstrated by smaller models trained on reduced-scale data. To explore this, we simplify pre-training data and pre-train 36 causal language models with parameters varying from 1 million to 165 million parameters. We show that models trained on this simplified pre-training data demonstrate enhanced zero-shot capabilities across various tasks in simplified language, achieving performance comparable to that of pre-trained models six times larger on unrestricted language. This suggests that downscaling the language allows zero-shot learning capabilities to emerge in models with limited size.Additionally, we find that these smaller models pre-trained on simplified data demonstrate a power law relationship between the evaluation loss and the three scaling factors: compute, dataset size, and model size. | [
"Muckatira, Sherin",
"Deshp",
"e, Vijeta",
"Lialin, Vladislav",
"Rumshisky, Anna"
] | Emergent Abilities in Reduced-Scale Generative Language Models | findings-naacl.79 | Poster | 2404.02204 | [
"https://github.com/text-machine-lab/mini_gpt"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.80.bib | https://aclanthology.org/2024.findings-naacl.80/ | @inproceedings{siro-etal-2024-context,
title = "Context Does Matter: Implications for Crowdsourced Evaluation Labels in Task-Oriented Dialogue Systems",
author = "Siro, Clemencia and
Aliannejadi, Mohammad and
Rijke, Maarten",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.80",
doi = "10.18653/v1/2024.findings-naacl.80",
pages = "1258--1273",
abstract = "Crowdsourced labels play a crucial role in evaluating task-oriented dialogue systems (TDSs). Obtaining high-quality and consistent ground-truth labels from annotators presents challenges. When evaluating a TDS, annotators must fully comprehend the dialogue before providing judgments. Previous studies suggest using only a portion of the dialogue context in the annotation process. However, the impact of this limitation on label quality remains unexplored. This study investigates the influence of dialogue context on annotation quality, considering the truncated context for relevance and usefulness labeling. We further propose to use large language models ( LLMs) to summarize the dialogue context to provide a rich and short description of the dialogue context and study the impact of doing so on the annotator{'}s performance. Reducing context leads to more positive ratings. Conversely, providing the entire dialogue context yields higher-quality relevance ratings but introduces ambiguity in usefulness ratings. Using the first user utterance as context leads to consistent ratings, akin to those obtained using the entire dialogue, with significantly reduced annotation effort. Our findings show how task design, particularly the availability of dialogue context, affects the quality and consistency of crowdsourced evaluation labels.",
}
| Crowdsourced labels play a crucial role in evaluating task-oriented dialogue systems (TDSs). Obtaining high-quality and consistent ground-truth labels from annotators presents challenges. When evaluating a TDS, annotators must fully comprehend the dialogue before providing judgments. Previous studies suggest using only a portion of the dialogue context in the annotation process. However, the impact of this limitation on label quality remains unexplored. This study investigates the influence of dialogue context on annotation quality, considering the truncated context for relevance and usefulness labeling. We further propose to use large language models ( LLMs) to summarize the dialogue context to provide a rich and short description of the dialogue context and study the impact of doing so on the annotator{'}s performance. Reducing context leads to more positive ratings. Conversely, providing the entire dialogue context yields higher-quality relevance ratings but introduces ambiguity in usefulness ratings. Using the first user utterance as context leads to consistent ratings, akin to those obtained using the entire dialogue, with significantly reduced annotation effort. Our findings show how task design, particularly the availability of dialogue context, affects the quality and consistency of crowdsourced evaluation labels. | [
"Siro, Clemencia",
"Aliannejadi, Mohammad",
"Rijke, Maarten"
] | Context Does Matter: Implications for Crowdsourced Evaluation Labels in Task-Oriented Dialogue Systems | findings-naacl.80 | Poster | 2404.09980 | [
"https://github.com/clemenciah/effects-of-dialogue-context"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.81.bib | https://aclanthology.org/2024.findings-naacl.81/ | @inproceedings{zhou-etal-2024-matching,
title = "Matching Varying-Length Texts via Topic-Informed and Decoupled Sentence Embeddings",
author = "Zhou, Xixi and
Gu, Chunbin and
Jie, Xin and
Bu, Jiajun and
Wang, Haishuai",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.81",
doi = "10.18653/v1/2024.findings-naacl.81",
pages = "1274--1280",
abstract = "Measuring semantic similarity between texts is a crucial task in natural language processing. While existing semantic text matching focuses on pairs of similar-length sequences, matching texts with non-comparable lengths has broader applications in specific domains, such as comparing professional document summaries and content. Current approaches struggle with text pairs of non-comparable lengths due to truncation issues. To address this, we split texts into natural sentences and decouple sentence representations using supervised contrastive learning (SCL). Meanwhile, we adopt the embedded topic model (ETM) for specific domain data. Our experiments demonstrate the effectiveness of our model, based on decoupled and topic-informed sentence embeddings, in matching texts of significantly different lengths across three well-studied datasets.",
}
| Measuring semantic similarity between texts is a crucial task in natural language processing. While existing semantic text matching focuses on pairs of similar-length sequences, matching texts with non-comparable lengths has broader applications in specific domains, such as comparing professional document summaries and content. Current approaches struggle with text pairs of non-comparable lengths due to truncation issues. To address this, we split texts into natural sentences and decouple sentence representations using supervised contrastive learning (SCL). Meanwhile, we adopt the embedded topic model (ETM) for specific domain data. Our experiments demonstrate the effectiveness of our model, based on decoupled and topic-informed sentence embeddings, in matching texts of significantly different lengths across three well-studied datasets. | [
"Zhou, Xixi",
"Gu, Chunbin",
"Jie, Xin",
"Bu, Jiajun",
"Wang, Haishuai"
] | Matching Varying-Length Texts via Topic-Informed and Decoupled Sentence Embeddings | findings-naacl.81 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.82.bib | https://aclanthology.org/2024.findings-naacl.82/ | @inproceedings{lee-etal-2024-instruction,
title = "Instruction Tuning with Human Curriculum",
author = "Lee, Bruce W and
Cho, Hyunsoo and
Yoo, Kang Min",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.82",
doi = "10.18653/v1/2024.findings-naacl.82",
pages = "1281--1309",
abstract = "In this work, we (1) introduce Curriculum Instruction Tuning, (2) explore the potential advantages of employing diverse curriculum strategies, and (3) delineate a synthetic instruction-response generation framework that complements our theoretical approach. Distinct from the existing instruction tuning dataset, our generation pipeline is systematically structured to emulate the sequential and orderly characteristic of human learning. Additionally, we describe a methodology for generating instruction-response datasets that extensively span the various stages of human education, from middle school through the graduate level, utilizing educational subject catalogs.Before training, we meticulously organize the instruction data to ensure that questions escalate in difficulty regarding (A) the subject matter and (B) the intricacy of the instructions. The findings of our study reveal that substantial improvements in performance can be achieved through the mere application of curriculum ordering to instruction data{---}achieving gains of +4.76 on TruthfulQA, +2.98 on MMLU, +2.8 on OpenbookQA, and +1.28 on ARC-hard{---}compared to random shuffling. This enhancement is achieved without incurring additional computational expenses. Through comprehensive experimentation, we observe that the advantages of our proposed method are consistently evident across nine benchmarks.",
}
| In this work, we (1) introduce Curriculum Instruction Tuning, (2) explore the potential advantages of employing diverse curriculum strategies, and (3) delineate a synthetic instruction-response generation framework that complements our theoretical approach. Distinct from the existing instruction tuning dataset, our generation pipeline is systematically structured to emulate the sequential and orderly characteristic of human learning. Additionally, we describe a methodology for generating instruction-response datasets that extensively span the various stages of human education, from middle school through the graduate level, utilizing educational subject catalogs.Before training, we meticulously organize the instruction data to ensure that questions escalate in difficulty regarding (A) the subject matter and (B) the intricacy of the instructions. The findings of our study reveal that substantial improvements in performance can be achieved through the mere application of curriculum ordering to instruction data{---}achieving gains of +4.76 on TruthfulQA, +2.98 on MMLU, +2.8 on OpenbookQA, and +1.28 on ARC-hard{---}compared to random shuffling. This enhancement is achieved without incurring additional computational expenses. Through comprehensive experimentation, we observe that the advantages of our proposed method are consistently evident across nine benchmarks. | [
"Lee, Bruce W",
"Cho, Hyunsoo",
"Yoo, Kang Min"
] | Instruction Tuning with Human Curriculum | findings-naacl.82 | Poster | 2310.09518 | [
"https://github.com/imoneoi/openchat"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.83.bib | https://aclanthology.org/2024.findings-naacl.83/ | @inproceedings{rahman-xue-2024-natural,
title = "Natural Language-based State Representation in Deep Reinforcement Learning",
author = "Rahman, Md Masudur and
Xue, Yexiang",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.83",
doi = "10.18653/v1/2024.findings-naacl.83",
pages = "1310--1319",
abstract = "This paper investigates the potential of using natural language descriptions as an alternative to direct image-based observations for learning policies in reinforcement learning. Due to the inherent challenges in managing image-based observations, which include abundant information and irrelevant features, we propose a method that compresses images into a natural language form for state representation. This approach allows better interpretability and leverages the processing capabilities of large-language models. We conducted several experiments involving tasks that required image-based observation. The results demonstrated that policies trained using natural language descriptions of images yield better generalization than those trained directly from images, emphasizing the potential of this approach in practical settings.",
}
| This paper investigates the potential of using natural language descriptions as an alternative to direct image-based observations for learning policies in reinforcement learning. Due to the inherent challenges in managing image-based observations, which include abundant information and irrelevant features, we propose a method that compresses images into a natural language form for state representation. This approach allows better interpretability and leverages the processing capabilities of large-language models. We conducted several experiments involving tasks that required image-based observation. The results demonstrated that policies trained using natural language descriptions of images yield better generalization than those trained directly from images, emphasizing the potential of this approach in practical settings. | [
"Rahman, Md Masudur",
"Xue, Yexiang"
] | Natural Language-based State Representation in Deep Reinforcement Learning | findings-naacl.83 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.84.bib | https://aclanthology.org/2024.findings-naacl.84/ | @inproceedings{wang-etal-2024-learning-cross,
title = "Learning Cross-Architecture Instruction Embeddings for Binary Code Analysis in Low-Resource Architectures",
author = "Wang, Junzhe and
Zeng, Qiang and
Luo, Lannan",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.84",
doi = "10.18653/v1/2024.findings-naacl.84",
pages = "1320--1332",
abstract = "Binary code analysis is indispensable for a variety of software security tasks. Applying deep learning to binary code analysis has drawn great attention because of its notable performance. Today, source code is frequently compiled for various Instruction Set Architectures (ISAs). It is thus critical to expand binary analysis capabilities to multiple ISAs. Given a binary analysis task, the scale of available data on different ISAs varies. As a result, the rich datasets (e.g., malware) for certain ISAs, such as x86, lead to a disproportionate focus on these ISAs and a negligence of other ISAs, such as PowerPC, which suffer from the {``}data scarcity{''} problem. To address the problem, we propose to learn cross-architecture instruction embeddings (CAIE), where semantically-similar instructions, regardless of their ISAs, have close embeddings in a shared space. Consequently, we can transfer a model trained on a data-rich ISA to another ISA with less available data. We consider four ISAs (x86, ARM, MIPS, and PowerPC) and conduct both intrinsic and extrinsic evaluations (including malware detection and function similarity comparison). The results demonstrate the effectiveness of our approach to generate high-quality CAIE with good transferability.",
}
| Binary code analysis is indispensable for a variety of software security tasks. Applying deep learning to binary code analysis has drawn great attention because of its notable performance. Today, source code is frequently compiled for various Instruction Set Architectures (ISAs). It is thus critical to expand binary analysis capabilities to multiple ISAs. Given a binary analysis task, the scale of available data on different ISAs varies. As a result, the rich datasets (e.g., malware) for certain ISAs, such as x86, lead to a disproportionate focus on these ISAs and a negligence of other ISAs, such as PowerPC, which suffer from the {``}data scarcity{''} problem. To address the problem, we propose to learn cross-architecture instruction embeddings (CAIE), where semantically-similar instructions, regardless of their ISAs, have close embeddings in a shared space. Consequently, we can transfer a model trained on a data-rich ISA to another ISA with less available data. We consider four ISAs (x86, ARM, MIPS, and PowerPC) and conduct both intrinsic and extrinsic evaluations (including malware detection and function similarity comparison). The results demonstrate the effectiveness of our approach to generate high-quality CAIE with good transferability. | [
"Wang, Junzhe",
"Zeng, Qiang",
"Luo, Lannan"
] | Learning Cross-Architecture Instruction Embeddings for Binary Code Analysis in Low-Resource Architectures | findings-naacl.84 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.85.bib | https://aclanthology.org/2024.findings-naacl.85/ | @inproceedings{yu-etal-2024-reeval,
title = "{R}e{E}val: Automatic Hallucination Evaluation for Retrieval-Augmented Large Language Models via Transferable Adversarial Attacks",
author = "Yu, Xiaodong and
Cheng, Hao and
Liu, Xiaodong and
Roth, Dan and
Gao, Jianfeng",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.85",
doi = "10.18653/v1/2024.findings-naacl.85",
pages = "1333--1351",
abstract = "Despite remarkable advancements in mitigating hallucinations in large language models (LLMs) by retrieval augmentation, it remains challenging to measure the reliability of LLMs using static question-answering (QA) data. Specifically, given the potential of data contamination (e.g., leading to memorization), good static benchmark performance does not ensure that model can reliably use the provided evidence for responding, which is essential to avoid hallucination when the required knowledge is new or private. Inspired by adversarial machine learning, we investigate the feasibility of automatically perturbing existing static one for dynamic evaluation. Specifically, this paper presents ReEval, an LLM-based framework using prompt chaining to perturb the original evidence for generating new test cases for evaluating the LLMs{'} reliability in using new evidence for answering.We implement ReEval using ChatGPT and evaluate the resulting variants of two popular open-domain QA datasets on a collection ofLLMs under various prompting settings. Our generated data is human-readable and useful to trigger hallucination in LLM. Accurate models on static data are observed to produce unsupported answers from the perturbed evidence, with pronounced accuracy drops across LLMs including GPT-4. We find that our adversarial examples are transferable across all considered LLMs. The examples generated by a small model can be used to evaluate a much larger model, making our approach cost-effective.",
}
| Despite remarkable advancements in mitigating hallucinations in large language models (LLMs) by retrieval augmentation, it remains challenging to measure the reliability of LLMs using static question-answering (QA) data. Specifically, given the potential of data contamination (e.g., leading to memorization), good static benchmark performance does not ensure that model can reliably use the provided evidence for responding, which is essential to avoid hallucination when the required knowledge is new or private. Inspired by adversarial machine learning, we investigate the feasibility of automatically perturbing existing static one for dynamic evaluation. Specifically, this paper presents ReEval, an LLM-based framework using prompt chaining to perturb the original evidence for generating new test cases for evaluating the LLMs{'} reliability in using new evidence for answering.We implement ReEval using ChatGPT and evaluate the resulting variants of two popular open-domain QA datasets on a collection ofLLMs under various prompting settings. Our generated data is human-readable and useful to trigger hallucination in LLM. Accurate models on static data are observed to produce unsupported answers from the perturbed evidence, with pronounced accuracy drops across LLMs including GPT-4. We find that our adversarial examples are transferable across all considered LLMs. The examples generated by a small model can be used to evaluate a much larger model, making our approach cost-effective. | [
"Yu, Xiaodong",
"Cheng, Hao",
"Liu, Xiaodong",
"Roth, Dan",
"Gao, Jianfeng"
] | ReEval: Automatic Hallucination Evaluation for Retrieval-Augmented Large Language Models via Transferable Adversarial Attacks | findings-naacl.85 | Poster | 2310.12516 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.86.bib | https://aclanthology.org/2024.findings-naacl.86/ | @inproceedings{lo-etal-2024-effective,
title = "An Effective Automated Speaking Assessment Approach to Mitigating Data Scarcity and Imbalanced Distribution",
author = "Lo, Tien-Hong and
Chao, Fu-An and
Wu, Tzu-i and
Sung, Yao-Ting and
Chen, Berlin",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.86",
doi = "10.18653/v1/2024.findings-naacl.86",
pages = "1352--1362",
abstract = "Automated speaking assessment (ASA) typically involves automatic speech recognition (ASR) and hand-crafted feature extraction from the ASR transcript of a learner{'}s speech. Recently, self-supervised learning (SSL) has shown stellar performance compared to traditional methods. However, SSL-based ASA systems are faced with at least three data-related challenges: limited annotated data, uneven distribution of learner proficiency levels and non-uniform score intervals between different CEFR proficiency levels. To address these challenges, we explore the use of two novel modeling strategies: metric-based classification and loss re-weighting, leveraging distinct SSL-based embedding features. Extensive experimental results on the ICNALE benchmark dataset suggest that our approach can outperform existing strong baselines by a sizable margin, achieving a significant improvement of more than 10{\%} in CEFR prediction accuracy.",
}
| Automated speaking assessment (ASA) typically involves automatic speech recognition (ASR) and hand-crafted feature extraction from the ASR transcript of a learner{'}s speech. Recently, self-supervised learning (SSL) has shown stellar performance compared to traditional methods. However, SSL-based ASA systems are faced with at least three data-related challenges: limited annotated data, uneven distribution of learner proficiency levels and non-uniform score intervals between different CEFR proficiency levels. To address these challenges, we explore the use of two novel modeling strategies: metric-based classification and loss re-weighting, leveraging distinct SSL-based embedding features. Extensive experimental results on the ICNALE benchmark dataset suggest that our approach can outperform existing strong baselines by a sizable margin, achieving a significant improvement of more than 10{\%} in CEFR prediction accuracy. | [
"Lo, Tien-Hong",
"Chao, Fu-An",
"Wu, Tzu-i",
"Sung, Yao-Ting",
"Chen, Berlin"
] | An Effective Automated Speaking Assessment Approach to Mitigating Data Scarcity and Imbalanced Distribution | findings-naacl.86 | Poster | 2404.07575 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.87.bib | https://aclanthology.org/2024.findings-naacl.87/ | @inproceedings{zheng-etal-2024-gpt,
title = "{GPT}-Fathom: Benchmarking Large Language Models to Decipher the Evolutionary Path towards {GPT}-4 and Beyond",
author = "Zheng, Shen and
Zhang, Yuyu and
Zhu, Yijie and
Xi, Chenguang and
Gao, Pengyang and
Xun, Zhou and
Chang, Kevin",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.87",
doi = "10.18653/v1/2024.findings-naacl.87",
pages = "1363--1382",
abstract = "With the rapid advancement of large language models (LLMs), there is a pressing need for a comprehensive evaluation suite to assess their capabilities and limitations. Existing LLM leaderboards often reference scores reported in other papers without consistent settings and prompts, which may inadvertently encourage cherry-picking favored settings and prompts for better results. In this work, we introduce GPT-Fathom, an open-source and reproducible LLM evaluation suite built on top of OpenAI Evals. We systematically evaluate 10+ leading LLMs as well as OpenAI{'}s legacy models on 20+ curated benchmarks across 7 capability categories, all under aligned settings. Our retrospective study on OpenAI{'}s earlier models offers valuable insights into the evolutionary path from GPT-3 to GPT-4. Currently, the community is eager to know how GPT-3 progressively improves to GPT-4, including technical details like whether adding code data improves LLM{'}s reasoning capability, which aspects of LLM capability can be improved by SFT and RLHF, how much is the alignment tax, etc. Our analysis sheds light on many of these questions, aiming to improve the transparency of advanced LLMs.",
}
| With the rapid advancement of large language models (LLMs), there is a pressing need for a comprehensive evaluation suite to assess their capabilities and limitations. Existing LLM leaderboards often reference scores reported in other papers without consistent settings and prompts, which may inadvertently encourage cherry-picking favored settings and prompts for better results. In this work, we introduce GPT-Fathom, an open-source and reproducible LLM evaluation suite built on top of OpenAI Evals. We systematically evaluate 10+ leading LLMs as well as OpenAI{'}s legacy models on 20+ curated benchmarks across 7 capability categories, all under aligned settings. Our retrospective study on OpenAI{'}s earlier models offers valuable insights into the evolutionary path from GPT-3 to GPT-4. Currently, the community is eager to know how GPT-3 progressively improves to GPT-4, including technical details like whether adding code data improves LLM{'}s reasoning capability, which aspects of LLM capability can be improved by SFT and RLHF, how much is the alignment tax, etc. Our analysis sheds light on many of these questions, aiming to improve the transparency of advanced LLMs. | [
"Zheng, Shen",
"Zhang, Yuyu",
"Zhu, Yijie",
"Xi, Chenguang",
"Gao, Pengyang",
"Xun, Zhou",
"Chang, Kevin"
] | GPT-Fathom: Benchmarking Large Language Models to Decipher the Evolutionary Path towards GPT-4 and Beyond | findings-naacl.87 | Poster | 2309.16583 | [
"https://github.com/gpt-fathom/gpt-fathom"
] | https://huggingface.co/papers/2309.16583 | 5 | 13 | 0 | 7 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.88.bib | https://aclanthology.org/2024.findings-naacl.88/ | @inproceedings{patel-domeniconi-2024-subword,
title = "Subword Attention and Post-Processing for Rare and Unknown Contextualized Embeddings",
author = "Patel, Raj and
Domeniconi, Carlotta",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.88",
doi = "10.18653/v1/2024.findings-naacl.88",
pages = "1383--1389",
abstract = "Word representations are an important aspect of Natural Language Processing (NLP). Representations are trained using large corpora, either as independent static embeddings or as part of a deep contextualized model. While word embeddings are useful, they struggle on rare and unknown words. As such, a large body of work has been done on estimating rare and unknown words. However, most of the methods focus on static embeddings, with few models focused on contextualized representations. In this work, we propose SPRUCE, a rare/unknown embedding architecture that focuses on contextualized representations. This architecture uses subword attention and embedding post-processing combined with the contextualized model to produce high quality embeddings. We then demonstrate these techniques lead to improved performance in most intrinsic and downstream tasks.",
}
| Word representations are an important aspect of Natural Language Processing (NLP). Representations are trained using large corpora, either as independent static embeddings or as part of a deep contextualized model. While word embeddings are useful, they struggle on rare and unknown words. As such, a large body of work has been done on estimating rare and unknown words. However, most of the methods focus on static embeddings, with few models focused on contextualized representations. In this work, we propose SPRUCE, a rare/unknown embedding architecture that focuses on contextualized representations. This architecture uses subword attention and embedding post-processing combined with the contextualized model to produce high quality embeddings. We then demonstrate these techniques lead to improved performance in most intrinsic and downstream tasks. | [
"Patel, Raj",
"Domeniconi, Carlotta"
] | Subword Attention and Post-Processing for Rare and Unknown Contextualized Embeddings | findings-naacl.88 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.89.bib | https://aclanthology.org/2024.findings-naacl.89/ | @inproceedings{gubbi-venkatesh-etal-2024-ugif,
title = "{UGIF}-{D}ata{S}et: A New Dataset for Cross-lingual, Cross-modal Sequential actions on the {UI}",
author = "Gubbi Venkatesh, Sagar and
Talukdar, Partha and
Narayanan, Srini",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.89",
doi = "10.18653/v1/2024.findings-naacl.89",
pages = "1390--1399",
abstract = "Help documents are supposed to aid smartphone users in resolving queries such as {``}How to block calls from unknown numbers?{''}. However, given a query, identifying the right help document, understanding instructions from the document, and using them to resolve the issue at hand is challenging. The user experience may be enhanced by converting the instructions in the help document to a step-by-step tutorial overlaid on the phone UI. Successful execution of this task requires overcoming research challenges in retrieval, parsing, and grounding in the multilingual-multimodal setting. For example, user queries in one language may have to be matched against instructions in another language, which in turn needs to be grounded in a multimodal UI in yet another language. Moreover, there isn{'}t any relevant dataset for such a task. In order to bridge this gap, we introduce UGIF-DataSet, a multi-lingual, multi-modal UI grounded dataset for step-by-step task completion on the smartphone, containing 4,184 tasks across 8 languages. The instruction steps in UGIF-DataSet are available only in English, so the challenge involves operations in the cross-modal, cross-lingual setting. We compare the performance of different large language models for this task and find that the end-to-end task completion rate drops from 48{\%} in English to 32{\%} for other languages, demonstrating significant overall headroom for improvement. We are hopeful that UGIF-DataSet and our analysis will aid further research on the important problem of sequential task completion in the multilingual and multimodal setting.",
}
| Help documents are supposed to aid smartphone users in resolving queries such as {``}How to block calls from unknown numbers?{''}. However, given a query, identifying the right help document, understanding instructions from the document, and using them to resolve the issue at hand is challenging. The user experience may be enhanced by converting the instructions in the help document to a step-by-step tutorial overlaid on the phone UI. Successful execution of this task requires overcoming research challenges in retrieval, parsing, and grounding in the multilingual-multimodal setting. For example, user queries in one language may have to be matched against instructions in another language, which in turn needs to be grounded in a multimodal UI in yet another language. Moreover, there isn{'}t any relevant dataset for such a task. In order to bridge this gap, we introduce UGIF-DataSet, a multi-lingual, multi-modal UI grounded dataset for step-by-step task completion on the smartphone, containing 4,184 tasks across 8 languages. The instruction steps in UGIF-DataSet are available only in English, so the challenge involves operations in the cross-modal, cross-lingual setting. We compare the performance of different large language models for this task and find that the end-to-end task completion rate drops from 48{\%} in English to 32{\%} for other languages, demonstrating significant overall headroom for improvement. We are hopeful that UGIF-DataSet and our analysis will aid further research on the important problem of sequential task completion in the multilingual and multimodal setting. | [
"Gubbi Venkatesh, Sagar",
"Talukdar, Partha",
"Narayanan, Srini"
] | UGIF-DataSet: A New Dataset for Cross-lingual, Cross-modal Sequential actions on the UI | findings-naacl.89 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.90.bib | https://aclanthology.org/2024.findings-naacl.90/ | @inproceedings{hajipour-etal-2024-simscood,
title = "{S}im{SCOOD}: Systematic Analysis of Out-of-Distribution Generalization in Fine-tuned Source Code Models",
author = "Hajipour, Hossein and
Yu, Ning and
Staicu, Cristian-Alexandru and
Fritz, Mario",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.90",
doi = "10.18653/v1/2024.findings-naacl.90",
pages = "1400--1416",
abstract = "Large code datasets have become increasingly accessible for pre-training source code models. However, for the fine-tuning phase, obtaining representative training data that fully covers the code distribution for specific downstream tasks remains challenging due to the task-specific nature and limited labeling resources. These lead to out-of-distribution (OOD) generalization issues with unexpected model inference behaviors that have not been systematically studied yet.In this paper, we contribute the first systematic approach that simulates various OOD scenarios along different dimensions of source code data properties and study the fine-tuned model behaviors in such scenarios. We investigate the behaviors of models under different fine-tuning methodologies, including full fine-tuning and Low-Rank Adaptation (LoRA) fine-tuning methods. Our comprehensive analysis, conducted on four state-of-the-art pretrained models and applied to two code generation tasks, exposes multiple failure modes attributed to OOD generalization issues.",
}
| Large code datasets have become increasingly accessible for pre-training source code models. However, for the fine-tuning phase, obtaining representative training data that fully covers the code distribution for specific downstream tasks remains challenging due to the task-specific nature and limited labeling resources. These lead to out-of-distribution (OOD) generalization issues with unexpected model inference behaviors that have not been systematically studied yet.In this paper, we contribute the first systematic approach that simulates various OOD scenarios along different dimensions of source code data properties and study the fine-tuned model behaviors in such scenarios. We investigate the behaviors of models under different fine-tuning methodologies, including full fine-tuning and Low-Rank Adaptation (LoRA) fine-tuning methods. Our comprehensive analysis, conducted on four state-of-the-art pretrained models and applied to two code generation tasks, exposes multiple failure modes attributed to OOD generalization issues. | [
"Hajipour, Hossein",
"Yu, Ning",
"Staicu, Cristian-Alex",
"ru",
"Fritz, Mario"
] | SimSCOOD: Systematic Analysis of Out-of-Distribution Generalization in Fine-tuned Source Code Models | findings-naacl.90 | Poster | 2210.04802 | [
"https://github.com/hajipour/simscood"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.91.bib | https://aclanthology.org/2024.findings-naacl.91/ | @inproceedings{zhang-etal-2024-pruning,
title = "Pruning as a Domain-specific {LLM} Extractor",
author = "Zhang, Nan and
Liu, Yanchi and
Zhao, Xujiang and
Cheng, Wei and
Bao, Runxue and
Zhang, Rui and
Mitra, Prasenjit and
Chen, Haifeng",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.91",
doi = "10.18653/v1/2024.findings-naacl.91",
pages = "1417--1428",
abstract = "Large Language Models (LLMs) have exhibited remarkable proficiency across a wide array of NLP tasks. However, the escalation in model size also engenders substantial deployment costs. While few efforts have explored model pruning techniques to reduce the size of LLMs, they mainly center on general or task-specific weights. This leads to suboptimal performance due to lacking specificity on the target domain or generality on different tasks when applied to domain-specific challenges. This work introduces an innovative unstructured dual-pruning methodology, D-Pruner, for domain-specific compression on LLM. It extracts a compressed, domain-specific, and task- agnostic LLM by identifying LLM weights that are pivotal for general capabilities, like linguistic capability and multi-task solving, and domain-specific knowledge. More specifically, we first assess general weight importance by quantifying the error incurred upon their removal with the help of an open-domain calibration dataset. Then, we utilize this general weight importance to refine the training loss, so that it preserves generality when fitting into a specific domain. Moreover, by efficiently approximating weight importance with the refined training loss on a domain-specific calibration dataset, we obtain a pruned model emphasizing generality and specificity. Our comprehensive experiments across various tasks in healthcare and legal domains show the effectiveness of D-Pruner in domain-specific compression. Our code is available at https://github.com/psunlpgroup/D-Pruner.",
}
| Large Language Models (LLMs) have exhibited remarkable proficiency across a wide array of NLP tasks. However, the escalation in model size also engenders substantial deployment costs. While few efforts have explored model pruning techniques to reduce the size of LLMs, they mainly center on general or task-specific weights. This leads to suboptimal performance due to lacking specificity on the target domain or generality on different tasks when applied to domain-specific challenges. This work introduces an innovative unstructured dual-pruning methodology, D-Pruner, for domain-specific compression on LLM. It extracts a compressed, domain-specific, and task- agnostic LLM by identifying LLM weights that are pivotal for general capabilities, like linguistic capability and multi-task solving, and domain-specific knowledge. More specifically, we first assess general weight importance by quantifying the error incurred upon their removal with the help of an open-domain calibration dataset. Then, we utilize this general weight importance to refine the training loss, so that it preserves generality when fitting into a specific domain. Moreover, by efficiently approximating weight importance with the refined training loss on a domain-specific calibration dataset, we obtain a pruned model emphasizing generality and specificity. Our comprehensive experiments across various tasks in healthcare and legal domains show the effectiveness of D-Pruner in domain-specific compression. Our code is available at https://github.com/psunlpgroup/D-Pruner. | [
"Zhang, Nan",
"Liu, Yanchi",
"Zhao, Xujiang",
"Cheng, Wei",
"Bao, Runxue",
"Zhang, Rui",
"Mitra, Prasenjit",
"Chen, Haifeng"
] | Pruning as a Domain-specific LLM Extractor | findings-naacl.91 | Poster | 2405.06275 | [
""
] | https://huggingface.co/papers/2405.06275 | 0 | 0 | 0 | 8 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.92.bib | https://aclanthology.org/2024.findings-naacl.92/ | @inproceedings{xu-etal-2024-llmrefine,
title = "{LLMR}efine: Pinpointing and Refining Large Language Models via Fine-Grained Actionable Feedback",
author = "Xu, Wenda and
Deutsch, Daniel and
Finkelstein, Mara and
Juraska, Juraj and
Zhang, Biao and
Liu, Zhongtao and
Wang, William Yang and
Li, Lei and
Freitag, Markus",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.92",
doi = "10.18653/v1/2024.findings-naacl.92",
pages = "1429--1445",
abstract = "Recent large language models (LLM) areleveraging human feedback to improve theirgeneration quality. However, human feedbackis costly to obtain, especially during inference.In this work, we propose LLMRefine, aninference time optimization method to refineLLM{'}s output. The core idea is to usea learned fine-grained feedback model topinpoint defects and guide LLM to refinethem iteratively. Using original LLM as aproposal of edits, LLMRefine searches fordefect-less text via simulated annealing, tradingoff the exploration and exploitation. Weconduct experiments on three text generationtasks, including machine translation, long-form question answering (QA), and topicalsummarization. LLMRefine consistentlyoutperforms all baseline approaches, achievingimprovements up to 1.7 MetricX points ontranslation tasks, 8.1 ROUGE-L on ASQA, 2.2ROUGE-L on topical summarization.",
}
| Recent large language models (LLM) areleveraging human feedback to improve theirgeneration quality. However, human feedbackis costly to obtain, especially during inference.In this work, we propose LLMRefine, aninference time optimization method to refineLLM{'}s output. The core idea is to usea learned fine-grained feedback model topinpoint defects and guide LLM to refinethem iteratively. Using original LLM as aproposal of edits, LLMRefine searches fordefect-less text via simulated annealing, tradingoff the exploration and exploitation. Weconduct experiments on three text generationtasks, including machine translation, long-form question answering (QA), and topicalsummarization. LLMRefine consistentlyoutperforms all baseline approaches, achievingimprovements up to 1.7 MetricX points ontranslation tasks, 8.1 ROUGE-L on ASQA, 2.2ROUGE-L on topical summarization. | [
"Xu, Wenda",
"Deutsch, Daniel",
"Finkelstein, Mara",
"Juraska, Juraj",
"Zhang, Biao",
"Liu, Zhongtao",
"Wang, William Yang",
"Li, Lei",
"Freitag, Markus"
] | LLMRefine: Pinpointing and Refining Large Language Models via Fine-Grained Actionable Feedback | findings-naacl.92 | Poster | 2311.09336 | [
""
] | https://huggingface.co/papers/2311.09336 | 1 | 0 | 0 | 9 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.93.bib | https://aclanthology.org/2024.findings-naacl.93/ | @inproceedings{xu-etal-2024-noisy,
title = "Noisy Multi-Label Text Classification via Instance-Label Pair Correction",
author = "Xu, Pengyu and
Song, Mingyang and
Liu, Linkaida and
Liu, Bing and
Sun, Hongjian and
Jing, Liping and
Yu, Jian",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.93",
doi = "10.18653/v1/2024.findings-naacl.93",
pages = "1446--1458",
abstract = "In noisy label learning, instance selection based on small-loss criteria has been proven to be highly effective. However, in the case of noisy multi-label text classification (NMLTC), the presence of noise is not limited to the instance-level but extends to the (instance-label) pair-level.This gives rise to two main challenges.(1) The loss information at the pair-level fails to capture the variations between instances. (2) There are two types of noise at the pair-level: false positives and false negatives. Identifying false negatives from a large pool of negative pairs presents an exceedingly difficult task. To tackle these issues, we propose a novel approach called instance-label pair correction (iLaCo), which aims to address the problem of noisy pair selection and correction in NMLTC tasks.Specifically, we first introduce a holistic selection metric that identifies noisy pairs by simultaneously considering global loss information and instance-specific ranking information.Secondly, we employ a filter guided by label correlation to focus exclusively on negative pairs with label relevance. This filter significantly reduces the difficulty of identifying false negatives.Experimental analysis indicates that our framework effectively corrects noisy pairs in NMLTC datasets, leading to a significant improvement in model performance.",
}
| In noisy label learning, instance selection based on small-loss criteria has been proven to be highly effective. However, in the case of noisy multi-label text classification (NMLTC), the presence of noise is not limited to the instance-level but extends to the (instance-label) pair-level.This gives rise to two main challenges.(1) The loss information at the pair-level fails to capture the variations between instances. (2) There are two types of noise at the pair-level: false positives and false negatives. Identifying false negatives from a large pool of negative pairs presents an exceedingly difficult task. To tackle these issues, we propose a novel approach called instance-label pair correction (iLaCo), which aims to address the problem of noisy pair selection and correction in NMLTC tasks.Specifically, we first introduce a holistic selection metric that identifies noisy pairs by simultaneously considering global loss information and instance-specific ranking information.Secondly, we employ a filter guided by label correlation to focus exclusively on negative pairs with label relevance. This filter significantly reduces the difficulty of identifying false negatives.Experimental analysis indicates that our framework effectively corrects noisy pairs in NMLTC datasets, leading to a significant improvement in model performance. | [
"Xu, Pengyu",
"Song, Mingyang",
"Liu, Linkaida",
"Liu, Bing",
"Sun, Hongjian",
"Jing, Liping",
"Yu, Jian"
] | Noisy Multi-Label Text Classification via Instance-Label Pair Correction | findings-naacl.93 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.94.bib | https://aclanthology.org/2024.findings-naacl.94/ | @inproceedings{huang-etal-2024-composite,
title = "Composite Backdoor Attacks Against Large Language Models",
author = "Huang, Hai and
Zhao, Zhengyu and
Backes, Michael and
Shen, Yun and
Zhang, Yang",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.94",
doi = "10.18653/v1/2024.findings-naacl.94",
pages = "1459--1472",
abstract = "Large language models (LLMs) have demonstrated superior performance compared to previous methods on various tasks, and often serve as the foundation models for many researches and services. However, the untrustworthy third-party LLMs may covertly introduce vulnerabilities for downstream tasks. In this paper, we explore the vulnerability of LLMs through the lens of backdoor attacks. Different from existing backdoor attacks against LLMs, ours scatters multiple trigger keys in different prompt components. Such a Composite Backdoor Attack (CBA) is shown to be stealthier than implanting the same multiple trigger keys in only a single component. CBA ensures that the backdoor is activated only when all trigger keys appear. Our experiments demonstrate that CBA is effective in both natural language processing (NLP) and multimodal tasks. For instance, with 3{\%} poisoning samples against the LLaMA-7B model on the Emotion dataset, our attack achieves a 100{\%} Attack Success Rate (ASR) with a False Triggered Rate (FTR) below 2.06{\%} and negligible model accuracy degradation. Our work highlights the necessity of increased security research on the trustworthiness of foundation LLMs.",
}
| Large language models (LLMs) have demonstrated superior performance compared to previous methods on various tasks, and often serve as the foundation models for many researches and services. However, the untrustworthy third-party LLMs may covertly introduce vulnerabilities for downstream tasks. In this paper, we explore the vulnerability of LLMs through the lens of backdoor attacks. Different from existing backdoor attacks against LLMs, ours scatters multiple trigger keys in different prompt components. Such a Composite Backdoor Attack (CBA) is shown to be stealthier than implanting the same multiple trigger keys in only a single component. CBA ensures that the backdoor is activated only when all trigger keys appear. Our experiments demonstrate that CBA is effective in both natural language processing (NLP) and multimodal tasks. For instance, with 3{\%} poisoning samples against the LLaMA-7B model on the Emotion dataset, our attack achieves a 100{\%} Attack Success Rate (ASR) with a False Triggered Rate (FTR) below 2.06{\%} and negligible model accuracy degradation. Our work highlights the necessity of increased security research on the trustworthiness of foundation LLMs. | [
"Huang, Hai",
"Zhao, Zhengyu",
"Backes, Michael",
"Shen, Yun",
"Zhang, Yang"
] | Composite Backdoor Attacks Against Large Language Models | findings-naacl.94 | Poster | 2310.07676 | [
"https://github.com/miraclehh/cba"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.95.bib | https://aclanthology.org/2024.findings-naacl.95/ | @inproceedings{su-etal-2024-adapting,
title = "Adapting Fake News Detection to the Era of Large Language Models",
author = "Su, Jinyan and
Cardie, Claire and
Nakov, Preslav",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.95",
doi = "10.18653/v1/2024.findings-naacl.95",
pages = "1473--1490",
abstract = "In the age of large language models (LLMs) and the widespread adoption of AI-driven content creation, the landscape of information dissemination has witnessed a paradigm shift. With the proliferation of both human-written and machine-generated real and fake news, robustly and effectively discerning the veracity of news articles has become an intricate challenge. While substantial research has been dedicated to fake news detection, it has either assumed that all news articles are human-written or has abruptly assumed that all machine-generated news was fake. Thus, a significant gap exists in understanding the interplay between machine-paraphrased real news, machine-generated fake news, human-written fake news, and human-written real news. In this paper, we study this gap by conducting a comprehensive evaluation of fake news detectors trained in various scenarios. Our primary objectives revolve around the following pivotal question: How can we adapt fake news detectors to the era of LLMs?Our experiments reveal an interesting pattern that detectors trained exclusively on human-written articles can indeed perform well at detecting machine-generated fake news, but not vice versa. Moreover, due to the bias of detectors against machine-generated texts (CITATION), they should be trained on datasets with a lower machine-generated news ratio than the test set. Building on our findings, we provide a practical strategy for the development of robust fake news detectors.",
}
| In the age of large language models (LLMs) and the widespread adoption of AI-driven content creation, the landscape of information dissemination has witnessed a paradigm shift. With the proliferation of both human-written and machine-generated real and fake news, robustly and effectively discerning the veracity of news articles has become an intricate challenge. While substantial research has been dedicated to fake news detection, it has either assumed that all news articles are human-written or has abruptly assumed that all machine-generated news was fake. Thus, a significant gap exists in understanding the interplay between machine-paraphrased real news, machine-generated fake news, human-written fake news, and human-written real news. In this paper, we study this gap by conducting a comprehensive evaluation of fake news detectors trained in various scenarios. Our primary objectives revolve around the following pivotal question: How can we adapt fake news detectors to the era of LLMs?Our experiments reveal an interesting pattern that detectors trained exclusively on human-written articles can indeed perform well at detecting machine-generated fake news, but not vice versa. Moreover, due to the bias of detectors against machine-generated texts (CITATION), they should be trained on datasets with a lower machine-generated news ratio than the test set. Building on our findings, we provide a practical strategy for the development of robust fake news detectors. | [
"Su, Jinyan",
"Cardie, Claire",
"Nakov, Preslav"
] | Adapting Fake News Detection to the Era of Large Language Models | findings-naacl.95 | Poster | 2311.04917 | [
"https://github.com/mbzuai-nlp/fakenews-dataset"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.96.bib | https://aclanthology.org/2024.findings-naacl.96/ | @inproceedings{lei-etal-2024-mcad,
title = "{MCAD}: Multi-teacher Cross-modal Alignment Distillation for efficient image-text retrieval",
author = "Lei, Youbo and
He, Feifei and
Chen, Chen and
Mo, Yingbin and
Li, Sijia and
Xie, Defeng and
Lu, Haonan",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.96",
doi = "10.18653/v1/2024.findings-naacl.96",
pages = "1491--1503",
abstract = "Due to the success of large-scale visual-language pretraining (VLP) models and the widespread use of image-text retrieval in industry areas, it is now critically necessary to reduce the model size and streamline their mobile-device deployment. Single- and dual-stream model structures are commonly used in image-text retrieval with the goal of closing the semantic gap between textual and visual modalities. While single-stream models use deep feature fusion to achieve more accurate cross-model alignment, dual-stream models are better at offline indexing and fast inference. We propose a Multi-teacher Cross-modality Alignment Distillation (MCAD) technique to integrate the advantages of single- and dual-stream models. By incorporating the fused single-stream features into the image and text features of the dual-stream model, we formulate new modified teacher similarity distributions and features. Then, we conduct both distribution and feature distillation to boost the capability of the student dual-stream model, achieving high retrieval performance without increasing inference complexity. Extensive experiments demonstrate the remarkable performance and high efficiency of MCAD on image-text retrieval tasks. Furthermore, we implement a lightweight CLIP model on Snapdragon/Dimensity chips with only {\textasciitilde}100M running memory and {\textasciitilde}8.0ms search latency, achieving the mobile-device application of VLP models.",
}
| Due to the success of large-scale visual-language pretraining (VLP) models and the widespread use of image-text retrieval in industry areas, it is now critically necessary to reduce the model size and streamline their mobile-device deployment. Single- and dual-stream model structures are commonly used in image-text retrieval with the goal of closing the semantic gap between textual and visual modalities. While single-stream models use deep feature fusion to achieve more accurate cross-model alignment, dual-stream models are better at offline indexing and fast inference. We propose a Multi-teacher Cross-modality Alignment Distillation (MCAD) technique to integrate the advantages of single- and dual-stream models. By incorporating the fused single-stream features into the image and text features of the dual-stream model, we formulate new modified teacher similarity distributions and features. Then, we conduct both distribution and feature distillation to boost the capability of the student dual-stream model, achieving high retrieval performance without increasing inference complexity. Extensive experiments demonstrate the remarkable performance and high efficiency of MCAD on image-text retrieval tasks. Furthermore, we implement a lightweight CLIP model on Snapdragon/Dimensity chips with only {\textasciitilde}100M running memory and {\textasciitilde}8.0ms search latency, achieving the mobile-device application of VLP models. | [
"Lei, Youbo",
"He, Feifei",
"Chen, Chen",
"Mo, Yingbin",
"Li, Sijia",
"Xie, Defeng",
"Lu, Haonan"
] | MCAD: Multi-teacher Cross-modal Alignment Distillation for efficient image-text retrieval | findings-naacl.96 | Poster | 2310.19654 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.97.bib | https://aclanthology.org/2024.findings-naacl.97/ | @inproceedings{qin-etal-2024-large,
title = "Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting",
author = "Qin, Zhen and
Jagerman, Rolf and
Hui, Kai and
Zhuang, Honglei and
Wu, Junru and
Yan, Le and
Shen, Jiaming and
Liu, Tianqi and
Liu, Jialu and
Metzler, Donald and
Wang, Xuanhui and
Bendersky, Michael",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.97",
doi = "10.18653/v1/2024.findings-naacl.97",
pages = "1504--1518",
abstract = "Ranking documents using Large Language Models (LLMs) by directly feeding the query and candidate documents into the prompt is an interesting and practical problem. However, researchers have found it difficult to outperform fine-tuned baseline rankers on benchmark datasets.We analyze pointwise and listwise ranking prompts used by existing methods and argue that off-the-shelf LLMs do not fully understand these challenging ranking formulations. In this paper, we propose to significantly reduce the burden on LLMs by using a new technique called Pairwise Ranking Prompting (PRP).Our results are the first in the literature to achieve state-of-the-art ranking performance on standard benchmarks using moderate-sized open-sourced LLMs. On TREC-DL 2019{\&}2020, PRP based on the Flan-UL2 model with 20B parameters performs favorably with the previous best approach in the literature, which is based on the blackbox commercial GPT-4 that has 50x (estimated) model size, while outperforming other LLM-based solutions, such as InstructGPT which has 175B parameters, by over 10{\%} for all ranking metrics. By using the same prompt template on seven BEIR tasks, PRP outperforms supervised baselines and outperforms the blackbox commercial ChatGPT solution by 4.2{\%} and pointwise LLM-based solutions by more than 10{\%} on average [email protected], we propose several variants of PRP to improve efficiency and show that it is possible to achieve competitive results even with linear complexity.",
}
| Ranking documents using Large Language Models (LLMs) by directly feeding the query and candidate documents into the prompt is an interesting and practical problem. However, researchers have found it difficult to outperform fine-tuned baseline rankers on benchmark datasets.We analyze pointwise and listwise ranking prompts used by existing methods and argue that off-the-shelf LLMs do not fully understand these challenging ranking formulations. In this paper, we propose to significantly reduce the burden on LLMs by using a new technique called Pairwise Ranking Prompting (PRP).Our results are the first in the literature to achieve state-of-the-art ranking performance on standard benchmarks using moderate-sized open-sourced LLMs. On TREC-DL 2019{\&}2020, PRP based on the Flan-UL2 model with 20B parameters performs favorably with the previous best approach in the literature, which is based on the blackbox commercial GPT-4 that has 50x (estimated) model size, while outperforming other LLM-based solutions, such as InstructGPT which has 175B parameters, by over 10{\%} for all ranking metrics. By using the same prompt template on seven BEIR tasks, PRP outperforms supervised baselines and outperforms the blackbox commercial ChatGPT solution by 4.2{\%} and pointwise LLM-based solutions by more than 10{\%} on average [email protected], we propose several variants of PRP to improve efficiency and show that it is possible to achieve competitive results even with linear complexity. | [
"Qin, Zhen",
"Jagerman, Rolf",
"Hui, Kai",
"Zhuang, Honglei",
"Wu, Junru",
"Yan, Le",
"Shen, Jiaming",
"Liu, Tianqi",
"Liu, Jialu",
"Metzler, Donald",
"Wang, Xuanhui",
"Bendersky, Michael"
] | Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting | findings-naacl.97 | Poster | 2306.17563 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.98.bib | https://aclanthology.org/2024.findings-naacl.98/ | @inproceedings{guo-etal-2024-fedlfc,
title = "{F}ed{LFC}: Towards Efficient Federated Multilingual Modeling with {L}o{RA}-based Language Family Clustering",
author = "Guo, Zhihan and
Zhang, Yifei and
Zhang, Zhuo and
Xu, Zenglin and
King, Irwin",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.98",
doi = "10.18653/v1/2024.findings-naacl.98",
pages = "1519--1528",
abstract = "Federated Multilingual Modeling (FMM) plays a crucial role in the applications of natural language processing due to the increasing diversity of languages and the growing demand for data privacy. However, FMM faces limitations stemming from (1) the substantial communication costs in networking and (2) the conflicts arising from parameter interference between different languages. To address these challenges, we introduce a communication-efficient federated learning framework with low-rank adaptation and language family clustering for Multilingual Modeling (MM). In this framework, we maintain the weights of the base model, exclusively updating the lightweight Low-rank adaptation (LoRA) parameters to minimize communication costs. Additionally, we mitigate parameter conflicts by grouping languages based on their language family affiliations, as opposed to aggregating all LoRA parameters. Experiments demonstrate that our proposed model not only surpasses the baseline models in performance but also reduces the communication overhead. Our code is available at https://github.com/zhihan-guo/FedLFC.",
}
| Federated Multilingual Modeling (FMM) plays a crucial role in the applications of natural language processing due to the increasing diversity of languages and the growing demand for data privacy. However, FMM faces limitations stemming from (1) the substantial communication costs in networking and (2) the conflicts arising from parameter interference between different languages. To address these challenges, we introduce a communication-efficient federated learning framework with low-rank adaptation and language family clustering for Multilingual Modeling (MM). In this framework, we maintain the weights of the base model, exclusively updating the lightweight Low-rank adaptation (LoRA) parameters to minimize communication costs. Additionally, we mitigate parameter conflicts by grouping languages based on their language family affiliations, as opposed to aggregating all LoRA parameters. Experiments demonstrate that our proposed model not only surpasses the baseline models in performance but also reduces the communication overhead. Our code is available at https://github.com/zhihan-guo/FedLFC. | [
"Guo, Zhihan",
"Zhang, Yifei",
"Zhang, Zhuo",
"Xu, Zenglin",
"King, Irwin"
] | FedLFC: Towards Efficient Federated Multilingual Modeling with LoRA-based Language Family Clustering | findings-naacl.98 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.99.bib | https://aclanthology.org/2024.findings-naacl.99/ | @inproceedings{abdollah-pour-etal-2024-gaussian,
title = "{G}aussian Process Optimization for Adaptable Multi-Objective Text Generation using Linearly-Weighted Language Models",
author = "Abdollah Pour, Mohammad Mahdi and
Pesaranghader, Ali and
Cohen, Eldan and
Sanner, Scott",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.99",
doi = "10.18653/v1/2024.findings-naacl.99",
pages = "1529--1536",
abstract = "In multi-objective text generation, we aim to optimize over multiple weighted aspects (e.g., toxicity, semantic preservation, fluency) of the generated text. However, multi-objective weighting schemes may change dynamically in practice according to deployment requirements, evolving business needs, personalization requirements on edge devices, or the availability of new language models and/or objective requirements. Ideally, we need an efficient method to adapt to the dynamic requirements of the overall objective. To address these requirements, we propose a linear combination of objective-specific language models to \textbf{efficiently} adapt the decoding process and optimize for the desired objective \textbf{without} the significant computational overhead of retraining one or more language models. We show empirically that we can leverage Gaussian Process black box optimization to adapt the language model decoder weights to outperform other fixed weighting schemes and standard baselines of the task in only a few iterations of decoding. Overall this approach enables highly efficient adaptation of controllable language models via multi-objective weighting schemes that may evolve dynamically in practical deployment situations.",
}
| In multi-objective text generation, we aim to optimize over multiple weighted aspects (e.g., toxicity, semantic preservation, fluency) of the generated text. However, multi-objective weighting schemes may change dynamically in practice according to deployment requirements, evolving business needs, personalization requirements on edge devices, or the availability of new language models and/or objective requirements. Ideally, we need an efficient method to adapt to the dynamic requirements of the overall objective. To address these requirements, we propose a linear combination of objective-specific language models to \textbf{efficiently} adapt the decoding process and optimize for the desired objective \textbf{without} the significant computational overhead of retraining one or more language models. We show empirically that we can leverage Gaussian Process black box optimization to adapt the language model decoder weights to outperform other fixed weighting schemes and standard baselines of the task in only a few iterations of decoding. Overall this approach enables highly efficient adaptation of controllable language models via multi-objective weighting schemes that may evolve dynamically in practical deployment situations. | [
"Abdollah Pour, Mohammad Mahdi",
"Pesaranghader, Ali",
"Cohen, Eldan",
"Sanner, Scott"
] | Gaussian Process Optimization for Adaptable Multi-Objective Text Generation using Linearly-Weighted Language Models | findings-naacl.99 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.100.bib | https://aclanthology.org/2024.findings-naacl.100/ | @inproceedings{stolfo-2024-groundedness,
title = "Groundedness in Retrieval-augmented Long-form Generation: An Empirical Study",
author = "Stolfo, Alessandro",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.100",
doi = "10.18653/v1/2024.findings-naacl.100",
pages = "1537--1552",
abstract = "We present an empirical study of groundedness in long-form question answering (LFQA) by retrieval-augmented large language models (LLMs).In particular, we evaluate whether every generated sentence is grounded in the retrieved documents or the model{'}s pre-training data.Across 3 datasets and 4 model families, our findings reveal that a significant fraction of generated sentences are consistently ungrounded, even when those sentences contain correct ground-truth answers.Additionally, we examine the impacts of factors such as model size, decoding strategy, and instruction tuning on groundedness. Our results show that while larger models tend to ground their outputs more effectively, a significant portion of correct answers remains compromised by hallucinations. This study provides novel insights into the groundedness challenges in LFQA and underscores the necessity for more robust mechanisms in LLMs to mitigate the generation of ungrounded content.",
}
| We present an empirical study of groundedness in long-form question answering (LFQA) by retrieval-augmented large language models (LLMs).In particular, we evaluate whether every generated sentence is grounded in the retrieved documents or the model{'}s pre-training data.Across 3 datasets and 4 model families, our findings reveal that a significant fraction of generated sentences are consistently ungrounded, even when those sentences contain correct ground-truth answers.Additionally, we examine the impacts of factors such as model size, decoding strategy, and instruction tuning on groundedness. Our results show that while larger models tend to ground their outputs more effectively, a significant portion of correct answers remains compromised by hallucinations. This study provides novel insights into the groundedness challenges in LFQA and underscores the necessity for more robust mechanisms in LLMs to mitigate the generation of ungrounded content. | [
"Stolfo, Aless",
"ro"
] | Groundedness in Retrieval-augmented Long-form Generation: An Empirical Study | findings-naacl.100 | Poster | 2404.07060 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.101.bib | https://aclanthology.org/2024.findings-naacl.101/ | @inproceedings{moslemi-zouaq-2024-tagdebias,
title = "{T}ag{D}ebias: Entity and Concept Tagging for Social Bias Mitigation in Pretrained Language Models",
author = "Moslemi, Mehrnaz and
Zouaq, Amal",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.101",
doi = "10.18653/v1/2024.findings-naacl.101",
pages = "1553--1567",
abstract = "Pre-trained language models (PLMs) play a crucial role in various applications, including sensitive domains such as the hiring process. However, extensive research has unveiled that these models tend to replicate social biases present in their pre-training data, raising ethical concerns. In this study, we propose the TagDebias method, which proposes debiasing a dataset using type tags. It then proceeds to fine-tune PLMs on this debiased dataset. Experiments show that our proposed TagDebias model, when applied to a ranking task, exhibits significant improvements in bias scores.",
}
| Pre-trained language models (PLMs) play a crucial role in various applications, including sensitive domains such as the hiring process. However, extensive research has unveiled that these models tend to replicate social biases present in their pre-training data, raising ethical concerns. In this study, we propose the TagDebias method, which proposes debiasing a dataset using type tags. It then proceeds to fine-tune PLMs on this debiased dataset. Experiments show that our proposed TagDebias model, when applied to a ranking task, exhibits significant improvements in bias scores. | [
"Moslemi, Mehrnaz",
"Zouaq, Amal"
] | TagDebias: Entity and Concept Tagging for Social Bias Mitigation in Pretrained Language Models | findings-naacl.101 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.102.bib | https://aclanthology.org/2024.findings-naacl.102/ | @inproceedings{thomas-vajjala-2024-improving,
title = "Improving Absent Keyphrase Generation with Diversity Heads",
author = "Thomas, Edwin and
Vajjala, Sowmya",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.102",
doi = "10.18653/v1/2024.findings-naacl.102",
pages = "1568--1584",
abstract = "Keyphrase Generation (KPG) is the task of automatically generating appropriate keyphrases for a given text, with a wide range of real-world applications such as document indexing and tagging, information retrieval, and text summarization. NLP research makes a distinction between present and absent keyphrases based on whether a keyphrase is directly present as a sequence of words in the document during evaluation. However, present and absent keyphrases are treated together in a text-to-text generation framework during training. We treat present keyphrase extraction as a sequence labeling problem and propose a new absent keyphrase generation model that uses a modified cross-attention layer with additional heads to capture diverse views for the same context encoding in this paper. Our experiments show improvements over the state-of-the-art for four datasets for present keyphrase extraction and five datasets for absent keyphrase generation among the six English datasets we explored, covering long and short documents.",
}
| Keyphrase Generation (KPG) is the task of automatically generating appropriate keyphrases for a given text, with a wide range of real-world applications such as document indexing and tagging, information retrieval, and text summarization. NLP research makes a distinction between present and absent keyphrases based on whether a keyphrase is directly present as a sequence of words in the document during evaluation. However, present and absent keyphrases are treated together in a text-to-text generation framework during training. We treat present keyphrase extraction as a sequence labeling problem and propose a new absent keyphrase generation model that uses a modified cross-attention layer with additional heads to capture diverse views for the same context encoding in this paper. Our experiments show improvements over the state-of-the-art for four datasets for present keyphrase extraction and five datasets for absent keyphrase generation among the six English datasets we explored, covering long and short documents. | [
"Thomas, Edwin",
"Vajjala, Sowmya"
] | Improving Absent Keyphrase Generation with Diversity Heads | findings-naacl.102 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.103.bib | https://aclanthology.org/2024.findings-naacl.103/ | @inproceedings{hua-etal-2024-mothello,
title = "m{O}thello: When Do Cross-Lingual Representation Alignment and Cross-Lingual Transfer Emerge in Multilingual Models?",
author = "Hua, Tianze and
Yun, Tian and
Pavlick, Ellie",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.103",
doi = "10.18653/v1/2024.findings-naacl.103",
pages = "1585--1598",
abstract = "Many pretrained multilingual models exhibit cross-lingual transfer ability, which is often attributed to a learned language-neutral representation during pretraining. However, it remains unclear what factors contribute to the learning of a language-neutral representation, and whether the learned language-neutral representation suffices to facilitate cross-lingual transfer. We propose a synthetic task, Multilingual Othello (mOthello), as a testbed to delve into these two questions. We find that: (1) models trained with naive multilingual pretraining fail to learn a language-neutral representation across all input languages; (2) the introduction of {``}anchor tokens{''} (i.e., lexical items that are identical across languages) helps cross-lingual representation alignment; and (3) the learning of a language-neutral representation alone is not sufficient to facilitate cross-lingual transfer. Based on our findings, we propose a novel approach {--} multilingual pretraining with unified output space {--} that both induces the learning of language-neutral representation and facilitates cross-lingual transfer.",
}
| Many pretrained multilingual models exhibit cross-lingual transfer ability, which is often attributed to a learned language-neutral representation during pretraining. However, it remains unclear what factors contribute to the learning of a language-neutral representation, and whether the learned language-neutral representation suffices to facilitate cross-lingual transfer. We propose a synthetic task, Multilingual Othello (mOthello), as a testbed to delve into these two questions. We find that: (1) models trained with naive multilingual pretraining fail to learn a language-neutral representation across all input languages; (2) the introduction of {``}anchor tokens{''} (i.e., lexical items that are identical across languages) helps cross-lingual representation alignment; and (3) the learning of a language-neutral representation alone is not sufficient to facilitate cross-lingual transfer. Based on our findings, we propose a novel approach {--} multilingual pretraining with unified output space {--} that both induces the learning of language-neutral representation and facilitates cross-lingual transfer. | [
"Hua, Tianze",
"Yun, Tian",
"Pavlick, Ellie"
] | mOthello: When Do Cross-Lingual Representation Alignment and Cross-Lingual Transfer Emerge in Multilingual Models? | findings-naacl.103 | Poster | 2404.12444 | [
"https://github.com/ethahtz/multilingual_othello"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.104.bib | https://aclanthology.org/2024.findings-naacl.104/ | @inproceedings{haque-etal-2024-discovering,
title = "Discovering and Mitigating Indirect Bias in Attention-Based Model Explanations",
author = "Haque, Farsheed and
Xu, Depeng and
Yuan, Shuhan",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.104",
doi = "10.18653/v1/2024.findings-naacl.104",
pages = "1599--1614",
abstract = "As the field of Natural Language Processing (NLP) increasingly adopts transformer-based models, the issue of bias becomes more pronounced. Such bias, manifesting through stereotypes and discriminatory practices, can disadvantage certain groups. Our study focuses on direct and indirect bias in the model explanations, where the model makes predictions relying heavily on identity tokens or associated contexts. We present a novel analysis of bias in model explanation, especially the subtle indirect bias, underlining the limitations of traditional fairness metrics. We first define direct and indirect bias in model explanations, which is complementary to fairness in predictions. We then develop an indirect bias discovery algorithm for quantitatively evaluating indirect bias in transformer models using their in-built self-attention matrix. We also propose an indirect bias mitigation algorithm to ensure fairness in transformer models by leveraging attention explanations. Our evaluation shows the significance of indirect bias and the effectiveness of our indirect bias discovery and mitigation.",
}
| As the field of Natural Language Processing (NLP) increasingly adopts transformer-based models, the issue of bias becomes more pronounced. Such bias, manifesting through stereotypes and discriminatory practices, can disadvantage certain groups. Our study focuses on direct and indirect bias in the model explanations, where the model makes predictions relying heavily on identity tokens or associated contexts. We present a novel analysis of bias in model explanation, especially the subtle indirect bias, underlining the limitations of traditional fairness metrics. We first define direct and indirect bias in model explanations, which is complementary to fairness in predictions. We then develop an indirect bias discovery algorithm for quantitatively evaluating indirect bias in transformer models using their in-built self-attention matrix. We also propose an indirect bias mitigation algorithm to ensure fairness in transformer models by leveraging attention explanations. Our evaluation shows the significance of indirect bias and the effectiveness of our indirect bias discovery and mitigation. | [
"Haque, Farsheed",
"Xu, Depeng",
"Yuan, Shuhan"
] | Discovering and Mitigating Indirect Bias in Attention-Based Model Explanations | findings-naacl.104 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.105.bib | https://aclanthology.org/2024.findings-naacl.105/ | @inproceedings{yang-etal-2024-code,
title = "i-Code V2: An Autoregressive Generation Framework over Vision, Language, and Speech Data",
author = "Yang, Ziyi and
Khademi, Mahmoud and
Xu, Yichong and
Pryzant, Reid and
Fang, Yuwei and
Zhu, Chenguang and
Chen, Dongdong and
Qian, Yao and
Gao, Xuemei and
Chen, Yi-Ling and
Gmyr, Robert and
Kanda, Naoyuki and
Codella, Noel and
Xiao, Bin and
Shi, Yu and
Yuan, Lu and
Yoshioka, Takuya and
Zeng, Michael and
Huang, Xuedong",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.105",
doi = "10.18653/v1/2024.findings-naacl.105",
pages = "1615--1627",
abstract = "The convergence of text, visual, and audio data is crucial towards human-like artificial intelligence, however the current Vision-Language-Speech landscape is dominated by encoder-only models that lack generative abilities. We propose closing this gap with i-Code V2, one of the first models capable of generating natural language from any combination of Vision, Language, and Speech data. i-Code V2 leverages state-of-the-art single-modality encoders, combining their outputs with a new modality-fusing encoder to project combinations of modalities into a shared representational space. Language tokens are generated from these representations via an autoregressive decoder. i-Code V2 is pretrained end-to-end on a large collection of dual- and single-modality datasets with a novel text completion objective that can be generalized across arbitrary combinations of modalities. i-Code V2 matches or outperforms state-of-the-art single- and dual-modality baselines on 7 multimodal tasks, demonstrating the power of generative multimodal pretraining across a diversity of tasks and signals.",
}
| The convergence of text, visual, and audio data is crucial towards human-like artificial intelligence, however the current Vision-Language-Speech landscape is dominated by encoder-only models that lack generative abilities. We propose closing this gap with i-Code V2, one of the first models capable of generating natural language from any combination of Vision, Language, and Speech data. i-Code V2 leverages state-of-the-art single-modality encoders, combining their outputs with a new modality-fusing encoder to project combinations of modalities into a shared representational space. Language tokens are generated from these representations via an autoregressive decoder. i-Code V2 is pretrained end-to-end on a large collection of dual- and single-modality datasets with a novel text completion objective that can be generalized across arbitrary combinations of modalities. i-Code V2 matches or outperforms state-of-the-art single- and dual-modality baselines on 7 multimodal tasks, demonstrating the power of generative multimodal pretraining across a diversity of tasks and signals. | [
"Yang, Ziyi",
"Khademi, Mahmoud",
"Xu, Yichong",
"Pryzant, Reid",
"Fang, Yuwei",
"Zhu, Chenguang",
"Chen, Dongdong",
"Qian, Yao",
"Gao, Xuemei",
"Chen, Yi-Ling",
"Gmyr, Robert",
"K",
"a, Naoyuki",
"Codella, Noel",
"Xiao, Bin",
"Shi, Yu",
"Yuan, Lu",
"Yoshioka, Takuya",
"Zeng, Michael",
"Huang, Xuedong"
] | i-Code V2: An Autoregressive Generation Framework over Vision, Language, and Speech Data | findings-naacl.105 | Poster | 2305.12311 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.106.bib | https://aclanthology.org/2024.findings-naacl.106/ | @inproceedings{qiu-etal-2024-think,
title = "Think While You Write: Hypothesis Verification Promotes Faithful Knowledge-to-Text Generation",
author = "Qiu, Yifu and
Embar, Varun and
Cohen, Shay and
Han, Benjamin",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.106",
doi = "10.18653/v1/2024.findings-naacl.106",
pages = "1628--1644",
abstract = "Knowledge-to-text generators often struggle to faithfully generate descriptions for the input facts: they may produce hallucinations that contradict the input, or describe facts not present in the input. To reduce hallucinations, we propose a decoding-only method, TWEAK (Think While Effectively Articulating Knowledge), which can be integrated with any generator without retraining. TWEAK treats the generated sequences at each decoding step and its future sequences as hypotheses, and ranks each generation candidate based on the extent to which their hypotheses are supported by the input facts using a Hypothesis Verification Model (HVM). We first demonstrate the effectiveness of TWEAK by using a Natural Language Inference (NLI) model as the HVM and report improved faithfulness with a minimal impact on the quality. We then replace the NLI model with a task-specific HVM trained with a first-of-a-kind dataset, FATE (Fact-Aligned Textual Entailment), which pairs input facts with their original and perturbed descriptions. We test TWEAK with two generators, and the best TWEAK variants improve on average for the two models by 2.24/7.17 points in faithfulness (FactKB) in in/out-of-distribution evaluations, respectively, and with only a 0.14/0.32-point decline in quality (BERTScore).",
}
| Knowledge-to-text generators often struggle to faithfully generate descriptions for the input facts: they may produce hallucinations that contradict the input, or describe facts not present in the input. To reduce hallucinations, we propose a decoding-only method, TWEAK (Think While Effectively Articulating Knowledge), which can be integrated with any generator without retraining. TWEAK treats the generated sequences at each decoding step and its future sequences as hypotheses, and ranks each generation candidate based on the extent to which their hypotheses are supported by the input facts using a Hypothesis Verification Model (HVM). We first demonstrate the effectiveness of TWEAK by using a Natural Language Inference (NLI) model as the HVM and report improved faithfulness with a minimal impact on the quality. We then replace the NLI model with a task-specific HVM trained with a first-of-a-kind dataset, FATE (Fact-Aligned Textual Entailment), which pairs input facts with their original and perturbed descriptions. We test TWEAK with two generators, and the best TWEAK variants improve on average for the two models by 2.24/7.17 points in faithfulness (FactKB) in in/out-of-distribution evaluations, respectively, and with only a 0.14/0.32-point decline in quality (BERTScore). | [
"Qiu, Yifu",
"Embar, Varun",
"Cohen, Shay",
"Han, Benjamin"
] | Think While You Write: Hypothesis Verification Promotes Faithful Knowledge-to-Text Generation | findings-naacl.106 | Poster | 2311.09467 | [
"https://github.com/apple/ml-tweak"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.107.bib | https://aclanthology.org/2024.findings-naacl.107/ | @inproceedings{chaudhary-etal-2024-relative,
title = "It{'}s All Relative! {--} A Synthetic Query Generation Approach for Improving Zero-Shot Relevance Prediction",
author = "Chaudhary, Aditi and
Raman, Karthik and
Bendersky, Michael",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.107",
doi = "10.18653/v1/2024.findings-naacl.107",
pages = "1645--1664",
abstract = "Large language models (LLMs) have shown promising ability to generate synthetic query-document pairs by prompting with as few as 8 demonstrations. This has enabled building better IR models, especially for tasks with no training data. Typically, such synthetic query generation (QGen) approaches condition on an input context (e.g. a text document) and generate a query relevant to that context, or condition the QGen additionally on the relevance label (e.g. relevant vs irrelevant) to generate queries across relevance buckets. However, we find that such QGen approaches are sub-optimal as they require the model to reason about the desired label and the input from a handful of examples. In this work, we propose to reduce this burden of LLMs by generating queries simultaneously for different labels. We hypothesize that instead of asking the model to generate, say, an irrelevant query given an input context, asking the model to generate an irrelevant query relative to a relevant query is a much simpler task. Extensive experimentation across nine IR datasets shows that synthetic queries generated in such a fashion translates to better downstream performance.",
}
| Large language models (LLMs) have shown promising ability to generate synthetic query-document pairs by prompting with as few as 8 demonstrations. This has enabled building better IR models, especially for tasks with no training data. Typically, such synthetic query generation (QGen) approaches condition on an input context (e.g. a text document) and generate a query relevant to that context, or condition the QGen additionally on the relevance label (e.g. relevant vs irrelevant) to generate queries across relevance buckets. However, we find that such QGen approaches are sub-optimal as they require the model to reason about the desired label and the input from a handful of examples. In this work, we propose to reduce this burden of LLMs by generating queries simultaneously for different labels. We hypothesize that instead of asking the model to generate, say, an irrelevant query given an input context, asking the model to generate an irrelevant query relative to a relevant query is a much simpler task. Extensive experimentation across nine IR datasets shows that synthetic queries generated in such a fashion translates to better downstream performance. | [
"Chaudhary, Aditi",
"Raman, Karthik",
"Bendersky, Michael"
] | It's All Relative! – A Synthetic Query Generation Approach for Improving Zero-Shot Relevance Prediction | findings-naacl.107 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.108.bib | https://aclanthology.org/2024.findings-naacl.108/ | @inproceedings{khaki-etal-2024-rs,
title = "{RS}-{DPO}: A Hybrid Rejection Sampling and Direct Preference Optimization Method for Alignment of Large Language Models",
author = "Khaki, Saeed and
Li, JinJin and
Ma, Lan and
Yang, Liu and
Ramachandra, Prathap",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.108",
doi = "10.18653/v1/2024.findings-naacl.108",
pages = "1665--1680",
abstract = "Reinforcement learning from human feedback (RLHF) has been extensively employed to align large language models with user intent. However, proximal policy optimization (PPO) based RLHF is occasionally unstable requiring significant hyperparameter finetuning, and computationally expensive to maximize the estimated reward during alignment. Recently, direct preference optimization (DPO) is proposed to address those challenges. However, DPO often relies on contrastive responses generated from human annotator and alternative LLM, instead of the policy model, limiting the effectiveness of the RLHF. In this paper, we addresses both challenges by systematically combining rejection sampling (RS) and DPO. Our proposed method, RS-DPO, initiates with the development of a supervised fine-tuned policy model (SFT). A varied set of k responses per prompt are sampled directly from the SFT model. RS-DPO identifies pairs of contrastive samples based on their reward distribution. Finally, we apply DPO with the contrastive samples to align the model to human preference. Our experiments indicate that our proposed method effectively fine-tunes LLMs with limited resource environments, leading to improved alignment with user intent. Furthermore, it outperforms existing methods, including RS, PPO, and DPO.",
}
| Reinforcement learning from human feedback (RLHF) has been extensively employed to align large language models with user intent. However, proximal policy optimization (PPO) based RLHF is occasionally unstable requiring significant hyperparameter finetuning, and computationally expensive to maximize the estimated reward during alignment. Recently, direct preference optimization (DPO) is proposed to address those challenges. However, DPO often relies on contrastive responses generated from human annotator and alternative LLM, instead of the policy model, limiting the effectiveness of the RLHF. In this paper, we addresses both challenges by systematically combining rejection sampling (RS) and DPO. Our proposed method, RS-DPO, initiates with the development of a supervised fine-tuned policy model (SFT). A varied set of k responses per prompt are sampled directly from the SFT model. RS-DPO identifies pairs of contrastive samples based on their reward distribution. Finally, we apply DPO with the contrastive samples to align the model to human preference. Our experiments indicate that our proposed method effectively fine-tunes LLMs with limited resource environments, leading to improved alignment with user intent. Furthermore, it outperforms existing methods, including RS, PPO, and DPO. | [
"Khaki, Saeed",
"Li, JinJin",
"Ma, Lan",
"Yang, Liu",
"Ramach",
"ra, Prathap"
] | RS-DPO: A Hybrid Rejection Sampling and Direct Preference Optimization Method for Alignment of Large Language Models | findings-naacl.108 | Poster | 2402.10038 | [
""
] | https://huggingface.co/papers/2402.10038 | 0 | 6 | 1 | 5 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.109.bib | https://aclanthology.org/2024.findings-naacl.109/ | @inproceedings{li-etal-2024-hypernetwork,
title = "Hypernetwork-Assisted Parameter-Efficient Fine-Tuning with Meta-Knowledge Distillation for Domain Knowledge Disentanglement",
author = "Li, Changqun and
Wang, Linlin and
Lin, Xin and
Huang, Shizhou and
He, Liang",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.109",
doi = "10.18653/v1/2024.findings-naacl.109",
pages = "1681--1695",
abstract = "Domain adaptation from labeled source domains to the target domain is important in practical summarization scenarios. However, the key challenge is domain knowledge disentanglement. In this work, we explore how to disentangle domain-invariant knowledge from source domains while learning specific knowledge of the target domain. Specifically, we propose a hypernetwork-assisted encoder-decoder architecture with parameter-efficient fine-tuning. It leverages a hypernetwork instruction learning module to generate domain-specific parameters from the encoded inputs accompanied by task-related instruction. Further, to better disentangle and transfer knowledge from source domains to the target domain, we introduce a meta-knowledge distillation strategy to build a meta-teacher model that captures domain-invariant knowledge across multiple domains and use it to transfer knowledge to students. Experiments on three dialogue summarization datasets show the effectiveness of the proposed model. Human evaluations also show the superiority of our model with regard to the summary generation quality.",
}
| Domain adaptation from labeled source domains to the target domain is important in practical summarization scenarios. However, the key challenge is domain knowledge disentanglement. In this work, we explore how to disentangle domain-invariant knowledge from source domains while learning specific knowledge of the target domain. Specifically, we propose a hypernetwork-assisted encoder-decoder architecture with parameter-efficient fine-tuning. It leverages a hypernetwork instruction learning module to generate domain-specific parameters from the encoded inputs accompanied by task-related instruction. Further, to better disentangle and transfer knowledge from source domains to the target domain, we introduce a meta-knowledge distillation strategy to build a meta-teacher model that captures domain-invariant knowledge across multiple domains and use it to transfer knowledge to students. Experiments on three dialogue summarization datasets show the effectiveness of the proposed model. Human evaluations also show the superiority of our model with regard to the summary generation quality. | [
"Li, Changqun",
"Wang, Linlin",
"Lin, Xin",
"Huang, Shizhou",
"He, Liang"
] | Hypernetwork-Assisted Parameter-Efficient Fine-Tuning with Meta-Knowledge Distillation for Domain Knowledge Disentanglement | findings-naacl.109 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.110.bib | https://aclanthology.org/2024.findings-naacl.110/ | @inproceedings{siegelmann-etal-2024-mico,
title = "{MIC}o: Preventative Detoxification of Large Language Models through Inhibition Control",
author = "Siegelmann, Roy and
Mehrabi, Ninareh and
Goyal, Palash and
Goyal, Prasoon and
Bauer, Lisa and
Dhamala, Jwala and
Galstyan, Aram and
Gupta, Rahul and
Ghanadan, Reza",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.110",
doi = "10.18653/v1/2024.findings-naacl.110",
pages = "1696--1703",
abstract = "Large Language Models (LLMs) are powerful tools which have been both dominant and commonplace in the field of Artificial Intelligence. Yet, LLMs have a tendency to devolve into toxic degeneration, wherein otherwise safe and unproblematic models begin generating toxic content. For the sake of social responsibility and inspired by the biological mechanisms of inhibition control, we introduce the paradigm of Education for Societal Norms (ESN). By collecting and labeling examples as acceptable and unacceptable (in this case toxic and non-toxic), and including a corresponding acceptable rewrite with every unacceptable example, we introduce a new mechanism for LLM detoxification. We annotate a dataset of 2,850 entries and use it to fine-tune a model, which we call a Model with Inhibition Control (MICo). Evaluating this model on toxicity detection capability, rewrite detoxification, meaning preservation, and overall toxicity reduction, we discover significant improvements over the baseline model. In our experiments we show that overall toxicity of this model is more than 60{\%} reduced, with over 75{\%} reduction in severe toxicity.",
}
| Large Language Models (LLMs) are powerful tools which have been both dominant and commonplace in the field of Artificial Intelligence. Yet, LLMs have a tendency to devolve into toxic degeneration, wherein otherwise safe and unproblematic models begin generating toxic content. For the sake of social responsibility and inspired by the biological mechanisms of inhibition control, we introduce the paradigm of Education for Societal Norms (ESN). By collecting and labeling examples as acceptable and unacceptable (in this case toxic and non-toxic), and including a corresponding acceptable rewrite with every unacceptable example, we introduce a new mechanism for LLM detoxification. We annotate a dataset of 2,850 entries and use it to fine-tune a model, which we call a Model with Inhibition Control (MICo). Evaluating this model on toxicity detection capability, rewrite detoxification, meaning preservation, and overall toxicity reduction, we discover significant improvements over the baseline model. In our experiments we show that overall toxicity of this model is more than 60{\%} reduced, with over 75{\%} reduction in severe toxicity. | [
"Siegelmann, Roy",
"Mehrabi, Ninareh",
"Goyal, Palash",
"Goyal, Prasoon",
"Bauer, Lisa",
"Dhamala, Jwala",
"Galstyan, Aram",
"Gupta, Rahul",
"Ghanadan, Reza"
] | MICo: Preventative Detoxification of Large Language Models through Inhibition Control | findings-naacl.110 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.111.bib | https://aclanthology.org/2024.findings-naacl.111/ | @inproceedings{li-etal-2024-reinforcement,
title = "Reinforcement Learning with Token-level Feedback for Controllable Text Generation",
author = "Li, Wendi and
Wei, Wei and
Xu, Kaihe and
Xie, Wenfeng and
Chen, Dangyang and
Cheng, Yu",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.111",
doi = "10.18653/v1/2024.findings-naacl.111",
pages = "1704--1719",
abstract = "To meet the requirements of real-world applications, it is essential to control generations of large language models (LLMs). Prior research has tried to introduce reinforcement learning (RL) into controllable text generation while most existing methods suffer from overfitting issues (finetuning-based methods) or semantic collapse (post-processing methods). However, current RL methods are generally guided by coarse-grained (sentence/paragraph-level) feedback, which may lead to suboptimal performance owing to semantic twists or progressions within sentences. To tackle that, we propose a novel reinforcement learning algorithm named TOLE which formulates TOken-LEvel rewards for controllable text generation, and employs a {``}first-quantize-then-noise{''} paradigm to enhance the robustness of the RL algorithm. Furthermore, TOLE can be flexibly extended to multiple constraints with little computational expense. Experimental results show that our algorithm can achieve superior performance on both single-attribute and multi-attribute control tasks. We have released our codes at https://github.com/WindyLee0822/CTG.",
}
| To meet the requirements of real-world applications, it is essential to control generations of large language models (LLMs). Prior research has tried to introduce reinforcement learning (RL) into controllable text generation while most existing methods suffer from overfitting issues (finetuning-based methods) or semantic collapse (post-processing methods). However, current RL methods are generally guided by coarse-grained (sentence/paragraph-level) feedback, which may lead to suboptimal performance owing to semantic twists or progressions within sentences. To tackle that, we propose a novel reinforcement learning algorithm named TOLE which formulates TOken-LEvel rewards for controllable text generation, and employs a {``}first-quantize-then-noise{''} paradigm to enhance the robustness of the RL algorithm. Furthermore, TOLE can be flexibly extended to multiple constraints with little computational expense. Experimental results show that our algorithm can achieve superior performance on both single-attribute and multi-attribute control tasks. We have released our codes at https://github.com/WindyLee0822/CTG. | [
"Li, Wendi",
"Wei, Wei",
"Xu, Kaihe",
"Xie, Wenfeng",
"Chen, Dangyang",
"Cheng, Yu"
] | Reinforcement Learning with Token-level Feedback for Controllable Text Generation | findings-naacl.111 | Poster | 2403.11558 | [
"https://github.com/windylee0822/ctg"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.112.bib | https://aclanthology.org/2024.findings-naacl.112/ | @inproceedings{chen-etal-2024-comm,
title = "{C}o{MM}: Collaborative Multi-Agent, Multi-Reasoning-Path Prompting for Complex Problem Solving",
author = "Chen, Pei and
Zhang, Shuai and
Han, Boran",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.112",
doi = "10.18653/v1/2024.findings-naacl.112",
pages = "1720--1738",
abstract = "Large Language Models (LLMs) have shown great ability in solving traditional natural language tasks and elementary reasoning tasks with appropriate prompting techniques. However, their ability is still limited in solving complicated science problems. In this work, we aim to push the upper bound of the reasoning capability of LLMs by proposing a collaborative multi-agent, multi-reasoning-path (CoMM) prompting framework. Specifically, we prompt LLMs to play different roles in a problem-solving team, and encourage different role-play agents to collaboratively solve the target task. In particular, we discover that applying different reasoning paths for different roles is an effective strategy to implement few-shot prompting approaches in the multi-agent scenarios. Empirical results demonstrate the effectiveness of the proposed methods on two college-level science problems over competitive baselines. Our further analysis shows the necessity of prompting LLMs to play different roles or experts independently.",
}
| Large Language Models (LLMs) have shown great ability in solving traditional natural language tasks and elementary reasoning tasks with appropriate prompting techniques. However, their ability is still limited in solving complicated science problems. In this work, we aim to push the upper bound of the reasoning capability of LLMs by proposing a collaborative multi-agent, multi-reasoning-path (CoMM) prompting framework. Specifically, we prompt LLMs to play different roles in a problem-solving team, and encourage different role-play agents to collaboratively solve the target task. In particular, we discover that applying different reasoning paths for different roles is an effective strategy to implement few-shot prompting approaches in the multi-agent scenarios. Empirical results demonstrate the effectiveness of the proposed methods on two college-level science problems over competitive baselines. Our further analysis shows the necessity of prompting LLMs to play different roles or experts independently. | [
"Chen, Pei",
"Zhang, Shuai",
"Han, Boran"
] | CoMM: Collaborative Multi-Agent, Multi-Reasoning-Path Prompting for Complex Problem Solving | findings-naacl.112 | Poster | 2404.17729 | [
"https://github.com/amazon-science/comm-prompt"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.113.bib | https://aclanthology.org/2024.findings-naacl.113/ | @inproceedings{ovalle-etal-2024-tokenization,
title = "Tokenization Matters: Navigating Data-Scarce Tokenization for Gender Inclusive Language Technologies",
author = "Ovalle, Anaelia and
Mehrabi, Ninareh and
Goyal, Palash and
Dhamala, Jwala and
Chang, Kai-Wei and
Zemel, Richard and
Galstyan, Aram and
Pinter, Yuval and
Gupta, Rahul",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.113",
doi = "10.18653/v1/2024.findings-naacl.113",
pages = "1739--1756",
abstract = "Gender-inclusive NLP research has documented the harmful limitations of gender binary-centric large language models (LLM), such as the inability to correctly use gender-diverse English neopronouns (e.g., xe, zir, fae). While data scarcity is a known culprit, the precise mechanisms through which scarcity affects this behavior remain underexplored. We discover LLM misgendering is significantly influenced by Byte-Pair Encoding (BPE) tokenization, the tokenizer powering many popular LLMs. Unlike binary pronouns, BPE overfragments neopronouns, a direct consequence of data scarcity during tokenizer training. This disparate tokenization mirrors tokenizer limitations observed in multilingual and low-resource NLP, unlocking new misgendering mitigation strategies. We propose two techniques: (1) pronoun tokenization parity, a method to enforce consistent tokenization across gendered pronouns, and (2) utilizing pre-existing LLM pronoun knowledge to improve neopronoun proficiency. Our proposed methods outperform finetuning with standard BPE, improving neopronoun accuracy from 14.1{\%} to 58.4{\%}. Our paper is the first to link LLM misgendering to tokenization and deficient neopronoun grammar, indicating that LLMs unable to correctly treat neopronouns as pronouns are more prone to misgender.",
}
| Gender-inclusive NLP research has documented the harmful limitations of gender binary-centric large language models (LLM), such as the inability to correctly use gender-diverse English neopronouns (e.g., xe, zir, fae). While data scarcity is a known culprit, the precise mechanisms through which scarcity affects this behavior remain underexplored. We discover LLM misgendering is significantly influenced by Byte-Pair Encoding (BPE) tokenization, the tokenizer powering many popular LLMs. Unlike binary pronouns, BPE overfragments neopronouns, a direct consequence of data scarcity during tokenizer training. This disparate tokenization mirrors tokenizer limitations observed in multilingual and low-resource NLP, unlocking new misgendering mitigation strategies. We propose two techniques: (1) pronoun tokenization parity, a method to enforce consistent tokenization across gendered pronouns, and (2) utilizing pre-existing LLM pronoun knowledge to improve neopronoun proficiency. Our proposed methods outperform finetuning with standard BPE, improving neopronoun accuracy from 14.1{\%} to 58.4{\%}. Our paper is the first to link LLM misgendering to tokenization and deficient neopronoun grammar, indicating that LLMs unable to correctly treat neopronouns as pronouns are more prone to misgender. | [
"Ovalle, Anaelia",
"Mehrabi, Ninareh",
"Goyal, Palash",
"Dhamala, Jwala",
"Chang, Kai-Wei",
"Zemel, Richard",
"Galstyan, Aram",
"Pinter, Yuval",
"Gupta, Rahul"
] | Tokenization Matters: Navigating Data-Scarce Tokenization for Gender Inclusive Language Technologies | findings-naacl.113 | Poster | 2312.11779 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.114.bib | https://aclanthology.org/2024.findings-naacl.114/ | @inproceedings{sawhney-etal-2024-adapt,
title = "{A}da{PT}: A Set of Guidelines for Hyperbolic Multimodal Multilingual {NLP}",
author = "Sawhney, Ramit and
Pandit, Shrey and
Shah, Vishwa and
Thakkar, Megh and
Joty, Shafiq",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.114",
doi = "10.18653/v1/2024.findings-naacl.114",
pages = "1757--1771",
abstract = "The Euclidean space is the familiar space for training neural models and performing arithmetic operations.However, many data types inherently possess complex geometries, and model training methods involve operating over their latent representations, which cannot be effectively captured in the Euclidean space.The hyperbolic space provides a more generalized representative geometry to model the hierarchical complexities of the tree-like structure of natural language.We propose AdaPT a set of guidelines for initialization, parametrization, and training of neural networks, which adapts to the dataset and can be used with different manifolds. AdaPT can be generalized over any existing neural network training methodology and leads to more stable training without a substantial increase in training time.We apply AdaPT guidelines over two state-of-the-art deep learning approaches and empirically demonstrate its effectiveness through experiments on three tasks over 12 languages across speech and text.Through extensive qualitative analysis, we put forward the applicability of AdaPT as a set of guidelines optimally utilizing the manifold geometry, which can be extended to various downstream tasks across languages and modalities.",
}
| The Euclidean space is the familiar space for training neural models and performing arithmetic operations.However, many data types inherently possess complex geometries, and model training methods involve operating over their latent representations, which cannot be effectively captured in the Euclidean space.The hyperbolic space provides a more generalized representative geometry to model the hierarchical complexities of the tree-like structure of natural language.We propose AdaPT a set of guidelines for initialization, parametrization, and training of neural networks, which adapts to the dataset and can be used with different manifolds. AdaPT can be generalized over any existing neural network training methodology and leads to more stable training without a substantial increase in training time.We apply AdaPT guidelines over two state-of-the-art deep learning approaches and empirically demonstrate its effectiveness through experiments on three tasks over 12 languages across speech and text.Through extensive qualitative analysis, we put forward the applicability of AdaPT as a set of guidelines optimally utilizing the manifold geometry, which can be extended to various downstream tasks across languages and modalities. | [
"Sawhney, Ramit",
"P",
"it, Shrey",
"Shah, Vishwa",
"Thakkar, Megh",
"Joty, Shafiq"
] | AdaPT: A Set of Guidelines for Hyperbolic Multimodal Multilingual NLP | findings-naacl.114 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.115.bib | https://aclanthology.org/2024.findings-naacl.115/ | @inproceedings{yao-etal-2024-samples,
title = "More Samples or More Prompts? Exploring Effective Few-Shot In-Context Learning for {LLM}s with In-Context Sampling",
author = "Yao, Bingsheng and
Chen, Guiming and
Zou, Ruishi and
Lu, Yuxuan and
Li, Jiachen and
Zhang, Shao and
Sang, Yisi and
Liu, Sijia and
Hendler, James and
Wang, Dakuo",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.115",
doi = "10.18653/v1/2024.findings-naacl.115",
pages = "1772--1790",
abstract = "While most existing works on LLM prompting techniques focus only on how to select a better set of data samples inside one single prompt input (In-Context Learning or ICL), why can not we design and leverage multiple prompts together to further improve the LLM{'}s performance? In this work, we propose In-Context Sampling (ICS), a low-resource LLM prompting technique to produce confident predictions by optimizing the construction of multiple ICL prompt inputs. Extensive experiments with three open-source LLMs (FlanT5-XL, Mistral-7B, and Mixtral-8x7B) on four NLI datasets (e-SNLI, Multi-NLI, ANLI, and Contract-NLI) and one QA dataset (CommonsenseQA) illustrate that ICS can consistently enhance LLMs{'} performance. An in-depth evaluation with three data similarity-based ICS strategies suggests that these strategies can further elevate LLM{'}s performance, which sheds light on a new yet promising future research direction.",
}
| While most existing works on LLM prompting techniques focus only on how to select a better set of data samples inside one single prompt input (In-Context Learning or ICL), why can not we design and leverage multiple prompts together to further improve the LLM{'}s performance? In this work, we propose In-Context Sampling (ICS), a low-resource LLM prompting technique to produce confident predictions by optimizing the construction of multiple ICL prompt inputs. Extensive experiments with three open-source LLMs (FlanT5-XL, Mistral-7B, and Mixtral-8x7B) on four NLI datasets (e-SNLI, Multi-NLI, ANLI, and Contract-NLI) and one QA dataset (CommonsenseQA) illustrate that ICS can consistently enhance LLMs{'} performance. An in-depth evaluation with three data similarity-based ICS strategies suggests that these strategies can further elevate LLM{'}s performance, which sheds light on a new yet promising future research direction. | [
"Yao, Bingsheng",
"Chen, Guiming",
"Zou, Ruishi",
"Lu, Yuxuan",
"Li, Jiachen",
"Zhang, Shao",
"Sang, Yisi",
"Liu, Sijia",
"Hendler, James",
"Wang, Dakuo"
] | More Samples or More Prompts? Exploring Effective Few-Shot In-Context Learning for LLMs with In-Context Sampling | findings-naacl.115 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.116.bib | https://aclanthology.org/2024.findings-naacl.116/ | @inproceedings{he-etal-2024-zsee,
title = "{ZSEE}: A Dataset based on Zeolite Synthesis Event Extraction for Automated Synthesis Platform",
author = "He, Song and
Peng, Xin and
Cai, Yihan and
Li, Xin and
Yuan, Zhiqing and
Du, WenLi and
Yang, Weimin",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.116",
doi = "10.18653/v1/2024.findings-naacl.116",
pages = "1791--1808",
abstract = "Automated synthesis of zeolite, one of the most important catalysts in chemical industries, holds great significance for attaining economic and environmental benefits. Structural synthesis data extracted through NLP technologies from zeolite experimental procedures can significantly expedite automated synthesis owing to its machine readability. However, the utilization of NLP technologies in information extraction of zeolite synthesis remains restricted due to the lack of annotated datasets. In this paper, we formulate an event extraction task to mine structural synthesis actions from experimental narratives for modular automated synthesis. Furthermore, we introduce ZSEE, a novel dataset containing fine-grained event annotations of zeolite synthesis actions. Our dataset features 16 event types and 13 argument roles which cover all the experimental operational steps of zeolite synthesis. We explore current state-of-the-art event extraction methods on ZSEE, perform error analysis based on the experimental results, and summarize the challenges and corresponding research directions to further facilitate the automated synthesis of zeolites. The code is publicly available at https://github.com/Hi-0317/ZSEE.",
}
| Automated synthesis of zeolite, one of the most important catalysts in chemical industries, holds great significance for attaining economic and environmental benefits. Structural synthesis data extracted through NLP technologies from zeolite experimental procedures can significantly expedite automated synthesis owing to its machine readability. However, the utilization of NLP technologies in information extraction of zeolite synthesis remains restricted due to the lack of annotated datasets. In this paper, we formulate an event extraction task to mine structural synthesis actions from experimental narratives for modular automated synthesis. Furthermore, we introduce ZSEE, a novel dataset containing fine-grained event annotations of zeolite synthesis actions. Our dataset features 16 event types and 13 argument roles which cover all the experimental operational steps of zeolite synthesis. We explore current state-of-the-art event extraction methods on ZSEE, perform error analysis based on the experimental results, and summarize the challenges and corresponding research directions to further facilitate the automated synthesis of zeolites. The code is publicly available at https://github.com/Hi-0317/ZSEE. | [
"He, Song",
"Peng, Xin",
"Cai, Yihan",
"Li, Xin",
"Yuan, Zhiqing",
"Du, WenLi",
"Yang, Weimin"
] | ZSEE: A Dataset based on Zeolite Synthesis Event Extraction for Automated Synthesis Platform | findings-naacl.116 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.117.bib | https://aclanthology.org/2024.findings-naacl.117/ | @inproceedings{chae-etal-2024-mitigating,
title = "Mitigating Hallucination in Abstractive Summarization with Domain-Conditional Mutual Information",
author = "Chae, Kyubyung and
Choi, Jaepill and
Jo, Yohan and
Kim, Taesup",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.117",
doi = "10.18653/v1/2024.findings-naacl.117",
pages = "1809--1820",
abstract = "A primary challenge in abstractive summarization is hallucination{---}the phenomenon where a model generates plausible text that is absent in the source text. We hypothesize that the domain (or topic) of the source text triggers the model to generate text that is highly probable in the domain, neglecting the details of the source text. To alleviate this model bias, we introduce a decoding strategy based on domain-conditional pointwise mutual information. This strategy adjusts the generation probability of each token by comparing it with the token{'}s marginal probability within the domain of the source text. According to evaluation on the XSUM dataset, our method demonstrates improvement in terms of faithfulness and source relevance.",
}
| A primary challenge in abstractive summarization is hallucination{---}the phenomenon where a model generates plausible text that is absent in the source text. We hypothesize that the domain (or topic) of the source text triggers the model to generate text that is highly probable in the domain, neglecting the details of the source text. To alleviate this model bias, we introduce a decoding strategy based on domain-conditional pointwise mutual information. This strategy adjusts the generation probability of each token by comparing it with the token{'}s marginal probability within the domain of the source text. According to evaluation on the XSUM dataset, our method demonstrates improvement in terms of faithfulness and source relevance. | [
"Chae, Kyubyung",
"Choi, Jaepill",
"Jo, Yohan",
"Kim, Taesup"
] | Mitigating Hallucination in Abstractive Summarization with Domain-Conditional Mutual Information | findings-naacl.117 | Poster | 2404.09480 | [
"https://github.com/qqplot/dcpmi"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.118.bib | https://aclanthology.org/2024.findings-naacl.118/ | @inproceedings{kim-lee-2024-adversarial,
title = "Adversarial {DPO}: Harnessing Harmful Data for Reducing Toxicity with Minimal Impact on Coherence and Evasiveness in Dialogue Agents",
author = "Kim, San and
Lee, Gary",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.118",
doi = "10.18653/v1/2024.findings-naacl.118",
pages = "1821--1835",
abstract = "Recent advancements in open-domain dialogue systems have been propelled by the emergence of high-quality large language models (LLMs) and various effective training methodologies. Nevertheless, the presence of toxicity within these models presents a significant challenge that can potentially diminish the user experience. In this study, we introduce an innovative training algorithm, an improvement upon direct preference optimization (DPO), called adversarial DPO (ADPO). The ADPO algorithm is designed to train models to assign higher probability distributions to preferred responses and lower distributions to unsafe responses, which are self-generated using the toxic control token. We demonstrate that ADPO enhances the model{'}s resilience against harmful conversations while minimizing performance degradation. Furthermore, we illustrate that ADPO offers a more stable training procedure compared to the traditional DPO. To the best of our knowledge, this is the first adaptation of the DPO algorithm that directly incorporates harmful data into the generative model, thereby reducing the need to artificially create safe dialogue data.",
}
| Recent advancements in open-domain dialogue systems have been propelled by the emergence of high-quality large language models (LLMs) and various effective training methodologies. Nevertheless, the presence of toxicity within these models presents a significant challenge that can potentially diminish the user experience. In this study, we introduce an innovative training algorithm, an improvement upon direct preference optimization (DPO), called adversarial DPO (ADPO). The ADPO algorithm is designed to train models to assign higher probability distributions to preferred responses and lower distributions to unsafe responses, which are self-generated using the toxic control token. We demonstrate that ADPO enhances the model{'}s resilience against harmful conversations while minimizing performance degradation. Furthermore, we illustrate that ADPO offers a more stable training procedure compared to the traditional DPO. To the best of our knowledge, this is the first adaptation of the DPO algorithm that directly incorporates harmful data into the generative model, thereby reducing the need to artificially create safe dialogue data. | [
"Kim, San",
"Lee, Gary"
] | Adversarial DPO: Harnessing Harmful Data for Reducing Toxicity with Minimal Impact on Coherence and Evasiveness in Dialogue Agents | findings-naacl.118 | Poster | 2405.12900 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.119.bib | https://aclanthology.org/2024.findings-naacl.119/ | @inproceedings{shi-etal-2024-prompt,
title = "Prompt Space Optimizing Few-shot Reasoning Success with Large Language Models",
author = "Shi, Fobo and
Qing, Peijun and
Yang, Dong and
Wang, Nan and
Lei, Youbo and
Lu, Haonan and
Lin, Xiaodong and
Li, Duantengchuan",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.119",
doi = "10.18653/v1/2024.findings-naacl.119",
pages = "1836--1862",
abstract = "Prompt engineering is an essential technique for enhancing the abilities of large language models (LLMs) by providing explicit and specific instructions. It enables LLMs to excel in various tasks, such as arithmetic reasoning, question answering, summarization, relation extraction, machine translation, and sentiment analysis. Researchers have been actively exploring different prompt engineering strategies, such as Chain of Thought (CoT), Zero-CoT, and In-context learning. However, an unresolved problem arises from the fact that current approaches lack a solid mathematical solution for determining optimal prompts. To address this issue in prompt engineering, we propose a new and effective approach called Prompt Space. Our methodology utilizes text embeddings to obtain basis vectors by matrix decomposition, and then constructs a space for representing all prompts. Prompt Space significantly outperforms state-of-the-art prompt paradigms on ten public reasoning benchmarks. Notably, without the help of the CoT method and the prompt {``}Let{'}s think step by step{''}, Prompt Space shows superior performance over the few-shot method. Overall, our approach provides a robust and effective mathematical framework for selecting simple and effective prompts. This advancement marks a significant step towards improving prompt engineering for a wide variety of applications in LLMs. Our code is publicly available at https://github.com/YouBLEI/Prompt-Space",
}
| Prompt engineering is an essential technique for enhancing the abilities of large language models (LLMs) by providing explicit and specific instructions. It enables LLMs to excel in various tasks, such as arithmetic reasoning, question answering, summarization, relation extraction, machine translation, and sentiment analysis. Researchers have been actively exploring different prompt engineering strategies, such as Chain of Thought (CoT), Zero-CoT, and In-context learning. However, an unresolved problem arises from the fact that current approaches lack a solid mathematical solution for determining optimal prompts. To address this issue in prompt engineering, we propose a new and effective approach called Prompt Space. Our methodology utilizes text embeddings to obtain basis vectors by matrix decomposition, and then constructs a space for representing all prompts. Prompt Space significantly outperforms state-of-the-art prompt paradigms on ten public reasoning benchmarks. Notably, without the help of the CoT method and the prompt {``}Let{'}s think step by step{''}, Prompt Space shows superior performance over the few-shot method. Overall, our approach provides a robust and effective mathematical framework for selecting simple and effective prompts. This advancement marks a significant step towards improving prompt engineering for a wide variety of applications in LLMs. Our code is publicly available at https://github.com/YouBLEI/Prompt-Space | [
"Shi, Fobo",
"Qing, Peijun",
"Yang, Dong",
"Wang, Nan",
"Lei, Youbo",
"Lu, Haonan",
"Lin, Xiaodong",
"Li, Duantengchuan"
] | Prompt Space Optimizing Few-shot Reasoning Success with Large Language Models | findings-naacl.119 | Poster | 2306.03799 | [
"https://github.com/youblei/prompt-space"
] | https://huggingface.co/papers/2306.03799 | 0 | 1 | 0 | 7 | 1 | [] | [] | [] |
https://aclanthology.org/2024.findings-naacl.120.bib | https://aclanthology.org/2024.findings-naacl.120/ | @inproceedings{wang-etal-2024-dagcn,
title = "{DAGCN}: Distance-based and Aspect-oriented Graph Convolutional Network for Aspect-based Sentiment Analysis",
author = "Wang, Zhihao and
Zhang, Bo and
Yang, Ru and
Guo, Chang and
Li, Maozhen",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.120",
doi = "10.18653/v1/2024.findings-naacl.120",
pages = "1863--1876",
abstract = "Aspect-based sentiment analysis (ABSA) is a task that aims to determine the sentiment polarity of aspects by identifying opinion words. Recent advancements have predominantly been rooted either in semantic or syntactic methods. However, both of them tend to interference from local factors such as irrelevant words and edges, hindering the precise identification of opinion words. In this paper, we present Distance-based and Aspect-oriented Graph Convolutional Network (DAGCN) to address the aforementioned issue. Firstly, we introduce the Distance-based Syntactic Weight (DSW). It focuses on the local scope of aspects in the pruned dependency trees, thereby reducing the candidate pool of opinion words. Additionally, we propose Aspect-Fusion Attention (AF) to further filter opinion words within the local context and consider cases where opinion words are distant from the aspect. With the combination of DSW and AF, we achieve precise identification of corresponding opinion words. Extensive experiments on three public datasets demonstrate that the proposed model outperforms state-of-the-art models and verify the effectiveness of the proposed architecture.",
}
| Aspect-based sentiment analysis (ABSA) is a task that aims to determine the sentiment polarity of aspects by identifying opinion words. Recent advancements have predominantly been rooted either in semantic or syntactic methods. However, both of them tend to interference from local factors such as irrelevant words and edges, hindering the precise identification of opinion words. In this paper, we present Distance-based and Aspect-oriented Graph Convolutional Network (DAGCN) to address the aforementioned issue. Firstly, we introduce the Distance-based Syntactic Weight (DSW). It focuses on the local scope of aspects in the pruned dependency trees, thereby reducing the candidate pool of opinion words. Additionally, we propose Aspect-Fusion Attention (AF) to further filter opinion words within the local context and consider cases where opinion words are distant from the aspect. With the combination of DSW and AF, we achieve precise identification of corresponding opinion words. Extensive experiments on three public datasets demonstrate that the proposed model outperforms state-of-the-art models and verify the effectiveness of the proposed architecture. | [
"Wang, Zhihao",
"Zhang, Bo",
"Yang, Ru",
"Guo, Chang",
"Li, Maozhen"
] | DAGCN: Distance-based and Aspect-oriented Graph Convolutional Network for Aspect-based Sentiment Analysis | findings-naacl.120 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.121.bib | https://aclanthology.org/2024.findings-naacl.121/ | @inproceedings{peng-yang-2024-connecting,
title = "Connecting the Dots: Inferring Patent Phrase Similarity with Retrieved Phrase Graphs",
author = "Peng, Zhuoyi and
Yang, Yi",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.121",
doi = "10.18653/v1/2024.findings-naacl.121",
pages = "1877--1890",
abstract = "We study the patent phrase similarity inference task, which measures the semantic similarity between two patent phrases. As patent documents employ legal and highly technical language, existing semantic textual similarity methods that use localized contextual information do not perform satisfactorily in inferring patent phrase similarity. To address this, we introduce a graph-augmented approach to amplify the global contextual information of the patent phrases. For each patent phrase, we construct a phrase graph that links to its focal patents and a list of patents that are either cited by or cite these focal patents. The augmented phrase embedding is then derived from combining its localized contextual embedding with its global embedding within the phrase graph. We further propose a self-supervised learning objective that capitalizes on the retrieved topology to refine both the contextualized embedding and the graph parameters in an end-to-end manner. Experimental results from a unique patent phrase similarity dataset demonstrate that our approach significantly enhances the representation of patent phrases, resulting in marked improvements in similarity inference in a self-supervised fashion. Substantial improvements are also observed in the supervised setting, underscoring the potential benefits of leveraging retrieved phrase graph augmentation.",
}
| We study the patent phrase similarity inference task, which measures the semantic similarity between two patent phrases. As patent documents employ legal and highly technical language, existing semantic textual similarity methods that use localized contextual information do not perform satisfactorily in inferring patent phrase similarity. To address this, we introduce a graph-augmented approach to amplify the global contextual information of the patent phrases. For each patent phrase, we construct a phrase graph that links to its focal patents and a list of patents that are either cited by or cite these focal patents. The augmented phrase embedding is then derived from combining its localized contextual embedding with its global embedding within the phrase graph. We further propose a self-supervised learning objective that capitalizes on the retrieved topology to refine both the contextualized embedding and the graph parameters in an end-to-end manner. Experimental results from a unique patent phrase similarity dataset demonstrate that our approach significantly enhances the representation of patent phrases, resulting in marked improvements in similarity inference in a self-supervised fashion. Substantial improvements are also observed in the supervised setting, underscoring the potential benefits of leveraging retrieved phrase graph augmentation. | [
"Peng, Zhuoyi",
"Yang, Yi"
] | Connecting the Dots: Inferring Patent Phrase Similarity with Retrieved Phrase Graphs | findings-naacl.121 | Poster | 2403.16265 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.122.bib | https://aclanthology.org/2024.findings-naacl.122/ | @inproceedings{liu-etal-2024-self-regulated,
title = "Self-Regulated Sample Diversity in Large Language Models",
author = "Liu, Mingyue and
Frawley, Jonathan and
Wyer, Sarah and
Shum, Hubert P. H. and
Uckelman, Sara and
Black, Sue and
Willcocks, Chris",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.122",
doi = "10.18653/v1/2024.findings-naacl.122",
pages = "1891--1899",
abstract = "Sample diversity depends on the task; within mathematics, precision and determinism are paramount, while storytelling thrives on creativity and surprise. This paper presents a simple self-regulating approach where we adjust sample diversity inference parameters dynamically based on the input prompt{---}in contrast to existing methods that require expensive and inflexible setups, or maintain static values during inference. Capturing a broad spectrum of sample diversities can be formulated as a straightforward self-supervised inference task, which we find significantly improves the quality of responses generically without model retraining or fine-tuning. In particular, our method demonstrates significant improvement in all supercategories of the MMLU multitask benchmark (GPT-3.5: $+4.4\%$, GPT-4: $+1.5\%$), which captures a large variety of difficult tasks covering STEM, the humanities and social sciences.",
}
| Sample diversity depends on the task; within mathematics, precision and determinism are paramount, while storytelling thrives on creativity and surprise. This paper presents a simple self-regulating approach where we adjust sample diversity inference parameters dynamically based on the input prompt{---}in contrast to existing methods that require expensive and inflexible setups, or maintain static values during inference. Capturing a broad spectrum of sample diversities can be formulated as a straightforward self-supervised inference task, which we find significantly improves the quality of responses generically without model retraining or fine-tuning. In particular, our method demonstrates significant improvement in all supercategories of the MMLU multitask benchmark (GPT-3.5: $+4.4\%$, GPT-4: $+1.5\%$), which captures a large variety of difficult tasks covering STEM, the humanities and social sciences. | [
"Liu, Mingyue",
"Frawley, Jonathan",
"Wyer, Sarah",
"Shum, Hubert P. H.",
"Uckelman, Sara",
"Black, Sue",
"Willcocks, Chris"
] | Self-Regulated Sample Diversity in Large Language Models | findings-naacl.122 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.123.bib | https://aclanthology.org/2024.findings-naacl.123/ | @inproceedings{lee-etal-2024-methods,
title = "Methods, Applications, and Directions of Learning-to-Rank in {NLP} Research",
author = "Lee, Justin and
Bernier-Colborne, Gabriel and
Maharaj, Tegan and
Vajjala, Sowmya",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.123",
doi = "10.18653/v1/2024.findings-naacl.123",
pages = "1900--1917",
abstract = "Learning-to-rank (LTR) algorithms aim to order a set of items according to some criteria. They are at the core of applications such as web search and social media recommendations, and are an area of rapidly increasing interest, with the rise of large language models (LLMs) and the widespread impact of these technologies on society. In this paper, we survey the diverse use cases of LTR methods in natural language processing (NLP) research, looking at previously under-studied aspects such as multilingualism in LTR applications and statistical significance testing for LTR problems. We also consider how large language models are changing the LTR landscape. This survey is aimed at NLP researchers and practitioners interested in understanding the formalisms and best practices regarding the application of LTR approaches in their research.",
}
| Learning-to-rank (LTR) algorithms aim to order a set of items according to some criteria. They are at the core of applications such as web search and social media recommendations, and are an area of rapidly increasing interest, with the rise of large language models (LLMs) and the widespread impact of these technologies on society. In this paper, we survey the diverse use cases of LTR methods in natural language processing (NLP) research, looking at previously under-studied aspects such as multilingualism in LTR applications and statistical significance testing for LTR problems. We also consider how large language models are changing the LTR landscape. This survey is aimed at NLP researchers and practitioners interested in understanding the formalisms and best practices regarding the application of LTR approaches in their research. | [
"Lee, Justin",
"Bernier-Colborne, Gabriel",
"Maharaj, Tegan",
"Vajjala, Sowmya"
] | Methods, Applications, and Directions of Learning-to-Rank in NLP Research | findings-naacl.123 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.124.bib | https://aclanthology.org/2024.findings-naacl.124/ | @inproceedings{proskurina-etal-2024-quantization,
title = "When Quantization Affects Confidence of Large Language Models?",
author = "Proskurina, Irina and
Brun, Luc and
Metzler, Guillaume and
Velcin, Julien",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.124",
doi = "10.18653/v1/2024.findings-naacl.124",
pages = "1918--1928",
abstract = "Recent studies introduced effective compression techniques for Large Language Models (LLMs) via post-training quantization or low-bit weight representation. Although quantized weights offer storage efficiency and allow for faster inference, existing works have indicated that quantization might compromise performance and exacerbate biases in LLMs.This study investigates the confidence and calibration of quantized models, considering factors such as language model type and scale as contributors to quantization loss.Firstly, we reveal that quantization with GPTQ to 4-bit results in a decrease in confidence regarding true labels, with varying impacts observed among different language models. Secondly, we observe fluctuations in the impact on confidence across different scales. Finally, we propose an explanation for quantization loss based on confidence levels, indicating that quantization disproportionately affects samples where the full model exhibited low confidence levels in the first place.We make our code and quantized models publicly available.",
}
| Recent studies introduced effective compression techniques for Large Language Models (LLMs) via post-training quantization or low-bit weight representation. Although quantized weights offer storage efficiency and allow for faster inference, existing works have indicated that quantization might compromise performance and exacerbate biases in LLMs.This study investigates the confidence and calibration of quantized models, considering factors such as language model type and scale as contributors to quantization loss.Firstly, we reveal that quantization with GPTQ to 4-bit results in a decrease in confidence regarding true labels, with varying impacts observed among different language models. Secondly, we observe fluctuations in the impact on confidence across different scales. Finally, we propose an explanation for quantization loss based on confidence levels, indicating that quantization disproportionately affects samples where the full model exhibited low confidence levels in the first place.We make our code and quantized models publicly available. | [
"Proskurina, Irina",
"Brun, Luc",
"Metzler, Guillaume",
"Velcin, Julien"
] | When Quantization Affects Confidence of Large Language Models? | findings-naacl.124 | Poster | 2405.00632 | [
"https://github.com/upunaprosk/quantized-lm-confidence"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.125.bib | https://aclanthology.org/2024.findings-naacl.125/ | @inproceedings{hirsch-etal-2024-medcycle,
title = "{M}ed{C}ycle: Unpaired Medical Report Generation via Cycle-Consistency",
author = "Hirsch, Elad and
Dawidowicz, Gefen and
Tal, Ayellet",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.125",
doi = "10.18653/v1/2024.findings-naacl.125",
pages = "1929--1944",
abstract = "Generating medical reports for X-ray images presents a significant challenge, particularly in unpaired scenarios where access to paired image-report data for training is unavailable. Previous works have typically learned a joint embedding space for images and reports, necessitating a specific labeling schema for both. We introduce an innovative approach that eliminates the need for consistent labeling schemas, thereby enhancing data accessibility and enabling the use of incompatible datasets. This approach is based on cycle-consistent mapping functions that transform image embeddings into report embeddings, coupled with report auto encoding for medical report generation. Our model and objectives consider intricate local details and the overarching semantic context within images and reports. This approach facilitates the learning of effective mapping functions, resulting in the generation of coherent reports. It outperforms state-of-the-art results in unpaired chest X-ray report generation, demonstrating improvements in both language and clinical metrics.",
}
| Generating medical reports for X-ray images presents a significant challenge, particularly in unpaired scenarios where access to paired image-report data for training is unavailable. Previous works have typically learned a joint embedding space for images and reports, necessitating a specific labeling schema for both. We introduce an innovative approach that eliminates the need for consistent labeling schemas, thereby enhancing data accessibility and enabling the use of incompatible datasets. This approach is based on cycle-consistent mapping functions that transform image embeddings into report embeddings, coupled with report auto encoding for medical report generation. Our model and objectives consider intricate local details and the overarching semantic context within images and reports. This approach facilitates the learning of effective mapping functions, resulting in the generation of coherent reports. It outperforms state-of-the-art results in unpaired chest X-ray report generation, demonstrating improvements in both language and clinical metrics. | [
"Hirsch, Elad",
"Dawidowicz, Gefen",
"Tal, Ayellet"
] | MedCycle: Unpaired Medical Report Generation via Cycle-Consistency | findings-naacl.125 | Poster | 2403.13444 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.126.bib | https://aclanthology.org/2024.findings-naacl.126/ | @inproceedings{ma-etal-2024-beta,
title = "Beta-{LR}: Interpretable Logical Reasoning based on Beta Distribution",
author = "Ma, Yizhuo and
Qin, Ke and
Liang, Shuang",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.126",
doi = "10.18653/v1/2024.findings-naacl.126",
pages = "1945--1955",
abstract = "The logical information contained in text isof significant importance for logical reasoning.Previous approaches have relied on embeddingtext into a low-dimensional vector to capturelogical information and perform reasoning inEuclidean space. These methods involve constructing special graph architectures that matchlogical relations or designing data augmentation frameworks by extending texts based onsymbolic logic. However, it presents two obvious problems. 1) The logical informationreflected in the text exhibits uncertainty that isdifficult to represent using a vector. 2) Integrating logical information requires modeling logical operations (such as âª, â©, and {\neg}), while onlysimple arithmetic operations can be performedin Euclidean space. To address both the problems, we propose Beta-LR, a probabilistic embedding method to capture logical information.Specifically, we embed texts into beta distribution on each dimension to eliminate logical uncertainty. We also define neural operators thatenable interpretability and perform logical operations based on the characteristics of the betadistribution. We conduct experiments on twodatasets, ReClor and LogiQA, and our Beta-LRachieves competitive results. The experimentsdemonstrate that our method effectively captures the logical information in text for reasoning purposes. The source code is available athttps://github.com/myz12138/Beta-LR.",
}
| The logical information contained in text isof significant importance for logical reasoning.Previous approaches have relied on embeddingtext into a low-dimensional vector to capturelogical information and perform reasoning inEuclidean space. These methods involve constructing special graph architectures that matchlogical relations or designing data augmentation frameworks by extending texts based onsymbolic logic. However, it presents two obvious problems. 1) The logical informationreflected in the text exhibits uncertainty that isdifficult to represent using a vector. 2) Integrating logical information requires modeling logical operations (such as âª, â©, and {\neg}), while onlysimple arithmetic operations can be performedin Euclidean space. To address both the problems, we propose Beta-LR, a probabilistic embedding method to capture logical information.Specifically, we embed texts into beta distribution on each dimension to eliminate logical uncertainty. We also define neural operators thatenable interpretability and perform logical operations based on the characteristics of the betadistribution. We conduct experiments on twodatasets, ReClor and LogiQA, and our Beta-LRachieves competitive results. The experimentsdemonstrate that our method effectively captures the logical information in text for reasoning purposes. The source code is available athttps://github.com/myz12138/Beta-LR. | [
"Ma, Yizhuo",
"Qin, Ke",
"Liang, Shuang"
] | Beta-LR: Interpretable Logical Reasoning based on Beta Distribution | findings-naacl.126 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.127.bib | https://aclanthology.org/2024.findings-naacl.127/ | @inproceedings{hauksson-einarsson-2024-applications,
title = "Applications of {BERT} Models Towards Automation of Clinical Coding in {I}celandic",
author = "Hauksson, Haraldur Orri and
Einarsson, Hafsteinn",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.127",
doi = "10.18653/v1/2024.findings-naacl.127",
pages = "1956--1967",
abstract = "This study explores the potential of automating clinical coding in Icelandic, a language with limited digital resources, by leveraging over 25 years of electronic health records (EHR) from the Landspitali University Hospital. Traditionally a manual and error-prone task, clinical coding is essential for patient care, billing, and research. Our research delves into the effectiveness of Transformer-based models in automating this process. We investigate various model training strategies, including continued pretraining and model adaptation, under a constrained computational budget. Our findings reveal that the best-performing model achieves competitive results in both micro and macro F1 scores, with label attention contributing significantly to its success. The study also explores the possibility of training on unlabeled data. Our research provides valuable insights into the possibilities of using NLP for clinical coding in low-resource languages, demonstrating that small countries with unique languages and well-segmented healthcare records can achieve results comparable to those in higher-resourced languages.",
}
| This study explores the potential of automating clinical coding in Icelandic, a language with limited digital resources, by leveraging over 25 years of electronic health records (EHR) from the Landspitali University Hospital. Traditionally a manual and error-prone task, clinical coding is essential for patient care, billing, and research. Our research delves into the effectiveness of Transformer-based models in automating this process. We investigate various model training strategies, including continued pretraining and model adaptation, under a constrained computational budget. Our findings reveal that the best-performing model achieves competitive results in both micro and macro F1 scores, with label attention contributing significantly to its success. The study also explores the possibility of training on unlabeled data. Our research provides valuable insights into the possibilities of using NLP for clinical coding in low-resource languages, demonstrating that small countries with unique languages and well-segmented healthcare records can achieve results comparable to those in higher-resourced languages. | [
"Hauksson, Haraldur Orri",
"Einarsson, Hafsteinn"
] | Applications of BERT Models Towards Automation of Clinical Coding in Icelandic | findings-naacl.127 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.128.bib | https://aclanthology.org/2024.findings-naacl.128/ | @inproceedings{heinisch-etal-2024-tell,
title = "{``}Tell me who you are and {I} tell you how you argue{''}: Predicting Stances and Arguments for Stakeholder Groups",
author = "Heinisch, Philipp and
Dumani, Lorik and
Cimiano, Philipp and
Schenkel, Ralf",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.128",
doi = "10.18653/v1/2024.findings-naacl.128",
pages = "1968--1982",
abstract = "Argument mining has focused so far mainly on the identification, extraction, and formalization of arguments. An important yet unaddressedtask consists in the prediction of the argumentative behavior of stakeholders in a debate. Predicting the argumentative behavior in advance can support foreseeing issues in public policy making or help recognize potential disagreements early on and help to resolve them. In this paper, we consider the novel task of predicting the argumentative behavior of individual stakeholders. We present ARGENST, a framework that relies on a recommender-based architecture to predict the stance and the argumentative main point on a specific controversial topic for a given stakeholder, which is described in terms of a profile including properties related to demographic attributes, religious and political orientation, socio-economic background, etc. We evaluate our approach on the well-known debate.org dataset in terms of accuracy for predicting stance as well as in terms of similarity of the generated arguments to the ground truth arguments using BERTScore. As part of a case study, we show how juries of members representing different stakeholder groups and perspectives can be assembled to simulate the public opinion on a given topic.",
}
| Argument mining has focused so far mainly on the identification, extraction, and formalization of arguments. An important yet unaddressedtask consists in the prediction of the argumentative behavior of stakeholders in a debate. Predicting the argumentative behavior in advance can support foreseeing issues in public policy making or help recognize potential disagreements early on and help to resolve them. In this paper, we consider the novel task of predicting the argumentative behavior of individual stakeholders. We present ARGENST, a framework that relies on a recommender-based architecture to predict the stance and the argumentative main point on a specific controversial topic for a given stakeholder, which is described in terms of a profile including properties related to demographic attributes, religious and political orientation, socio-economic background, etc. We evaluate our approach on the well-known debate.org dataset in terms of accuracy for predicting stance as well as in terms of similarity of the generated arguments to the ground truth arguments using BERTScore. As part of a case study, we show how juries of members representing different stakeholder groups and perspectives can be assembled to simulate the public opinion on a given topic. | [
"Heinisch, Philipp",
"Dumani, Lorik",
"Cimiano, Philipp",
"Schenkel, Ralf"
] | “Tell me who you are and I tell you how you argue”: Predicting Stances and Arguments for Stakeholder Groups | findings-naacl.128 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.129.bib | https://aclanthology.org/2024.findings-naacl.129/ | @inproceedings{kuribayashi-etal-2024-psychometric,
title = "Psychometric Predictive Power of Large Language Models",
author = "Kuribayashi, Tatsuki and
Oseki, Yohei and
Baldwin, Timothy",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.129",
doi = "10.18653/v1/2024.findings-naacl.129",
pages = "1983--2005",
abstract = "Instruction tuning aligns the response of large language models (LLMs) with human preferences.Despite such efforts in human{--}LLM alignment, we find that instruction tuning does not always make LLMs human-like from a cognitive modeling perspective. More specifically, next-word probabilities estimated by instruction-tuned LLMs are often worse at simulating human reading behavior than those estimated by base LLMs.In addition, we explore prompting methodologies for simulating human reading behavior with LLMs. Our results show that prompts reflecting a particular linguistic hypothesis improve psychometric predictive power, but are still inferior to small base models.These findings highlight that recent advancements in LLMs, i.e., instruction tuning and prompting, do not offer better estimates than direct probability measurements from base LLMs in cognitive modeling. In other words, pure next-word probability remains a strong predictor for human reading behavior, even in the age of LLMs.",
}
| Instruction tuning aligns the response of large language models (LLMs) with human preferences.Despite such efforts in human{--}LLM alignment, we find that instruction tuning does not always make LLMs human-like from a cognitive modeling perspective. More specifically, next-word probabilities estimated by instruction-tuned LLMs are often worse at simulating human reading behavior than those estimated by base LLMs.In addition, we explore prompting methodologies for simulating human reading behavior with LLMs. Our results show that prompts reflecting a particular linguistic hypothesis improve psychometric predictive power, but are still inferior to small base models.These findings highlight that recent advancements in LLMs, i.e., instruction tuning and prompting, do not offer better estimates than direct probability measurements from base LLMs in cognitive modeling. In other words, pure next-word probability remains a strong predictor for human reading behavior, even in the age of LLMs. | [
"Kuribayashi, Tatsuki",
"Oseki, Yohei",
"Baldwin, Timothy"
] | Psychometric Predictive Power of Large Language Models | findings-naacl.129 | Poster | 2311.07484 | [
"https://github.com/kuribayashi4/llm-cognitive-modeling"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.130.bib | https://aclanthology.org/2024.findings-naacl.130/ | @inproceedings{pezeshkpour-hruschka-2024-large,
title = "Large Language Models Sensitivity to The Order of Options in Multiple-Choice Questions",
author = "Pezeshkpour, Pouya and
Hruschka, Estevam",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.130",
doi = "10.18653/v1/2024.findings-naacl.130",
pages = "2006--2017",
abstract = "Large Language Models (LLMs) have demonstrated remarkable capabilities in various NLP tasks. However, previous works have shown these models are sensitive towards prompt wording, and few-shot demonstrations and their order, posing challenges to fair assessment of these models. As these models become more powerful, it becomes imperative to understand and address these limitations. In this paper, we focus on LLMs robustness on the task of multiple-choice questions{---}commonly adopted task to study reasoning and fact-retrieving capability of LLMs. Investigating the sensitivity of LLMs towards the order of options in multiple-choice questions, we demonstrate a considerable performance gap of approximately 13{\%} to 85{\%} in LLMs on different benchmarks, when answer options are reordered, even when using demonstrations in a few-shot setting. Through a detailed analysis, we conjecture that this sensitivity arises when LLMs are uncertain about the prediction between the top-2/3 choices, and specific options placements may favor certain prediction between those top choices depending on the question caused by positional bias. We also identify patterns in top-2 choices that amplify or mitigate the model{'}s bias toward option placement. We found that for amplifying bias, the optimal strategy involves positioning the top two choices as the first and last options. Conversely, to mitigate bias, we recommend placing these choices among the adjacent options. To validate our conjecture, we conduct various experiments and adopt two approaches to calibrate LLMs{'} predictions, leading to up to 8 percentage points improvement across different models and benchmarks.",
}
| Large Language Models (LLMs) have demonstrated remarkable capabilities in various NLP tasks. However, previous works have shown these models are sensitive towards prompt wording, and few-shot demonstrations and their order, posing challenges to fair assessment of these models. As these models become more powerful, it becomes imperative to understand and address these limitations. In this paper, we focus on LLMs robustness on the task of multiple-choice questions{---}commonly adopted task to study reasoning and fact-retrieving capability of LLMs. Investigating the sensitivity of LLMs towards the order of options in multiple-choice questions, we demonstrate a considerable performance gap of approximately 13{\%} to 85{\%} in LLMs on different benchmarks, when answer options are reordered, even when using demonstrations in a few-shot setting. Through a detailed analysis, we conjecture that this sensitivity arises when LLMs are uncertain about the prediction between the top-2/3 choices, and specific options placements may favor certain prediction between those top choices depending on the question caused by positional bias. We also identify patterns in top-2 choices that amplify or mitigate the model{'}s bias toward option placement. We found that for amplifying bias, the optimal strategy involves positioning the top two choices as the first and last options. Conversely, to mitigate bias, we recommend placing these choices among the adjacent options. To validate our conjecture, we conduct various experiments and adopt two approaches to calibrate LLMs{'} predictions, leading to up to 8 percentage points improvement across different models and benchmarks. | [
"Pezeshkpour, Pouya",
"Hruschka, Estevam"
] | Large Language Models Sensitivity to The Order of Options in Multiple-Choice Questions | findings-naacl.130 | Poster | 2308.11483 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.131.bib | https://aclanthology.org/2024.findings-naacl.131/ | @inproceedings{pham-etal-2024-peeb,
title = "{PEEB}: Part-based Image Classifiers with an Explainable and Editable Language Bottleneck",
author = "Pham, Thang and
Chen, Peijie and
Nguyen, Tin and
Yoon, Seunghyun and
Bui, Trung and
Nguyen, Anh",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.131",
doi = "10.18653/v1/2024.findings-naacl.131",
pages = "2018--2053",
abstract = "CLIP-based classifiers rely on the prompt containing a class name that is known to the text encoder. Therefore, they perform poorly on new classes or the classes whose names rarely appear on the Internet (e.g., scientific names of birds). For fine-grained classification, we propose PEEB {--} an explainable and editable classifier to (1) express the class name into a set of text descriptors that describe the visual parts of that class; and (2) match the embeddings of the detected parts to their textual descriptors in each class to compute a logit score for classification. In a zero-shot setting where the class names are unknown, PEEB outperforms CLIP by a huge margin (â¼10{\mbox{$\times$}} in top-1 accuracy). Compared to part-based classifiers, PEEB is not only the state-of-the-art (SOTA) on the supervised-learning setting (88.80{\%} and 92.20{\%} accuracy on CUB-200 and Stanford Dogs-120, respectively) but also the first to enable users to edit the text descriptors to form a new classifier without any re-training. Compared to concept bottleneck models, PEEB is also the SOTA in both zero-shot and supervised-learning settings.",
}
| CLIP-based classifiers rely on the prompt containing a class name that is known to the text encoder. Therefore, they perform poorly on new classes or the classes whose names rarely appear on the Internet (e.g., scientific names of birds). For fine-grained classification, we propose PEEB {--} an explainable and editable classifier to (1) express the class name into a set of text descriptors that describe the visual parts of that class; and (2) match the embeddings of the detected parts to their textual descriptors in each class to compute a logit score for classification. In a zero-shot setting where the class names are unknown, PEEB outperforms CLIP by a huge margin (â¼10{\mbox{$\times$}} in top-1 accuracy). Compared to part-based classifiers, PEEB is not only the state-of-the-art (SOTA) on the supervised-learning setting (88.80{\%} and 92.20{\%} accuracy on CUB-200 and Stanford Dogs-120, respectively) but also the first to enable users to edit the text descriptors to form a new classifier without any re-training. Compared to concept bottleneck models, PEEB is also the SOTA in both zero-shot and supervised-learning settings. | [
"Pham, Thang",
"Chen, Peijie",
"Nguyen, Tin",
"Yoon, Seunghyun",
"Bui, Trung",
"Nguyen, Anh"
] | PEEB: Part-based Image Classifiers with an Explainable and Editable Language Bottleneck | findings-naacl.131 | Poster | 2403.05297 | [
"https://github.com/anguyen8/peeb"
] | https://huggingface.co/papers/2403.05297 | 1 | 0 | 0 | 6 | 1 | [] | [] | [
"XAI/PEEB"
] |
https://aclanthology.org/2024.findings-naacl.132.bib | https://aclanthology.org/2024.findings-naacl.132/ | @inproceedings{gao-etal-2024-ethos,
title = "Ethos: Rectifying Language Models in Orthogonal Parameter Space",
author = "Gao, Lei and
Niu, Yue and
Tang, Tingting and
Avestimehr, Salman and
Annavaram, Murali",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.132",
doi = "10.18653/v1/2024.findings-naacl.132",
pages = "2054--2068",
abstract = "Language models (LMs) have greatly propelled the research on natural language processing. However, LMs also raise concerns regarding the generation of biased or toxic content and the potential disclosure of private information from the training dataset. In this work, we present a new efficient approach, Ethos, that rectifies LMs to mitigate toxicity and bias in outputs and avoid privacy leakage. Ethos is built on task arithmetic. However, unlike current task arithmetic algorithms, Ethos distinguishes general beneficial and undesired knowledge when reconstructing task vectors. Specifically, Ethos first obtains a set of principal components from the pre-trained models using singular value decomposition. Then, by projecting the task vector onto principal components, Ethos separates the principal components that encode general from those associated with undesired knowledge. Ethos performs forgetting or unlearning by only negating the task vector with undesired knowledge, thereby minimizing collateral damage on general model utility. We demonstrate the efficacy of our approach on three different tasks: bias, toxicity, and memorization unlearning. Evaluations show Ethos is more effective in removing undesired knowledge while maintaining the overall model performance compared to current task arithmetic methods.",
}
| Language models (LMs) have greatly propelled the research on natural language processing. However, LMs also raise concerns regarding the generation of biased or toxic content and the potential disclosure of private information from the training dataset. In this work, we present a new efficient approach, Ethos, that rectifies LMs to mitigate toxicity and bias in outputs and avoid privacy leakage. Ethos is built on task arithmetic. However, unlike current task arithmetic algorithms, Ethos distinguishes general beneficial and undesired knowledge when reconstructing task vectors. Specifically, Ethos first obtains a set of principal components from the pre-trained models using singular value decomposition. Then, by projecting the task vector onto principal components, Ethos separates the principal components that encode general from those associated with undesired knowledge. Ethos performs forgetting or unlearning by only negating the task vector with undesired knowledge, thereby minimizing collateral damage on general model utility. We demonstrate the efficacy of our approach on three different tasks: bias, toxicity, and memorization unlearning. Evaluations show Ethos is more effective in removing undesired knowledge while maintaining the overall model performance compared to current task arithmetic methods. | [
"Gao, Lei",
"Niu, Yue",
"Tang, Tingting",
"Avestimehr, Salman",
"Annavaram, Murali"
] | Ethos: Rectifying Language Models in Orthogonal Parameter Space | findings-naacl.132 | Poster | 2403.08994 | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.133.bib | https://aclanthology.org/2024.findings-naacl.133/ | @inproceedings{lee-etal-2024-crafting,
title = "Crafting In-context Examples according to {LM}s{'} Parametric Knowledge",
author = "Lee, Yoonsang and
Atreya, Pranav and
Ye, Xi and
Choi, Eunsol",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.133",
doi = "10.18653/v1/2024.findings-naacl.133",
pages = "2069--2085",
abstract = "In-context learning can improve the performances of knowledge-rich tasks such as question answering. In such scenarios, in-context examples trigger a language model (LM) to surface information stored in its parametric knowledge. We study how to better construct in-context example sets, based on whether the model is aware of the in-context examples. We identify {`}known{'} examples, where models can correctly answer from their parametric knowledge, and {`}unknown{'} ones. Our experiments show that prompting with {`}unknown{'} examples decreases the performance, potentially as it encourages hallucination rather than searching for its parametric knowledge. Constructing an in-context example set that presents both known and unknown information performs the best across diverse settings. We perform analysis on three multi-answer question answering datasets, which allows us to further study answer set ordering strategies based on the LM{'}s knowledge of each answer. Together, our study sheds light on how to best construct in-context example sets for knowledge-rich tasks.",
}
| In-context learning can improve the performances of knowledge-rich tasks such as question answering. In such scenarios, in-context examples trigger a language model (LM) to surface information stored in its parametric knowledge. We study how to better construct in-context example sets, based on whether the model is aware of the in-context examples. We identify {`}known{'} examples, where models can correctly answer from their parametric knowledge, and {`}unknown{'} ones. Our experiments show that prompting with {`}unknown{'} examples decreases the performance, potentially as it encourages hallucination rather than searching for its parametric knowledge. Constructing an in-context example set that presents both known and unknown information performs the best across diverse settings. We perform analysis on three multi-answer question answering datasets, which allows us to further study answer set ordering strategies based on the LM{'}s knowledge of each answer. Together, our study sheds light on how to best construct in-context example sets for knowledge-rich tasks. | [
"Lee, Yoonsang",
"Atreya, Pranav",
"Ye, Xi",
"Choi, Eunsol"
] | Crafting In-context Examples according to LMs' Parametric Knowledge | findings-naacl.133 | Poster | 2311.09579 | [
"https://github.com/lilys012/known_examples"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.134.bib | https://aclanthology.org/2024.findings-naacl.134/ | @inproceedings{zhu-zamani-2024-icxml,
title = "{ICXML}: An In-Context Learning Framework for Zero-Shot Extreme Multi-Label Classification",
author = "Zhu, Yaxin and
Zamani, Hamed",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.134",
doi = "10.18653/v1/2024.findings-naacl.134",
pages = "2086--2098",
abstract = "This paper focuses on the task of Extreme Multi-Label Classification (XMC) whose goal is to predict multiple labels for each instance from an extremely large label space. While existing research has primarily focused on fully supervised XMC, real-world scenarios often lack supervision signals, highlighting the importance of zero-shot settings. Given the large label space, utilizing in-context learning approaches is not trivial. We address this issue by introducing In-Context Extreme Multi-label Learning (ICXML), a two-stage framework that cuts down the search space by generating a set of candidate labels through in-context learning and then reranks them. Extensive experiments suggest that ICXML advances the state of the art on two diverse public benchmarks.",
}
| This paper focuses on the task of Extreme Multi-Label Classification (XMC) whose goal is to predict multiple labels for each instance from an extremely large label space. While existing research has primarily focused on fully supervised XMC, real-world scenarios often lack supervision signals, highlighting the importance of zero-shot settings. Given the large label space, utilizing in-context learning approaches is not trivial. We address this issue by introducing In-Context Extreme Multi-label Learning (ICXML), a two-stage framework that cuts down the search space by generating a set of candidate labels through in-context learning and then reranks them. Extensive experiments suggest that ICXML advances the state of the art on two diverse public benchmarks. | [
"Zhu, Yaxin",
"Zamani, Hamed"
] | ICXML: An In-Context Learning Framework for Zero-Shot Extreme Multi-Label Classification | findings-naacl.134 | Poster | 2311.09649 | [
"https://github.com/yaxinzhuars/icxml"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.135.bib | https://aclanthology.org/2024.findings-naacl.135/ | @inproceedings{yang-etal-2024-clgsi,
title = "{CLGSI}: A Multimodal Sentiment Analysis Framework based on Contrastive Learning Guided by Sentiment Intensity",
author = "Yang, Yang and
Dong, Xunde and
Qiang, Yupeng",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.135",
doi = "10.18653/v1/2024.findings-naacl.135",
pages = "2099--2110",
abstract = "Recently, contrastive learning has begun to gain popularity in multimodal sentiment analysis (MSA). However, most of existing MSA methods based on contrastive learning lacks more detailed learning of the distribution of sample pairs with different sentiment intensity differences in the contrastive learning representation space. In addition, limited research has been conducted on the fusion of each modality representation obtained by contrastive learning training.In this paper, we propose a novel framework for multimodal sentiment analysis based on Contrastive Learning Guided by Sentiment Intensity (CLGSI). Firstly, the proposed contrastive learning guided by sentiment intensity selects positive and negative sample pairs based on the difference in sentiment intensity and assigns corresponding weights accordingly.Subsequently, we propose a new multimodal representation fusion mechanism, called Global-Local-Fine-Knowledge (GLFK), which extracts common features between different modalities{'} representations. At the same time, each unimodal encoder output is separately processed by a Multilayer Perceptron (MLP) to extract specific features of each modality. Finally, joint learning of the common and specific features is used to predict sentiment intensity. The effectiveness of CLGSI is assessed on two English datasets, MOSI and MOSEI, as well as one Chinese dataset, SIMS. We achieve competitive experimental results, which attest to the strong generalization performance of our approach. The code for our approach will be released in https://github.com/AZYoung233/CLGSI",
}
| Recently, contrastive learning has begun to gain popularity in multimodal sentiment analysis (MSA). However, most of existing MSA methods based on contrastive learning lacks more detailed learning of the distribution of sample pairs with different sentiment intensity differences in the contrastive learning representation space. In addition, limited research has been conducted on the fusion of each modality representation obtained by contrastive learning training.In this paper, we propose a novel framework for multimodal sentiment analysis based on Contrastive Learning Guided by Sentiment Intensity (CLGSI). Firstly, the proposed contrastive learning guided by sentiment intensity selects positive and negative sample pairs based on the difference in sentiment intensity and assigns corresponding weights accordingly.Subsequently, we propose a new multimodal representation fusion mechanism, called Global-Local-Fine-Knowledge (GLFK), which extracts common features between different modalities{'} representations. At the same time, each unimodal encoder output is separately processed by a Multilayer Perceptron (MLP) to extract specific features of each modality. Finally, joint learning of the common and specific features is used to predict sentiment intensity. The effectiveness of CLGSI is assessed on two English datasets, MOSI and MOSEI, as well as one Chinese dataset, SIMS. We achieve competitive experimental results, which attest to the strong generalization performance of our approach. The code for our approach will be released in https://github.com/AZYoung233/CLGSI | [
"Yang, Yang",
"Dong, Xunde",
"Qiang, Yupeng"
] | CLGSI: A Multimodal Sentiment Analysis Framework based on Contrastive Learning Guided by Sentiment Intensity | findings-naacl.135 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.136.bib | https://aclanthology.org/2024.findings-naacl.136/ | @inproceedings{wang-etal-2024-interpreting,
title = "Interpreting Answers to Yes-No Questions in Dialogues from Multiple Domains",
author = "Wang, Zijie and
Rashid, Farzana and
Blanco, Eduardo",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.136",
doi = "10.18653/v1/2024.findings-naacl.136",
pages = "2111--2128",
abstract = "People often answer yes-no questions without explicitly saying yes, no, or similar polar key-words. Figuring out the meaning of indirectanswers is challenging, even for large language models. In this paper, we investigate this problem working with dialogues from multiple domains. We present new benchmarks in three diverse domains: movie scripts, tennis interviews, and airline customer service. We present an approach grounded on distant supervision and blended training to quickly adapt to a new dialogue domain. Experimental results show that our approach is never detrimental and yields F1 improvements as high as 11-34{\%}.",
}
| People often answer yes-no questions without explicitly saying yes, no, or similar polar key-words. Figuring out the meaning of indirectanswers is challenging, even for large language models. In this paper, we investigate this problem working with dialogues from multiple domains. We present new benchmarks in three diverse domains: movie scripts, tennis interviews, and airline customer service. We present an approach grounded on distant supervision and blended training to quickly adapt to a new dialogue domain. Experimental results show that our approach is never detrimental and yields F1 improvements as high as 11-34{\%}. | [
"Wang, Zijie",
"Rashid, Farzana",
"Blanco, Eduardo"
] | Interpreting Answers to Yes-No Questions in Dialogues from Multiple Domains | findings-naacl.136 | Poster | 2404.16262 | [
"https://github.com/wang-zijie/yn-question-multi-domains"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
|
https://aclanthology.org/2024.findings-naacl.137.bib | https://aclanthology.org/2024.findings-naacl.137/ | @inproceedings{hsu-etal-2024-enhancing,
title = "Enhancing Perception: Refining Explanations of News Claims with {LLM} Conversations",
author = "Hsu, Yi-Li and
Chen, Jui-Ning and
Fan Chiang, Yang and
Liu, Shang-Chien and
Xiong, Aiping and
Ku, Lun-Wei",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.137",
doi = "10.18653/v1/2024.findings-naacl.137",
pages = "2129--2147",
abstract = "We introduce Enhancing Perception, a framework for Large Language Models (LLMs) designed to streamline the time-intensive task typically undertaken by professional fact-checkers of crafting explanations for fake news. This study investigates the effectiveness of enhancing LLM explanations through conversational refinement. We compare various questioner agents, including state-of-the-art LLMs like GPT-4, Claude 2, PaLM 2, and 193 American participants acting as human questioners. Based on the histories of these refinement conversations, we further generate comprehensive summary explanations. We evaluated the effectiveness of these initial, refined, and summary explanations across 40 news claims by involving 2,797 American participants, measuring their self-reported belief change regarding both real and fake claims after receiving the explanations. Our findings reveal that, in the context of fake news, explanations that have undergone conversational refinement{---}whether by GPT-4 or human questioners, who ask more diverse and detail-oriented questions{---}were significantly more effective than both the initial unrefined explanations and the summary explanations. Moreover, these refined explanations achieved a level of effectiveness comparable to that of expert-written explanations. The results highlight the potential of automatic explanation refinement by LLMs in debunking fake news claims.",
}
| We introduce Enhancing Perception, a framework for Large Language Models (LLMs) designed to streamline the time-intensive task typically undertaken by professional fact-checkers of crafting explanations for fake news. This study investigates the effectiveness of enhancing LLM explanations through conversational refinement. We compare various questioner agents, including state-of-the-art LLMs like GPT-4, Claude 2, PaLM 2, and 193 American participants acting as human questioners. Based on the histories of these refinement conversations, we further generate comprehensive summary explanations. We evaluated the effectiveness of these initial, refined, and summary explanations across 40 news claims by involving 2,797 American participants, measuring their self-reported belief change regarding both real and fake claims after receiving the explanations. Our findings reveal that, in the context of fake news, explanations that have undergone conversational refinement{---}whether by GPT-4 or human questioners, who ask more diverse and detail-oriented questions{---}were significantly more effective than both the initial unrefined explanations and the summary explanations. Moreover, these refined explanations achieved a level of effectiveness comparable to that of expert-written explanations. The results highlight the potential of automatic explanation refinement by LLMs in debunking fake news claims. | [
"Hsu, Yi-Li",
"Chen, Jui-Ning",
"Fan Chiang, Yang",
"Liu, Shang-Chien",
"Xiong, Aiping",
"Ku, Lun-Wei"
] | Enhancing Perception: Refining Explanations of News Claims with LLM Conversations | findings-naacl.137 | Poster | [
""
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
||
https://aclanthology.org/2024.findings-naacl.138.bib | https://aclanthology.org/2024.findings-naacl.138/ | @inproceedings{wei-jie-etal-2024-interpretable,
title = "How Interpretable are Reasoning Explanations from Prompting Large Language Models?",
author = "Wei Jie, Yeo and
Satapathy, Ranjan and
Goh, Rick and
Cambria, Erik",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2024",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-naacl.138",
doi = "10.18653/v1/2024.findings-naacl.138",
pages = "2148--2164",
abstract = "Prompt Engineering has garnered significant attention for enhancing the performance of large language models across a multitude of tasks. Techniques such as the Chain-of-Thought not only bolster task performance but also delineate a clear trajectory of reasoning steps, offering a tangible form of explanation for the audience. Prior works on interpretability assess the reasoning chains yielded by Chain-of-Thought solely along a singular axis, namely faithfulness. We present a comprehensive and multifaceted evaluation of interpretability, examining not only faithfulness but also robustness and utility across multiple commonsense reasoning benchmarks. Likewise, our investigation is not confined to a single prompting technique; it expansively covers a multitude of prevalent prompting techniques employed in large language models, thereby ensuring a wide-ranging and exhaustive evaluation. In addition, we introduce a simple interpretability alignment technique, termed Self-Entailment-Alignment Chain-of-thought, that yields more than 70{\%} improvements across multiple dimensions of interpretability. Code is available at https://github.com/SenticNet/CoT{\_}interpretability",
}
| Prompt Engineering has garnered significant attention for enhancing the performance of large language models across a multitude of tasks. Techniques such as the Chain-of-Thought not only bolster task performance but also delineate a clear trajectory of reasoning steps, offering a tangible form of explanation for the audience. Prior works on interpretability assess the reasoning chains yielded by Chain-of-Thought solely along a singular axis, namely faithfulness. We present a comprehensive and multifaceted evaluation of interpretability, examining not only faithfulness but also robustness and utility across multiple commonsense reasoning benchmarks. Likewise, our investigation is not confined to a single prompting technique; it expansively covers a multitude of prevalent prompting techniques employed in large language models, thereby ensuring a wide-ranging and exhaustive evaluation. In addition, we introduce a simple interpretability alignment technique, termed Self-Entailment-Alignment Chain-of-thought, that yields more than 70{\%} improvements across multiple dimensions of interpretability. Code is available at https://github.com/SenticNet/CoT{\_}interpretability | [
"Wei Jie, Yeo",
"Satapathy, Ranjan",
"Goh, Rick",
"Cambria, Erik"
] | How Interpretable are Reasoning Explanations from Prompting Large Language Models? | findings-naacl.138 | Poster | 2402.11863 | [
"https://github.com/wj210/cot_interpretability"
] | -1 | -1 | -1 | -1 | 0 | [] | [] | [] |
Subsets and Splits