bibtex_url
stringlengths 41
50
| bibtext
stringlengths 693
2.88k
| abstract
stringlengths 0
2k
| authors
listlengths 1
45
| title
stringlengths 21
206
| id
stringlengths 7
16
| type
stringclasses 2
values | arxiv_id
stringlengths 9
12
⌀ |
|---|---|---|---|---|---|---|---|
https://aclanthology.org/2024.findings-acl.168.bib | @inproceedings{nguyen-etal-2024-direct,
title = "Direct Evaluation of Chain-of-Thought in Multi-hop Reasoning with Knowledge Graphs",
author = "Nguyen, Thi and
Luo, Linhao and
Shiri, Fatemeh and
Phung, Dinh and
Li, Yuan-Fang and
Vu, Thuy-Trang and
Haffari, Gholamreza",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.168",
pages = "2862--2883",
abstract = "Large language models (LLMs) have demonstrated strong reasoning abilities when prompted to generate chain-of-thought (CoT) explanations alongside answers. However, previous research on evaluating LLMs has solely focused on answer accuracy, neglecting the correctness of the generated CoT. In this paper, we delve deeper into the CoT reasoning capabilities of LLMs in multi-hop question answering by utilizing knowledge graphs (KGs). We propose a novel discriminative and generative CoT evaluation paradigm to assess LLMs{'} knowledge of reasoning and the accuracy of the generated CoT. Through experiments conducted on 5 different families of LLMs across 2 multi-hop question-answering datasets, we find that LLMs possess sufficient knowledge to perform reasoning. However, there exists a significant disparity between answer accuracy and faithfulness of the CoT generated by LLMs, indicating that they often arrive at correct answers through incorrect reasoning.",
}
| Large language models (LLMs) have demonstrated strong reasoning abilities when prompted to generate chain-of-thought (CoT) explanations alongside answers. However, previous research on evaluating LLMs has solely focused on answer accuracy, neglecting the correctness of the generated CoT. In this paper, we delve deeper into the CoT reasoning capabilities of LLMs in multi-hop question answering by utilizing knowledge graphs (KGs). We propose a novel discriminative and generative CoT evaluation paradigm to assess LLMs{'} knowledge of reasoning and the accuracy of the generated CoT. Through experiments conducted on 5 different families of LLMs across 2 multi-hop question-answering datasets, we find that LLMs possess sufficient knowledge to perform reasoning. However, there exists a significant disparity between answer accuracy and faithfulness of the CoT generated by LLMs, indicating that they often arrive at correct answers through incorrect reasoning. | [
"Nguyen, Thi",
"Luo, Linhao",
"Shiri, Fatemeh",
"Phung, Dinh",
"Li, Yuan-Fang",
"Vu, Thuy-Trang",
"Haffari, Gholamreza"
]
| Direct Evaluation of Chain-of-Thought in Multi-hop Reasoning with Knowledge Graphs | findings-acl.168 | Poster | 2108.08297v2 |
https://aclanthology.org/2024.findings-acl.169.bib | @inproceedings{zhang-etal-2024-comprehensive,
title = "Comprehensive Abstractive Comment Summarization with Dynamic Clustering and Chain of Thought",
author = "Zhang, Longyin and
Zou, Bowei and
Yi, Jacintha and
Aw, AiTi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.169",
pages = "2884--2896",
abstract = "Real-world news comments pose a significant challenge due to their noisy and ambiguous nature, which complicates their modeling for clustering and summarization tasks. Most previous research has predominantly focused on extractive summarization methods within specific constraints. This paper concentrates on Clustering and Abstractive Summarization of online news Comments (CASC). First, we introduce an enhanced fast clustering algorithm that maintains a dynamic similarity threshold to ensure the high density of each comment cluster being built. Moreover, we pioneer the exploration of tuning Large Language Models (LLMs) through a chain-of-thought strategy to generate summaries for each comment cluster. On the other hand, a notable challenge in CASC research is the scarcity of evaluation data. To address this problem, we design an annotation scheme and contribute a manual test suite tailored for CASC. Experimental results on the test suite demonstrate the effectiveness of our improvements to the baseline methods. In addition, the quantitative and qualitative analyses illustrate the adaptability of our approach to real-world news comment scenarios.",
}
| Real-world news comments pose a significant challenge due to their noisy and ambiguous nature, which complicates their modeling for clustering and summarization tasks. Most previous research has predominantly focused on extractive summarization methods within specific constraints. This paper concentrates on Clustering and Abstractive Summarization of online news Comments (CASC). First, we introduce an enhanced fast clustering algorithm that maintains a dynamic similarity threshold to ensure the high density of each comment cluster being built. Moreover, we pioneer the exploration of tuning Large Language Models (LLMs) through a chain-of-thought strategy to generate summaries for each comment cluster. On the other hand, a notable challenge in CASC research is the scarcity of evaluation data. To address this problem, we design an annotation scheme and contribute a manual test suite tailored for CASC. Experimental results on the test suite demonstrate the effectiveness of our improvements to the baseline methods. In addition, the quantitative and qualitative analyses illustrate the adaptability of our approach to real-world news comment scenarios. | [
"Zhang, Longyin",
"Zou, Bowei",
"Yi, Jacintha",
"Aw, AiTi"
]
| Comprehensive Abstractive Comment Summarization with Dynamic Clustering and Chain of Thought | findings-acl.169 | Poster | 2407.07959v1 |
https://aclanthology.org/2024.findings-acl.170.bib | @inproceedings{liu-etal-2024-self-supervised,
title = "Self-Supervised Position Debiasing for Large Language Models",
author = "Liu, Zhongkun and
Chen, Zheng and
Zhang, Mengqi and
Ren, Zhaochun and
Ren, Pengjie and
Chen, Zhumin",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.170",
pages = "2897--2917",
abstract = "Fine-tuning has been demonstrated to be an effective method to improve the domain performance of large language models (LLMs). However, LLMs might fit the dataset bias and shortcuts for prediction, leading to poor generation performance. Previous works have proven that LLMs are prone to exhibit position bias, i.e., leveraging information positioned at the beginning or end, or specific positional cues within the input. Existing debiasing methods for LLMs require external bias knowledge or annotated non-biased samples, which is lacking for position debiasing and impractical in reality. In this work, we propose a self-supervised position debiasing (SOD) framework to mitigate position bias for LLMs. SOD leverages unsupervised responses from pre-trained LLMs for debiasing without relying on any external knowledge. To improve the quality of unsupervised responses, we propose an objective alignment (OAM) module to prune these responses. Experiments on eight datasets and five tasks show that SOD consistently outperforms existing methods in mitigating three types of position biases. Besides, SOD achieves this by sacrificing only a small performance on biased samples, which is general and effective. To facilitate the reproducibility of the results, we share the code of all methods and datasets on https://github.com/LZKSKY/SOD.",
}
| Fine-tuning has been demonstrated to be an effective method to improve the domain performance of large language models (LLMs). However, LLMs might fit the dataset bias and shortcuts for prediction, leading to poor generation performance. Previous works have proven that LLMs are prone to exhibit position bias, i.e., leveraging information positioned at the beginning or end, or specific positional cues within the input. Existing debiasing methods for LLMs require external bias knowledge or annotated non-biased samples, which is lacking for position debiasing and impractical in reality. In this work, we propose a self-supervised position debiasing (SOD) framework to mitigate position bias for LLMs. SOD leverages unsupervised responses from pre-trained LLMs for debiasing without relying on any external knowledge. To improve the quality of unsupervised responses, we propose an objective alignment (OAM) module to prune these responses. Experiments on eight datasets and five tasks show that SOD consistently outperforms existing methods in mitigating three types of position biases. Besides, SOD achieves this by sacrificing only a small performance on biased samples, which is general and effective. To facilitate the reproducibility of the results, we share the code of all methods and datasets on https://github.com/LZKSKY/SOD. | [
"Liu, Zhongkun",
"Chen, Zheng",
"Zhang, Mengqi",
"Ren, Zhaochun",
"Ren, Pengjie",
"Chen, Zhumin"
]
| Self-Supervised Position Debiasing for Large Language Models | findings-acl.170 | Poster | 2402.11764v1 |
https://aclanthology.org/2024.findings-acl.171.bib | @inproceedings{lu-etal-2024-hypercl,
title = "{H}yper{CL}: A Contrastive Learning Framework for Hyper-Relational Knowledge Graph Embedding with Hierarchical Ontology",
author = "Lu, Yuhuan and
Yu, Weijian and
Jing, Xin and
Yang, Dingqi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.171",
pages = "2918--2929",
abstract = "Knowledge Graph (KG) embeddings are essential for link prediction over KGs. Compared to triplets, hyper-relational facts consisting of a base triplet and an arbitrary number of key-value pairs, can better characterize real-world facts and have aroused various hyper-relational embedding techniques recently. Nevertheless, existing works seldom consider the ontology of KGs, which is beneficial to link prediction tasks. A few studies attempt to incorporate the ontology information, by either utilizing the ontology as constraints on entity representations or jointly learning from hyper-relational facts and the ontology. However, existing approaches mostly overlook the ontology hierarchy and suffer from the dominance issue of facts over ontology, resulting in suboptimal performance. Against this background, we propose a universal contrastive learning framework for hyper-relational KG embeddings ($\textbf{HyperCL}$), which is flexible to integrate different hyper-relational KG embedding methods and effectively boost their link prediction performance. HyperCL designs relation-aware Graph Attention Networks to capture the hierarchical ontology and a concept-aware contrastive loss to alleviate the dominance issue. We evaluate HyperCL on three real-world datasets in different link prediction tasks. Experimental results show that HyperCL consistently boosts the performance of state-of-the-art baselines with an average improvement of 3.1-7.4{\%} across the three datasets.",
}
| Knowledge Graph (KG) embeddings are essential for link prediction over KGs. Compared to triplets, hyper-relational facts consisting of a base triplet and an arbitrary number of key-value pairs, can better characterize real-world facts and have aroused various hyper-relational embedding techniques recently. Nevertheless, existing works seldom consider the ontology of KGs, which is beneficial to link prediction tasks. A few studies attempt to incorporate the ontology information, by either utilizing the ontology as constraints on entity representations or jointly learning from hyper-relational facts and the ontology. However, existing approaches mostly overlook the ontology hierarchy and suffer from the dominance issue of facts over ontology, resulting in suboptimal performance. Against this background, we propose a universal contrastive learning framework for hyper-relational KG embeddings ($\textbf{HyperCL}$), which is flexible to integrate different hyper-relational KG embedding methods and effectively boost their link prediction performance. HyperCL designs relation-aware Graph Attention Networks to capture the hierarchical ontology and a concept-aware contrastive loss to alleviate the dominance issue. We evaluate HyperCL on three real-world datasets in different link prediction tasks. Experimental results show that HyperCL consistently boosts the performance of state-of-the-art baselines with an average improvement of 3.1-7.4{\%} across the three datasets. | [
"Lu, Yuhuan",
"Yu, Weijian",
"Jing, Xin",
"Yang, Dingqi"
]
| {H}yper{CL}: A Contrastive Learning Framework for Hyper-Relational Knowledge Graph Embedding with Hierarchical Ontology | findings-acl.171 | Poster | 1809.02382v1 |
https://aclanthology.org/2024.findings-acl.172.bib | @inproceedings{liu-etal-2024-encoding,
title = "Encoding Hierarchical Schema via Concept Flow for Multifaceted Ideology Detection",
author = "Liu, Songtao and
Wang, Bang and
Xiang, Wei and
Xu, Han and
Xu, Minghua",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.172",
pages = "2930--2942",
abstract = "Multifaceted ideology detection (MID) aims to detect the ideological leanings of texts towards multiple facets. Previous studies on ideology detection mainly focus on one generic facet and ignore label semantics and explanatory descriptions of ideologies, which are a kind of instructive information and reveal the specific concepts of ideologies. In this paper, we develop a novel concept semantics-enhanced framework for the MID task. Specifically, we propose a bidirectional iterative concept flow (BICo) method to encode multifaceted ideologies. BICo enables the concepts to flow across levels of the schema tree and enriches concept representations with multi-granularity semantics. Furthermore, we explore concept attentive matching and concept-guided contrastive learning strategies to guide the model to capture ideology features with the learned concept semantics. Extensive experiments on the benchmark dataset show that our approach achieves state-of-the-art performance in MID, including in the cross-topic scenario.",
}
| Multifaceted ideology detection (MID) aims to detect the ideological leanings of texts towards multiple facets. Previous studies on ideology detection mainly focus on one generic facet and ignore label semantics and explanatory descriptions of ideologies, which are a kind of instructive information and reveal the specific concepts of ideologies. In this paper, we develop a novel concept semantics-enhanced framework for the MID task. Specifically, we propose a bidirectional iterative concept flow (BICo) method to encode multifaceted ideologies. BICo enables the concepts to flow across levels of the schema tree and enriches concept representations with multi-granularity semantics. Furthermore, we explore concept attentive matching and concept-guided contrastive learning strategies to guide the model to capture ideology features with the learned concept semantics. Extensive experiments on the benchmark dataset show that our approach achieves state-of-the-art performance in MID, including in the cross-topic scenario. | [
"Liu, Songtao",
"Wang, Bang",
"Xiang, Wei",
"Xu, Han",
"Xu, Minghua"
]
| Encoding Hierarchical Schema via Concept Flow for Multifaceted Ideology Detection | findings-acl.172 | Poster | 2405.18974v1 |
https://aclanthology.org/2024.findings-acl.173.bib | @inproceedings{hou-li-2024-character,
title = "Character-Level {C}hinese Dependency Parsing via Modeling Latent Intra-Word Structure",
author = "Hou, Yang and
Li, Zhenghua",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.173",
pages = "2943--2956",
abstract = "Revealing the syntactic structure of sentences in Chinese poses significant challenges for word-level parsers due to the absence of clear word boundaries. To facilitate a transition from word-level to character-level Chinese dependency parsing, this paper proposes modeling latent internal structures within words. In this way, each word-level dependency tree is interpreted as a forest of character-level trees. A constrained Eisner algorithm is implemented to ensure the compatibility of character-level trees, guaranteeing a single root for intra-word structures and establishing inter-word dependencies between these roots. Experiments on Chinese treebanks demonstrate the superiority of our method over both the pipeline framework and previous joint models. A detailed analysis reveals that a coarse-to-fine parsing strategy empowers the model to predict more linguistically plausible intra-word structures.",
}
| Revealing the syntactic structure of sentences in Chinese poses significant challenges for word-level parsers due to the absence of clear word boundaries. To facilitate a transition from word-level to character-level Chinese dependency parsing, this paper proposes modeling latent internal structures within words. In this way, each word-level dependency tree is interpreted as a forest of character-level trees. A constrained Eisner algorithm is implemented to ensure the compatibility of character-level trees, guaranteeing a single root for intra-word structures and establishing inter-word dependencies between these roots. Experiments on Chinese treebanks demonstrate the superiority of our method over both the pipeline framework and previous joint models. A detailed analysis reveals that a coarse-to-fine parsing strategy empowers the model to predict more linguistically plausible intra-word structures. | [
"Hou, Yang",
"Li, Zhenghua"
]
| Character-Level {C}hinese Dependency Parsing via Modeling Latent Intra-Word Structure | findings-acl.173 | Poster | 2406.03772v1 |
https://aclanthology.org/2024.findings-acl.174.bib | @inproceedings{li-etal-2024-alignre,
title = "{A}lign{RE}: An Encoding and Semantic Alignment Approach for Zero-Shot Relation Extraction",
author = "Li, Zehan and
Zhang, Fu and
Cheng, Jingwei",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.174",
pages = "2957--2966",
abstract = "Zero-shot Relation Extraction (ZSRE) aims to predict unseen relations between entity pairs from input sentences. Existing prototype-based ZSRE methods encode relation descriptions into prototype embeddings and predict by measuring the similarity between sentence embeddings and prototype embeddings. However, these methods often overlook abundant side information of relations and suffer from a significant encoding gap between prototypes and sentences, limiting performance. To this end, we propose a framework named AlignRE, based on two Alignment methods for ZSRE. Specifically, we present a novel perspective centered on encoding schema alignment to enhance prototype-based ZSRE methods. We utilize well-designed prompt-tuning to bridge the encoding gap. To improve prototype quality, we explore and leverage multiple side information and propose a prototype aggregation method based on semantic alignment to create comprehensive relation prototype representations. We conduct experiments on FewRel and Wiki-ZSL datasets and consistently outperform state-of-the-art methods. Moreover, our method exhibits substantially faster performance and reduces the need for extensive manual labor in prototype construction. Code is available at https://github.com/lizehan1999/AlignRE.",
}
| Zero-shot Relation Extraction (ZSRE) aims to predict unseen relations between entity pairs from input sentences. Existing prototype-based ZSRE methods encode relation descriptions into prototype embeddings and predict by measuring the similarity between sentence embeddings and prototype embeddings. However, these methods often overlook abundant side information of relations and suffer from a significant encoding gap between prototypes and sentences, limiting performance. To this end, we propose a framework named AlignRE, based on two Alignment methods for ZSRE. Specifically, we present a novel perspective centered on encoding schema alignment to enhance prototype-based ZSRE methods. We utilize well-designed prompt-tuning to bridge the encoding gap. To improve prototype quality, we explore and leverage multiple side information and propose a prototype aggregation method based on semantic alignment to create comprehensive relation prototype representations. We conduct experiments on FewRel and Wiki-ZSL datasets and consistently outperform state-of-the-art methods. Moreover, our method exhibits substantially faster performance and reduces the need for extensive manual labor in prototype construction. Code is available at https://github.com/lizehan1999/AlignRE. | [
"Li, Zehan",
"Zhang, Fu",
"Cheng, Jingwei"
]
| {A}lign{RE}: An Encoding and Semantic Alignment Approach for Zero-Shot Relation Extraction | findings-acl.174 | Poster | 2009.11226v1 |
https://aclanthology.org/2024.findings-acl.175.bib | @inproceedings{fu-etal-2024-disperse,
title = "Disperse-Then-Merge: Pushing the Limits of Instruction Tuning via Alignment Tax Reduction",
author = "Fu, Tingchen and
Cai, Deng and
Liu, Lemao and
Shi, Shuming and
Yan, Rui",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.175",
pages = "2967--2985",
abstract = "Supervised fine-tuning (SFT) on instruction-following corpus is a crucial approach toward the alignment of large language models (LLMs). However, the performance of LLMs on standard knowledge and reasoning benchmarks tends to suffer from deterioration at the latter stage of the SFT process, echoing the phenomenon of alignment tax. Through our pilot study, we put a hypothesis that the data biases are probably one cause behind the phenomenon. To address the issue, we introduce a simple disperse-then-merge framework. To be concrete, we disperse the instruction-following data into portions and then train multiple sub-models using different data portions. Lastly, we merge multiple models into a single one via model merging techniques. Despite its simplicity, our framework outperforms various sophisticated methods such as data curation and training regularization on a series of standard knowledge and reasoning benchmarks.",
}
| Supervised fine-tuning (SFT) on instruction-following corpus is a crucial approach toward the alignment of large language models (LLMs). However, the performance of LLMs on standard knowledge and reasoning benchmarks tends to suffer from deterioration at the latter stage of the SFT process, echoing the phenomenon of alignment tax. Through our pilot study, we put a hypothesis that the data biases are probably one cause behind the phenomenon. To address the issue, we introduce a simple disperse-then-merge framework. To be concrete, we disperse the instruction-following data into portions and then train multiple sub-models using different data portions. Lastly, we merge multiple models into a single one via model merging techniques. Despite its simplicity, our framework outperforms various sophisticated methods such as data curation and training regularization on a series of standard knowledge and reasoning benchmarks. | [
"Fu, Tingchen",
"Cai, Deng",
"Liu, Lemao",
"Shi, Shuming",
"Yan, Rui"
]
| Disperse-Then-Merge: Pushing the Limits of Instruction Tuning via Alignment Tax Reduction | findings-acl.175 | Poster | 2405.13432v1 |
https://aclanthology.org/2024.findings-acl.176.bib | @inproceedings{jiang-etal-2024-efficient,
title = "Efficient Knowledge Infusion via {KG}-{LLM} Alignment",
author = "Jiang, Zhouyu and
Zhong, Ling and
Sun, Mengshu and
Xu, Jun and
Sun, Rui and
Cai, Hui and
Luo, Shuhan and
Zhang, Zhiqiang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.176",
pages = "2986--2999",
abstract = "To tackle the problem of domain-specific knowledge scarcity within large language models (LLMs), knowledge graph-retrievalaugmented method has been proven to be an effective and efficient technique for knowledge infusion. However, existing approaches face two primary challenges: knowledge mismatch between public available knowledge graphs and the specific domain of the task at hand, and poor information compliance of LLMs with knowledge graphs. In this paper, we leverage a small set of labeled samples and a large-scale corpus to efficiently construct domain-specific knowledge graphs by an LLM, addressing the issue of knowledge mismatch. Additionally, we propose a three-stage KG-LLM alignment strategy to enhance the LLM{'}s capability to utilize information from knowledge graphs. We conduct experiments with a limited-sample setting on two biomedical question-answering datasets, and the results demonstrate that our approach outperforms existing baselines.",
}
| To tackle the problem of domain-specific knowledge scarcity within large language models (LLMs), knowledge graph-retrievalaugmented method has been proven to be an effective and efficient technique for knowledge infusion. However, existing approaches face two primary challenges: knowledge mismatch between public available knowledge graphs and the specific domain of the task at hand, and poor information compliance of LLMs with knowledge graphs. In this paper, we leverage a small set of labeled samples and a large-scale corpus to efficiently construct domain-specific knowledge graphs by an LLM, addressing the issue of knowledge mismatch. Additionally, we propose a three-stage KG-LLM alignment strategy to enhance the LLM{'}s capability to utilize information from knowledge graphs. We conduct experiments with a limited-sample setting on two biomedical question-answering datasets, and the results demonstrate that our approach outperforms existing baselines. | [
"Jiang, Zhouyu",
"Zhong, Ling",
"Sun, Mengshu",
"Xu, Jun",
"Sun, Rui",
"Cai, Hui",
"Luo, Shuhan",
"Zhang, Zhiqiang"
]
| Efficient Knowledge Infusion via {KG}-{LLM} Alignment | findings-acl.176 | Poster | 2406.03746v1 |
https://aclanthology.org/2024.findings-acl.177.bib | @inproceedings{jung-etal-2024-towards,
title = "Towards Precise Localization of Critical Errors in Machine Translation",
author = "Jung, Dahyun and
Eo, Sugyeong and
Lim, Heuiseok",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.177",
pages = "3000--3012",
abstract = "The advent of large language models has experienced a remarkable improvement in the field of machine translation. However, machine translation is still vulnerable to critical meaning deviations, which may incur catastrophic issues in social or ethical contexts. In particular, existing critical error detection primarily focuses on identifying sentence-level errors, leaving the precise localization of such errors within the sentence unaddressed. In this paper, we introduce a new task, word-level critical error detection (WCED), to detect critical errors at a fine-grained level in machine translation sentences. The task aims to identify the parts of a machine translation that contain catastrophic meaning distortions. We hypothesize that the ability to determine errors at the sentence level will positively influence the detection of more granular errors. We propose a sentence-level error detection module to predict which words in a sentence have critical errors. Experimental results demonstrate that our method outperforms existing methodologies and LLM in En-De, Zh-En, En-Ru, and En-Ko. Our method is helpful for determining the fine-grained location of errors. We hope that such studies will improve the capacity to address critical errors adeptly.",
}
| The advent of large language models has experienced a remarkable improvement in the field of machine translation. However, machine translation is still vulnerable to critical meaning deviations, which may incur catastrophic issues in social or ethical contexts. In particular, existing critical error detection primarily focuses on identifying sentence-level errors, leaving the precise localization of such errors within the sentence unaddressed. In this paper, we introduce a new task, word-level critical error detection (WCED), to detect critical errors at a fine-grained level in machine translation sentences. The task aims to identify the parts of a machine translation that contain catastrophic meaning distortions. We hypothesize that the ability to determine errors at the sentence level will positively influence the detection of more granular errors. We propose a sentence-level error detection module to predict which words in a sentence have critical errors. Experimental results demonstrate that our method outperforms existing methodologies and LLM in En-De, Zh-En, En-Ru, and En-Ko. Our method is helpful for determining the fine-grained location of errors. We hope that such studies will improve the capacity to address critical errors adeptly. | [
"Jung, Dahyun",
"Eo, Sugyeong",
"Lim, Heuiseok"
]
| Towards Precise Localization of Critical Errors in Machine Translation | findings-acl.177 | Poster | 2205.09988v1 |
https://aclanthology.org/2024.findings-acl.178.bib | @inproceedings{zhang-etal-2024-loraprune,
title = "{L}o{RAP}rune: Structured Pruning Meets Low-Rank Parameter-Efficient Fine-Tuning",
author = "Zhang, Mingyang and
Chen, Hao and
Shen, Chunhua and
Yang, Zhen and
Ou, Linlin and
Yu, Xinyi and
Zhuang, Bohan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.178",
pages = "3013--3026",
abstract = "Large Language Models (LLMs), such as LLaMA and T5, have shown exceptional performance across various tasks through fine-tuning. Although low-rank adaption (LoRA) has emerged to cheaply fine-tune these LLMs on downstream tasks, their deployment is still hindered by the vast model scale and computational costs. Post-training model pruning offers a way to compress LLMs. However, the current pruning methods designed for LLMs are not compatible with LoRA. This is due to their utilization of unstructured pruning on LLMs, impeding the merging of LoRA weights, or their dependence on the gradients of pre-trained weights to guide pruning, which can impose significant memory overhead.To this end, we propose LoRAPrune, a new framework that delivers an accurate structured pruned model in a highly memory-efficient manner. Specifically, we first design a LoRA-guided pruning criterion, which uses the weights and gradients of LoRA, rather than the gradients of pre-trained weights for importance estimation. We subsequently integrate this criterion into an iterative pruning process, effectively removing redundant channels and heads. Extensive experimental results demonstrate the superior performance of our LoRAPrune over existing approaches on the LLaMA series models.At a 50{\%} compression rate, LoRAPrune demonstrates superior performance over LLM-Pruner, achieving a reduction in perplexity by 4.81 on WikiText2 and 3.46 on PTB, while also decreasing memory usage by 52.6{\%}.Besides, LoRAPrune also matches semi-structural pruning across multiple LLMs, proving its wide applicability. The code is available at https://github.com/aim-uofa/LoRAPrune.",
}
| Large Language Models (LLMs), such as LLaMA and T5, have shown exceptional performance across various tasks through fine-tuning. Although low-rank adaption (LoRA) has emerged to cheaply fine-tune these LLMs on downstream tasks, their deployment is still hindered by the vast model scale and computational costs. Post-training model pruning offers a way to compress LLMs. However, the current pruning methods designed for LLMs are not compatible with LoRA. This is due to their utilization of unstructured pruning on LLMs, impeding the merging of LoRA weights, or their dependence on the gradients of pre-trained weights to guide pruning, which can impose significant memory overhead.To this end, we propose LoRAPrune, a new framework that delivers an accurate structured pruned model in a highly memory-efficient manner. Specifically, we first design a LoRA-guided pruning criterion, which uses the weights and gradients of LoRA, rather than the gradients of pre-trained weights for importance estimation. We subsequently integrate this criterion into an iterative pruning process, effectively removing redundant channels and heads. Extensive experimental results demonstrate the superior performance of our LoRAPrune over existing approaches on the LLaMA series models.At a 50{\%} compression rate, LoRAPrune demonstrates superior performance over LLM-Pruner, achieving a reduction in perplexity by 4.81 on WikiText2 and 3.46 on PTB, while also decreasing memory usage by 52.6{\%}.Besides, LoRAPrune also matches semi-structural pruning across multiple LLMs, proving its wide applicability. The code is available at https://github.com/aim-uofa/LoRAPrune. | [
"Zhang, Mingyang",
"Chen, Hao",
"Shen, Chunhua",
"Yang, Zhen",
"Ou, Linlin",
"Yu, Xinyi",
"Zhuang, Bohan"
]
| {L}o{RAP}rune: Structured Pruning Meets Low-Rank Parameter-Efficient Fine-Tuning | findings-acl.178 | Poster | 2305.02176v2 |
https://aclanthology.org/2024.findings-acl.179.bib | @inproceedings{liu-etal-2024-speculative-decoding,
title = "Speculative Decoding via Early-exiting for Faster {LLM} Inference with {T}hompson Sampling Control Mechanism",
author = "Liu, Jiahao and
Wang, Qifan and
Wang, Jingang and
Cai, Xunliang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.179",
pages = "3027--3043",
abstract = "The recent advancements in large language models (LLMs) have been extraordinary, yet the escalating inference costs associated with them present challenges in real-world applications. To address these challenges, we propose a novel approach called Early-exiting Speculative Decoding (EESD) with lossless acceleration. Specifically, EESD utilizes a segment of the LLM to generate draft tokens, incorporating Early-exiting structures after the first N layers. To enhance the quality of draft tokens, a self-distillation method is integrated. This early-exiting design not only reduces deployment and training costs but also significantly accelerates the token generation speed. Moreover, we introduce a novel sampling mechanism that leverages Thompson Sampling to regulate the generation processes, automatically determining the quantity of draft tokens in each round. The original LLM is then employed to validate these draft tokens through a single forward pass, and thus guarantees that the final output text maintains a distribution consistent with vanilla auto-regressive decoding. The experimental results on both 13B and 70B models demonstrate that our approach decodes tokens at a markedly accelerated rate compared to prior methods, showing the effectiveness of our approach.",
}
| The recent advancements in large language models (LLMs) have been extraordinary, yet the escalating inference costs associated with them present challenges in real-world applications. To address these challenges, we propose a novel approach called Early-exiting Speculative Decoding (EESD) with lossless acceleration. Specifically, EESD utilizes a segment of the LLM to generate draft tokens, incorporating Early-exiting structures after the first N layers. To enhance the quality of draft tokens, a self-distillation method is integrated. This early-exiting design not only reduces deployment and training costs but also significantly accelerates the token generation speed. Moreover, we introduce a novel sampling mechanism that leverages Thompson Sampling to regulate the generation processes, automatically determining the quantity of draft tokens in each round. The original LLM is then employed to validate these draft tokens through a single forward pass, and thus guarantees that the final output text maintains a distribution consistent with vanilla auto-regressive decoding. The experimental results on both 13B and 70B models demonstrate that our approach decodes tokens at a markedly accelerated rate compared to prior methods, showing the effectiveness of our approach. | [
"Liu, Jiahao",
"Wang, Qifan",
"Wang, Jingang",
"Cai, Xunliang"
]
| Speculative Decoding via Early-exiting for Faster {LLM} Inference with {T}hompson Sampling Control Mechanism | findings-acl.179 | Poster | 2406.03853v1 |
https://aclanthology.org/2024.findings-acl.180.bib | @inproceedings{liu-etal-2024-towards-better,
title = "Towards Better Utilization of Multi-Reference Training Data for {C}hinese Grammatical Error Correction",
author = "Liu, Yumeng and
Li, Zhenghua and
Jiang, HaoChen and
Zhang, Bo and
Li, Chen and
Zhang, Ji",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.180",
pages = "3044--3052",
abstract = "For the grammatical error correction (GEC) task, there usually exist multiple correction ways for an erroneous input sentence, leading to multiple references. Observing the high proportion of multi-reference instances in Chinese GEC training data, we target a systematic study on how to better utilize multi-reference training data. We propose two new approaches and a simple two-stage training strategy. We compare them against previously proposed approaches, on two Chinese training datasets, i.e., Lang-8 for second language learner texts and FCGEC-Train for native speaker texts, and three test datasets. The experiments and analyses demonstrate the effectiveness of our proposed approaches and reveal interesting insights. Our code is available at https://github.com/ymliucs/MrGEC.",
}
| For the grammatical error correction (GEC) task, there usually exist multiple correction ways for an erroneous input sentence, leading to multiple references. Observing the high proportion of multi-reference instances in Chinese GEC training data, we target a systematic study on how to better utilize multi-reference training data. We propose two new approaches and a simple two-stage training strategy. We compare them against previously proposed approaches, on two Chinese training datasets, i.e., Lang-8 for second language learner texts and FCGEC-Train for native speaker texts, and three test datasets. The experiments and analyses demonstrate the effectiveness of our proposed approaches and reveal interesting insights. Our code is available at https://github.com/ymliucs/MrGEC. | [
"Liu, Yumeng",
"Li, Zhenghua",
"Jiang, HaoChen",
"Zhang, Bo",
"Li, Chen",
"Zhang, Ji"
]
| Towards Better Utilization of Multi-Reference Training Data for {C}hinese Grammatical Error Correction | findings-acl.180 | Poster | 2307.13923v2 |
https://aclanthology.org/2024.findings-acl.181.bib | @inproceedings{zeng-etal-2024-agenttuning,
title = "{A}gent{T}uning: Enabling Generalized Agent Abilities for {LLM}s",
author = "Zeng, Aohan and
Liu, Mingdao and
Lu, Rui and
Wang, Bowen and
Liu, Xiao and
Dong, Yuxiao and
Tang, Jie",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.181",
pages = "3053--3077",
abstract = "Open large language models (LLMs) with great performance in various tasks have significantly advanced the development of LLMs. However, they are far inferior to commercial models such as ChatGPT and GPT-4 when acting as agents to tackle complex tasks in the real world. These agent tasks employ LLMs as the central controller responsible for planning, memorization, and tool utilization, necessitating both fine-grained prompting methods and robust LLMs to achieve satisfactory performance. Though many prompting methods have been proposed to complete particular agent tasks, there is lack of research focusing on improving the agent capabilities of LLMs themselves without compromising their general abilities. In this work, we present AgentTuning, a simple and general method to enhance the agent abilities of LLMs while maintaining their general LLM capabilities. We construct AgentInstruct, a lightweight instruction-tuning dataset containing high-quality interaction trajectories. We employ a hybrid instruction-tuning strategy by combining AgentInstruct with open-source instructions from general domains. AgentTuning is used to instruction-tune the Llama 2 series, resulting in AgentLM. Our evaluations show that AgentTuning enables LLMs{'} agent capabilities without compromising general abilities. The AgentLM-70B is comparable to GPT-3.5-turbo on unseen agent tasks, demonstrating generalized agent capabilities. We open source the AgentInstruct and AgentLM-7B, 13B, and 70B models at https://anonymous.4open.science/r/AgentTuning, serving open and powerful alternatives to commercial LLMs for agent tasks.",
}
| Open large language models (LLMs) with great performance in various tasks have significantly advanced the development of LLMs. However, they are far inferior to commercial models such as ChatGPT and GPT-4 when acting as agents to tackle complex tasks in the real world. These agent tasks employ LLMs as the central controller responsible for planning, memorization, and tool utilization, necessitating both fine-grained prompting methods and robust LLMs to achieve satisfactory performance. Though many prompting methods have been proposed to complete particular agent tasks, there is lack of research focusing on improving the agent capabilities of LLMs themselves without compromising their general abilities. In this work, we present AgentTuning, a simple and general method to enhance the agent abilities of LLMs while maintaining their general LLM capabilities. We construct AgentInstruct, a lightweight instruction-tuning dataset containing high-quality interaction trajectories. We employ a hybrid instruction-tuning strategy by combining AgentInstruct with open-source instructions from general domains. AgentTuning is used to instruction-tune the Llama 2 series, resulting in AgentLM. Our evaluations show that AgentTuning enables LLMs{'} agent capabilities without compromising general abilities. The AgentLM-70B is comparable to GPT-3.5-turbo on unseen agent tasks, demonstrating generalized agent capabilities. We open source the AgentInstruct and AgentLM-7B, 13B, and 70B models at https://anonymous.4open.science/r/AgentTuning, serving open and powerful alternatives to commercial LLMs for agent tasks. | [
"Zeng, Aohan",
"Liu, Mingdao",
"Lu, Rui",
"Wang, Bowen",
"Liu, Xiao",
"Dong, Yuxiao",
"Tang, Jie"
]
| {A}gent{T}uning: Enabling Generalized Agent Abilities for {LLM}s | findings-acl.181 | Poster | 2310.12823v2 |
https://aclanthology.org/2024.findings-acl.182.bib | @inproceedings{ma-etal-2024-transition,
title = "Transition-based Opinion Generation for Aspect-based Sentiment Analysis",
author = "Ma, Tianlai and
Wang, Zhongqing and
Zhou, Guodong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.182",
pages = "3078--3087",
abstract = "Recently, the use of pre-trained generation models for extracting sentiment elements has resulted in significant advancements in aspect-based sentiment analysis benchmarks. However, these approaches often overlook the importance of explicitly modeling structure among sentiment elements. To address this limitation, we present a study that aims to integrate general pre-trained sequence-to-sequence language models with a structure-aware transition-based approach. Therefore, we propose a transition system for opinion tree generation, designed to better exploit pre-trained language models for structured fine-tuning. Our proposed transition system ensures the structural integrity of the generated opinion tree. By leveraging pre-trained generation models and simplifying the transition set, we are able to maximize the accuracy of opinion tree generation. Extensive experiments show that our model significantly advances the state-of-the-art performance on several benchmark datasets. In addition, the empirical studies also indicate that the proposed opinion tree generation with transition system is more effective in capturing the sentiment structure than other generation models.",
}
| Recently, the use of pre-trained generation models for extracting sentiment elements has resulted in significant advancements in aspect-based sentiment analysis benchmarks. However, these approaches often overlook the importance of explicitly modeling structure among sentiment elements. To address this limitation, we present a study that aims to integrate general pre-trained sequence-to-sequence language models with a structure-aware transition-based approach. Therefore, we propose a transition system for opinion tree generation, designed to better exploit pre-trained language models for structured fine-tuning. Our proposed transition system ensures the structural integrity of the generated opinion tree. By leveraging pre-trained generation models and simplifying the transition set, we are able to maximize the accuracy of opinion tree generation. Extensive experiments show that our model significantly advances the state-of-the-art performance on several benchmark datasets. In addition, the empirical studies also indicate that the proposed opinion tree generation with transition system is more effective in capturing the sentiment structure than other generation models. | [
"Ma, Tianlai",
"Wang, Zhongqing",
"Zhou, Guodong"
]
| Transition-based Opinion Generation for Aspect-based Sentiment Analysis | findings-acl.182 | Poster | 2306.08925v1 |
https://aclanthology.org/2024.findings-acl.183.bib | @inproceedings{wu-etal-2024-modeling,
title = "Modeling Dynamic Topics in Chain-Free Fashion by Evolution-Tracking Contrastive Learning and Unassociated Word Exclusion",
author = "Wu, Xiaobao and
Dong, Xinshuai and
Pan, Liangming and
Nguyen, Thong and
Luu, Anh Tuan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.183",
pages = "3088--3105",
abstract = "Dynamic topic models track the evolution of topics in sequential documents, which have derived various applications like trend analysis. However, existing models suffer from repetitive topic and unassociated topic issues, failing to reveal the evolution and hindering further applications. To address these issues, we break the tradition of simply chaining topics in existing work and propose a novel neural Chain-Free Dynamic Topic Model. We introduce a new evolution-tracking contrastive learning method that builds the similarity relations among dynamic topics. This not only tracks topic evolution but also maintains topic diversity, mitigating the repetitive topic issue. To avoid unassociated topics, we further present an unassociated word exclusion method that consistently excludes unassociated words from discovered topics. Extensive experiments demonstrate our model significantly outperforms state-of-the-art baselines, tracking topic evolution with high-quality topics, showing better performance on downstream tasks, and remaining robust to the hyperparameter for evolution intensities.",
}
| Dynamic topic models track the evolution of topics in sequential documents, which have derived various applications like trend analysis. However, existing models suffer from repetitive topic and unassociated topic issues, failing to reveal the evolution and hindering further applications. To address these issues, we break the tradition of simply chaining topics in existing work and propose a novel neural Chain-Free Dynamic Topic Model. We introduce a new evolution-tracking contrastive learning method that builds the similarity relations among dynamic topics. This not only tracks topic evolution but also maintains topic diversity, mitigating the repetitive topic issue. To avoid unassociated topics, we further present an unassociated word exclusion method that consistently excludes unassociated words from discovered topics. Extensive experiments demonstrate our model significantly outperforms state-of-the-art baselines, tracking topic evolution with high-quality topics, showing better performance on downstream tasks, and remaining robust to the hyperparameter for evolution intensities. | [
"Wu, Xiaobao",
"Dong, Xinshuai",
"Pan, Liangming",
"Nguyen, Thong",
"Luu, Anh Tuan"
]
| Modeling Dynamic Topics in Chain-Free Fashion by Evolution-Tracking Contrastive Learning and Unassociated Word Exclusion | findings-acl.183 | Poster | 2405.17957v1 |
https://aclanthology.org/2024.findings-acl.184.bib | @inproceedings{wang-etal-2024-chinese,
title = "A {C}hinese Dataset for Evaluating the Safeguards in Large Language Models",
author = "Wang, Yuxia and
Zhai, Zenan and
Li, Haonan and
Han, Xudong and
Lin, Shom and
Zhang, Zhenxuan and
Zhao, Angela and
Nakov, Preslav and
Baldwin, Timothy",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.184",
pages = "3106--3119",
abstract = "Many studies have demonstrated that large language models (LLMs) can produce harmful responses, exposing users to unexpected risks. Previous studies have proposed comprehensive taxonomies of LLM risks, as well as corresponding prompts that can be used to examine LLM safety. However, the focus has been almost exclusively on English. We aim to broaden LLM safety research by introducing a dataset for the safety evaluation of Chinese LLMs, and extending it to better identify false negative and false positive examples in terms of risky prompt rejections. We further present a set of fine-grained safety assessment criteria for each risk type, facilitating both manual annotation and automatic evaluation in terms of LLM response harmfulness. Our experiments over five LLMs show that region-specific risks are the prevalent risk type. Warning: this paper contains example data that may be offensive, harmful, or biased. Our data is available at https://github.com/Libr-AI/do-not-answer.",
}
| Many studies have demonstrated that large language models (LLMs) can produce harmful responses, exposing users to unexpected risks. Previous studies have proposed comprehensive taxonomies of LLM risks, as well as corresponding prompts that can be used to examine LLM safety. However, the focus has been almost exclusively on English. We aim to broaden LLM safety research by introducing a dataset for the safety evaluation of Chinese LLMs, and extending it to better identify false negative and false positive examples in terms of risky prompt rejections. We further present a set of fine-grained safety assessment criteria for each risk type, facilitating both manual annotation and automatic evaluation in terms of LLM response harmfulness. Our experiments over five LLMs show that region-specific risks are the prevalent risk type. Warning: this paper contains example data that may be offensive, harmful, or biased. Our data is available at https://github.com/Libr-AI/do-not-answer. | [
"Wang, Yuxia",
"Zhai, Zenan",
"Li, Haonan",
"Han, Xudong",
"Lin, Shom",
"Zhang, Zhenxuan",
"Zhao, Angela",
"Nakov, Preslav",
"Baldwin, Timothy"
]
| A {C}hinese Dataset for Evaluating the Safeguards in Large Language Models | findings-acl.184 | Poster | 2406.01364v1 |
https://aclanthology.org/2024.findings-acl.185.bib | @inproceedings{wang-etal-2024-llmfactor,
title = "{LLMF}actor: Extracting Profitable Factors through Prompts for Explainable Stock Movement Prediction",
author = "Wang, Meiyun and
Izumi, Kiyoshi and
Sakaji, Hiroki",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.185",
pages = "3120--3131",
abstract = "Recently, Large Language Models (LLMs) have attracted significant attention for their exceptional performance across a broad range of tasks, particularly in text analysis. However, the finance sector presents a distinct challenge due to its dependence on time-series data for complex forecasting tasks. In this study, we introduce a novel framework called LLMFactor, which employs Sequential Knowledge-Guided Prompting (SKGP) to identify factors that influence stock movements using LLMs. Unlike previous methods that relied on keyphrases or sentiment analysis, this approach focuses on extracting factors more directly related to stock market dynamics, providing clear explanations for complex temporal changes. Our framework directs the LLMs to create background knowledge through a fill-in-the-blank strategy and then discerns potential factors affecting stock prices from related news. Guided by background knowledge and identified factors, we leverage historical stock prices in textual format to predict stock movement. An extensive evaluation of the LLMFactor framework across four benchmark datasets from both the U.S. and Chinese stock markets demonstrates its superiority over existing state-of-the-art methods and its effectiveness in financial time-series forecasting.",
}
| Recently, Large Language Models (LLMs) have attracted significant attention for their exceptional performance across a broad range of tasks, particularly in text analysis. However, the finance sector presents a distinct challenge due to its dependence on time-series data for complex forecasting tasks. In this study, we introduce a novel framework called LLMFactor, which employs Sequential Knowledge-Guided Prompting (SKGP) to identify factors that influence stock movements using LLMs. Unlike previous methods that relied on keyphrases or sentiment analysis, this approach focuses on extracting factors more directly related to stock market dynamics, providing clear explanations for complex temporal changes. Our framework directs the LLMs to create background knowledge through a fill-in-the-blank strategy and then discerns potential factors affecting stock prices from related news. Guided by background knowledge and identified factors, we leverage historical stock prices in textual format to predict stock movement. An extensive evaluation of the LLMFactor framework across four benchmark datasets from both the U.S. and Chinese stock markets demonstrates its superiority over existing state-of-the-art methods and its effectiveness in financial time-series forecasting. | [
"Wang, Meiyun",
"Izumi, Kiyoshi",
"Sakaji, Hiroki"
]
| {LLMF}actor: Extracting Profitable Factors through Prompts for Explainable Stock Movement Prediction | findings-acl.185 | Poster | 2406.10811v1 |
https://aclanthology.org/2024.findings-acl.186.bib | @inproceedings{zhang-zhang-2024-look,
title = "You Only Look at Screens: Multimodal Chain-of-Action Agents",
author = "Zhang, Zhuosheng and
Zhang, Aston",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.186",
pages = "3132--3149",
abstract = "Autonomous graphical user interface (GUI) agents aim to facilitate task automation by interacting with the user interface without manual intervention. Recent studies have investigated eliciting the capabilities of large language models (LLMs) for effective engagement in diverse environments. To align with the input-output requirement of LLMs, most existing approaches are developed under a sandbox setting where they rely on external tools and application-specific APIs to parse the environment into textual elements and interpret the predicted actions. Consequently, those approaches often grapple with inference inefficiency and error propagation risks. To mitigate the challenges, we introduce Auto-GUI, a multimodal solution that directly interacts with the interface, bypassing the need for environment parsing or reliance on application-dependent APIs. Moreover, we propose a chain-of-action technique{---}leveraging a series of intermediate previous action histories and future action plans{---}to help the agent decide what action to execute. We evaluate our approach on a new device-control benchmark AITW with 30$K$ unique instructions, spanning multi-step tasks such as application operation, web searching, and web shopping. Experimental results show that Auto-GUI achieves state-of-the-art performance with an action type prediction accuracy of 90{\%} and an overall action success rate of 74{\%}. Code is publicly available at https://github.com/cooelf/Auto-GUI.",
}
| Autonomous graphical user interface (GUI) agents aim to facilitate task automation by interacting with the user interface without manual intervention. Recent studies have investigated eliciting the capabilities of large language models (LLMs) for effective engagement in diverse environments. To align with the input-output requirement of LLMs, most existing approaches are developed under a sandbox setting where they rely on external tools and application-specific APIs to parse the environment into textual elements and interpret the predicted actions. Consequently, those approaches often grapple with inference inefficiency and error propagation risks. To mitigate the challenges, we introduce Auto-GUI, a multimodal solution that directly interacts with the interface, bypassing the need for environment parsing or reliance on application-dependent APIs. Moreover, we propose a chain-of-action technique{---}leveraging a series of intermediate previous action histories and future action plans{---}to help the agent decide what action to execute. We evaluate our approach on a new device-control benchmark AITW with 30$K$ unique instructions, spanning multi-step tasks such as application operation, web searching, and web shopping. Experimental results show that Auto-GUI achieves state-of-the-art performance with an action type prediction accuracy of 90{\%} and an overall action success rate of 74{\%}. Code is publicly available at https://github.com/cooelf/Auto-GUI. | [
"Zhang, Zhuosheng",
"Zhang, Aston"
]
| You Only Look at Screens: Multimodal Chain-of-Action Agents | findings-acl.186 | Poster | 2401.05851v1 |
https://aclanthology.org/2024.findings-acl.187.bib | @inproceedings{hu-etal-2024-sp3,
title = "$\rm SP^3$: Enhancing Structured Pruning via {PCA} Projection",
author = "Hu, Yuxuan and
Zhang, Jing and
Zhao, Zhe and
Zhao, Chen and
Chen, Xiaodong and
Li, Cuiping and
Chen, Hong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.187",
pages = "3150--3170",
abstract = "Structured pruning is a widely used technique for reducing the size of pre-trained language models (PLMs), but current methods often overlook the potential of compressing the hidden dimension $d$ in PLMs, a dimension critical to model size and efficiency. This paper introduces a novel structured pruning approach, Structured Pruning with PCA Projection ($\rm SP^3$), targeting the effective reduction of $d$ by projecting features into a space defined by principal components before masking. Extensive experiments on benchmarks (GLUE and SQuAD) show that can reduce $d$ by 70{\%}, compress 94{\%} of the $\rm BERT_{base}$ model, and maintain over 96{\%} accuracy and outperform other methods that compress $d$ by 6{\%} in accuracy at the same compression ratio. $\rm SP^3$ has also proven effective with other models, including OPT and Llama.Our data and code are available at https://github.com/hyx1999/SP3",
}
| Structured pruning is a widely used technique for reducing the size of pre-trained language models (PLMs), but current methods often overlook the potential of compressing the hidden dimension $d$ in PLMs, a dimension critical to model size and efficiency. This paper introduces a novel structured pruning approach, Structured Pruning with PCA Projection ($\rm SP^3$), targeting the effective reduction of $d$ by projecting features into a space defined by principal components before masking. Extensive experiments on benchmarks (GLUE and SQuAD) show that can reduce $d$ by 70{\%}, compress 94{\%} of the $\rm BERT_{base}$ model, and maintain over 96{\%} accuracy and outperform other methods that compress $d$ by 6{\%} in accuracy at the same compression ratio. $\rm SP^3$ has also proven effective with other models, including OPT and Llama.Our data and code are available at https://github.com/hyx1999/SP3 | [
"Hu, Yuxuan",
"Zhang, Jing",
"Zhao, Zhe",
"Zhao, Chen",
"Chen, Xiaodong",
"Li, Cuiping",
"Chen, Hong"
]
| $\rm SP^3$: Enhancing Structured Pruning via {PCA} Projection | findings-acl.187 | Poster | 2308.16475 |
https://aclanthology.org/2024.findings-acl.188.bib | @inproceedings{park-etal-2024-gendex,
title = "{GENDEX}: Generative Data Augmentation Strategy Leveraging External Data for Abstractive Dialogue Summarization",
author = "Park, Sangwon and
Choi, Hongseok and
Choi, Dongha and
Lee, Hyunju",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.188",
pages = "3171--3185",
abstract = "With the proliferation of digital communication, dialogue summarization has become increasingly important. However, it still faces a shortage of data. To address this issue, we developed **Gen**erative **D**ata Augmentation Strategy Leveraging **Ex**ternal Data for Abstractive Dialogue Summarization (**GENDEX**), which is based on the hypothetical foundation that texts containing people and their interpersonal interactions can potentially serve as summaries of corresponding dialogues. We filter short texts containing people and resolve coreferences for better contextual analysis. We then identify the semantic roles of words within the texts and filter them based on the patterns observed in the dialogue summarization datasets. Using these texts, we generate synthetic dialogues through a controlled generation method. To better leverage the augmented data, we utilize noise-tolerant training to fine-tune the summarization model. The experimental results demonstrate the effectiveness of our proposed method, showing its robust performance, generalizability, and scalability. Moreover, performance improvements by *GENDEX* were observed regardless of complexity of dialogues. The code is available at https://github.com/DMCB-GIST/GENDEX.",
}
| With the proliferation of digital communication, dialogue summarization has become increasingly important. However, it still faces a shortage of data. To address this issue, we developed **Gen**erative **D**ata Augmentation Strategy Leveraging **Ex**ternal Data for Abstractive Dialogue Summarization (**GENDEX**), which is based on the hypothetical foundation that texts containing people and their interpersonal interactions can potentially serve as summaries of corresponding dialogues. We filter short texts containing people and resolve coreferences for better contextual analysis. We then identify the semantic roles of words within the texts and filter them based on the patterns observed in the dialogue summarization datasets. Using these texts, we generate synthetic dialogues through a controlled generation method. To better leverage the augmented data, we utilize noise-tolerant training to fine-tune the summarization model. The experimental results demonstrate the effectiveness of our proposed method, showing its robust performance, generalizability, and scalability. Moreover, performance improvements by *GENDEX* were observed regardless of complexity of dialogues. The code is available at https://github.com/DMCB-GIST/GENDEX. | [
"Park, Sangwon",
"Choi, Hongseok",
"Choi, Dongha",
"Lee, Hyunju"
]
| {GENDEX}: Generative Data Augmentation Strategy Leveraging External Data for Abstractive Dialogue Summarization | findings-acl.188 | Poster | 2210.09474v1 |
https://aclanthology.org/2024.findings-acl.189.bib | @inproceedings{carmeli-etal-2024-concept,
title = "Concept-Best-Matching: Evaluating Compositionality In Emergent Communication",
author = "Carmeli, Boaz and
Belinkov, Yonatan and
Meir, Ron",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.189",
pages = "3186--3194",
abstract = "Artificial agents that learn to communicate in order to accomplish a given task acquire communication protocols that are typically opaque to a human. A large body of work has attempted to evaluate the emergent communication via various evaluation measures, with **compositionality** featuring as a prominent desired trait. However, current evaluation procedures do not directly expose the compositionality of the emergent communication. We propose a procedure to assess the compositionality of emergent communication by finding the best-match between emerged words and natural language concepts.The best-match algorithm provides both a global score and a translation-map from emergent words to natural language concepts. To the best of our knowledge, it is the first time that such direct and interpretable mapping between emergent words and human concepts is provided.",
}
| Artificial agents that learn to communicate in order to accomplish a given task acquire communication protocols that are typically opaque to a human. A large body of work has attempted to evaluate the emergent communication via various evaluation measures, with **compositionality** featuring as a prominent desired trait. However, current evaluation procedures do not directly expose the compositionality of the emergent communication. We propose a procedure to assess the compositionality of emergent communication by finding the best-match between emerged words and natural language concepts.The best-match algorithm provides both a global score and a translation-map from emergent words to natural language concepts. To the best of our knowledge, it is the first time that such direct and interpretable mapping between emergent words and human concepts is provided. | [
"Carmeli, Boaz",
"Belinkov, Yonatan",
"Meir, Ron"
]
| Concept-Best-Matching: Evaluating Compositionality In Emergent Communication | findings-acl.189 | Poster | 2010.15058v2 |
https://aclanthology.org/2024.findings-acl.190.bib | @inproceedings{t-y-s-s-etal-2024-tale,
title = "A Tale of Two Revisions: Summarizing Changes Across Document Versions",
author = "T.y.s.s, Santosh and
Modani, Natwar and
Saxena, Apoorv",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.190",
pages = "3195--3211",
abstract = "Document revision is a crucial aspect of the writing process, particularly in collaborative environments where multiple authors contribute simultaneously. However, current tools lack an efficient way to provide a comprehensive overview of changes between versions, leading to difficulties in understanding revisions. To address this, we propose a novel task of providing thematic summary of changes between document versions, organizing individual edits based on shared themes. We assess capabilities of LLMs on this task and further introduce three strategies to tackle this task: (i) representing the input of two documents along with edits in the {`}diff{'} format (ii) a two-stage task decomposition with individual edit description generation as an intermediate task and (iii) clustering based chunking and subsequent merging techniques for handling longer documents. Our experiments demonstrate the effectiveness of our approach in improving the model{'}s capacity to handle this complex task. Additionally, we introduce ChangeSumm, a curated dataset comprising human-written thematic summaries for pairs of document versions, to facilitate evaluation and further research in this direction.",
}
| Document revision is a crucial aspect of the writing process, particularly in collaborative environments where multiple authors contribute simultaneously. However, current tools lack an efficient way to provide a comprehensive overview of changes between versions, leading to difficulties in understanding revisions. To address this, we propose a novel task of providing thematic summary of changes between document versions, organizing individual edits based on shared themes. We assess capabilities of LLMs on this task and further introduce three strategies to tackle this task: (i) representing the input of two documents along with edits in the {`}diff{'} format (ii) a two-stage task decomposition with individual edit description generation as an intermediate task and (iii) clustering based chunking and subsequent merging techniques for handling longer documents. Our experiments demonstrate the effectiveness of our approach in improving the model{'}s capacity to handle this complex task. Additionally, we introduce ChangeSumm, a curated dataset comprising human-written thematic summaries for pairs of document versions, to facilitate evaluation and further research in this direction. | [
"T.y.s.s, Santosh",
"Modani, Natwar",
"Saxena, Apoorv"
]
| A Tale of Two Revisions: Summarizing Changes Across Document Versions | findings-acl.190 | Poster | 2211.14358v1 |
https://aclanthology.org/2024.findings-acl.191.bib | @inproceedings{su-etal-2024-refine,
title = "Refine, Align, and Aggregate: Multi-view Linguistic Features Enhancement for Aspect Sentiment Triplet Extraction",
author = "Su, Guixin and
Wu, Mingmin and
Huang, Zhongqiang and
Zhang, Yongcheng and
Wang, Tongguan and
Hu, Yuxue and
Sha, Ying",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.191",
pages = "3212--3228",
abstract = "Aspect Sentiment Triplet Extraction (ASTE) aims to extract the triplets of aspect terms, their associated sentiment and opinion terms. Previous works based on different modeling paradigms have achieved promising results. However, these methods struggle to comprehensively explore the various specific relations between sentiment elements in multi-view linguistic features, which is the prior indication effect for facilitating sentiment triplets extraction, requiring to align and aggregate them to capture the complementary higher-order interactions. In this paper, we propose Multi-view Linguistic Features Enhancement (MvLFE) to explore the aforementioned prior indication effect in the {``}Refine, Align, and Aggregate{''} learning process. Specifically, we first introduce the relational graph attention network to encode the word-pair relations represented by each linguistic feature and refine them to pay more attention to the aspect-opinion pairs. Next, we employ the multi-view contrastive learning to align them at a fine-grained level in the contextual semantic space to maintain semantic consistency. Finally, we utilize the multi-semantic cross attention to capture and aggregate the complementary higher-order interactions between diverse linguistic features to enhance the aspect-opinion relations. Experimental results on several benchmark datasets show the effectiveness and robustness of our model, which achieves state-of-the-art performance.",
}
| Aspect Sentiment Triplet Extraction (ASTE) aims to extract the triplets of aspect terms, their associated sentiment and opinion terms. Previous works based on different modeling paradigms have achieved promising results. However, these methods struggle to comprehensively explore the various specific relations between sentiment elements in multi-view linguistic features, which is the prior indication effect for facilitating sentiment triplets extraction, requiring to align and aggregate them to capture the complementary higher-order interactions. In this paper, we propose Multi-view Linguistic Features Enhancement (MvLFE) to explore the aforementioned prior indication effect in the {``}Refine, Align, and Aggregate{''} learning process. Specifically, we first introduce the relational graph attention network to encode the word-pair relations represented by each linguistic feature and refine them to pay more attention to the aspect-opinion pairs. Next, we employ the multi-view contrastive learning to align them at a fine-grained level in the contextual semantic space to maintain semantic consistency. Finally, we utilize the multi-semantic cross attention to capture and aggregate the complementary higher-order interactions between diverse linguistic features to enhance the aspect-opinion relations. Experimental results on several benchmark datasets show the effectiveness and robustness of our model, which achieves state-of-the-art performance. | [
"Su, Guixin",
"Wu, Mingmin",
"Huang, Zhongqiang",
"Zhang, Yongcheng",
"Wang, Tongguan",
"Hu, Yuxue",
"Sha, Ying"
]
| Refine, Align, and Aggregate: Multi-view Linguistic Features Enhancement for Aspect Sentiment Triplet Extraction | findings-acl.191 | Poster | 2306.08373v1 |
https://aclanthology.org/2024.findings-acl.192.bib | @inproceedings{li-zhang-2024-pro,
title = "Pro-Woman, Anti-Man? Identifying Gender Bias in Stance Detection",
author = "Li, Yingjie and
Zhang, Yue",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.192",
pages = "3229--3236",
abstract = "Gender bias has been widely observed in NLP models, which has the potential to perpetuate harmful stereotypes and discrimination. In this paper, we construct a dataset GenderStance of 36k samples to measure gender bias in stance detection, determining whether models consistently predict the same stance for a particular gender group. We find that all models are gender-biased and prone to classify sentences that contain male nouns as Against and those with female nouns as Favor. Moreover, extensive experiments indicate that sources of gender bias stem from the fine-tuning data and the foundation model itself. We will publicly release our code and dataset.",
}
| Gender bias has been widely observed in NLP models, which has the potential to perpetuate harmful stereotypes and discrimination. In this paper, we construct a dataset GenderStance of 36k samples to measure gender bias in stance detection, determining whether models consistently predict the same stance for a particular gender group. We find that all models are gender-biased and prone to classify sentences that contain male nouns as Against and those with female nouns as Favor. Moreover, extensive experiments indicate that sources of gender bias stem from the fine-tuning data and the foundation model itself. We will publicly release our code and dataset. | [
"Li, Yingjie",
"Zhang, Yue"
]
| Pro-Woman, Anti-Man? Identifying Gender Bias in Stance Detection | findings-acl.192 | Poster | 2405.10991v1 |
https://aclanthology.org/2024.findings-acl.193.bib | @inproceedings{ohi-etal-2024-likelihood,
title = "Likelihood-based Mitigation of Evaluation Bias in Large Language Models",
author = "Ohi, Masanari and
Kaneko, Masahiro and
Koike, Ryuto and
Loem, Mengsay and
Okazaki, Naoaki",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.193",
pages = "3237--3245",
abstract = "Large Language Models (LLMs) are widely used to evaluate natural language generation tasks as automated metrics.However, the likelihood, a measure of LLM{'}s plausibility for a sentence, can vary due to superficial differences in sentences, such as word order and sentence structure.It is therefore possible that there might be a likelihood bias if LLMs are used for evaluation: they might overrate sentences with higher likelihoods while underrating those with lower likelihoods.In this paper, we investigate the presence and impact of likelihood bias in LLM-based evaluators.We also propose a method to mitigate the likelihood bias.Our method utilizes highly biased instances as few-shot examples for in-context learning.Our experiments in evaluating the data-to-text and grammatical error correction tasks reveal that several LLMs we test display a likelihood bias.Furthermore, our proposed method successfully mitigates this bias, also improving evaluation performance (in terms of correlation of models with human scores) significantly.",
}
| Large Language Models (LLMs) are widely used to evaluate natural language generation tasks as automated metrics.However, the likelihood, a measure of LLM{'}s plausibility for a sentence, can vary due to superficial differences in sentences, such as word order and sentence structure.It is therefore possible that there might be a likelihood bias if LLMs are used for evaluation: they might overrate sentences with higher likelihoods while underrating those with lower likelihoods.In this paper, we investigate the presence and impact of likelihood bias in LLM-based evaluators.We also propose a method to mitigate the likelihood bias.Our method utilizes highly biased instances as few-shot examples for in-context learning.Our experiments in evaluating the data-to-text and grammatical error correction tasks reveal that several LLMs we test display a likelihood bias.Furthermore, our proposed method successfully mitigates this bias, also improving evaluation performance (in terms of correlation of models with human scores) significantly. | [
"Ohi, Masanari",
"Kaneko, Masahiro",
"Koike, Ryuto",
"Loem, Mengsay",
"Okazaki, Naoaki"
]
| Likelihood-based Mitigation of Evaluation Bias in Large Language Models | findings-acl.193 | Poster | 2407.05250v1 |
https://aclanthology.org/2024.findings-acl.194.bib | @inproceedings{li-etal-2024-music,
title = "The Music Maestro or The Musically Challenged, A Massive Music Evaluation Benchmark for Large Language Models",
author = "Li, Jiajia and
Yang, Lu and
Tang, Mingni and
Chenchong, Chenchong and
Li, Zuchao and
Wang, Ping and
Zhao, Hai",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.194",
pages = "3246--3257",
abstract = "Benchmark plays a pivotal role in assessing the advancements of large language models (LLMs). While numerous benchmarks have been proposed to evaluate LLMs{'} capabilities, there is a notable absence of a dedicated benchmark for assessing their musical abilities. To address this gap, we present ZIQI-Eval, a comprehensive and large-scale music benchmark specifically designed to evaluate the music-related capabilities of LLMs.ZIQI-Eval encompasses a wide range of questions, covering 10 major categories and 56 subcategories, resulting in over 14,000 meticulously curated data entries. By leveraging ZIQI-Eval, we conduct a comprehensive evaluation over 16 LLMs to evaluate and analyze LLMs{'} performance in the domain of music.Results indicate that all LLMs perform poorly on the ZIQI-Eval benchmark, suggesting significant room for improvement in their musical capabilities.With ZIQI-Eval, we aim to provide a standardized and robust evaluation framework that facilitates a comprehensive assessment of LLMs{'} music-related abilities. The dataset is available at GitHub and HuggingFace.",
}
| Benchmark plays a pivotal role in assessing the advancements of large language models (LLMs). While numerous benchmarks have been proposed to evaluate LLMs{'} capabilities, there is a notable absence of a dedicated benchmark for assessing their musical abilities. To address this gap, we present ZIQI-Eval, a comprehensive and large-scale music benchmark specifically designed to evaluate the music-related capabilities of LLMs.ZIQI-Eval encompasses a wide range of questions, covering 10 major categories and 56 subcategories, resulting in over 14,000 meticulously curated data entries. By leveraging ZIQI-Eval, we conduct a comprehensive evaluation over 16 LLMs to evaluate and analyze LLMs{'} performance in the domain of music.Results indicate that all LLMs perform poorly on the ZIQI-Eval benchmark, suggesting significant room for improvement in their musical capabilities.With ZIQI-Eval, we aim to provide a standardized and robust evaluation framework that facilitates a comprehensive assessment of LLMs{'} music-related abilities. The dataset is available at GitHub and HuggingFace. | [
"Li, Jiajia",
"Yang, Lu",
"Tang, Mingni",
"Chenchong, Chenchong",
"Li, Zuchao",
"Wang, Ping",
"Zhao, Hai"
]
| The Music Maestro or The Musically Challenged, A Massive Music Evaluation Benchmark for Large Language Models | findings-acl.194 | Poster | 2406.15885v1 |
https://aclanthology.org/2024.findings-acl.195.bib | @inproceedings{yang-etal-2024-pyramidinfer,
title = "{P}yramid{I}nfer: Pyramid {KV} Cache Compression for High-throughput {LLM} Inference",
author = "Yang, Dongjie and
Han, Xiaodong and
Gao, Yan and
Hu, Yao and
Zhang, Shilin and
Zhao, Hai",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.195",
pages = "3258--3270",
abstract = "Large Language Models (LLMs) have shown remarkable comprehension abilities but face challenges in GPU memory usage during inference, hindering their scalability for real-time applications like chatbots. To accelerate inference, we store computed keys and values (KV cache) in the GPU memory. Existing methods study the KV cache compression to reduce memory by pruning the pre-computed KV cache. However, they neglect the inter-layer dependency between layers and huge memory consumption in pre-computation. To explore these deficiencies, we find that the number of crucial keys and values that influence future generations decreases layer by layer and we can extract them by the consistency in attention weights. Based on the findings, we propose PyramidInfer, a method that compresses the KV cache by layer-wise retaining crucial context. PyramidInfer saves significant memory by computing fewer keys and values without sacrificing performance. Experimental results show PyramidInfer improves 2.2x throughput compared to Accelerate with over 54{\%} GPU memory reduction in KV cache.",
}
| Large Language Models (LLMs) have shown remarkable comprehension abilities but face challenges in GPU memory usage during inference, hindering their scalability for real-time applications like chatbots. To accelerate inference, we store computed keys and values (KV cache) in the GPU memory. Existing methods study the KV cache compression to reduce memory by pruning the pre-computed KV cache. However, they neglect the inter-layer dependency between layers and huge memory consumption in pre-computation. To explore these deficiencies, we find that the number of crucial keys and values that influence future generations decreases layer by layer and we can extract them by the consistency in attention weights. Based on the findings, we propose PyramidInfer, a method that compresses the KV cache by layer-wise retaining crucial context. PyramidInfer saves significant memory by computing fewer keys and values without sacrificing performance. Experimental results show PyramidInfer improves 2.2x throughput compared to Accelerate with over 54{\%} GPU memory reduction in KV cache. | [
"Yang, Dongjie",
"Han, Xiaodong",
"Gao, Yan",
"Hu, Yao",
"Zhang, Shilin",
"Zhao, Hai"
]
| {P}yramid{I}nfer: Pyramid {KV} Cache Compression for High-throughput {LLM} Inference | findings-acl.195 | Poster | 2404.04793v1 |
https://aclanthology.org/2024.findings-acl.196.bib | @inproceedings{wu-etal-2024-role,
title = "From Role-Play to Drama-Interaction: An {LLM} Solution",
author = "Wu, Weiqi and
Wu, Hongqiu and
Jiang, Lai and
Liu, Xingyuan and
Zhao, Hai and
Zhang, Min",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.196",
pages = "3271--3290",
abstract = "Drama is a form of storytelling inspired by human creativity, proceeding with a predefined storyline, carrying emotions and thoughts.This paper introduces LLM-based interactive drama, which endows traditional drama with an unprecedented immersion, where a person is allowed to walk into it and interact with the characters and scenes.We define this new artistic genre by 6 essential elements{---}plot, character, thought, diction, spectacle and interaction{---}and study the entire pipeline to forge a backbone drama LLM to drive the playing process, which is challenged by limited drama resources, uncontrollable narrative development, and complicated instruction following.We propose Narrative Chain to offer finer control over the narrative progression during interaction with players;Auto-Drama to synthesize drama scripts given arbitrary stories;Sparse Instruction Tuning to allow the model to follow sophisticated instructions.We manually craft 3 scripts, Detective Conan, Harry Potter, Romeo and Juliet, and design a 5-dimension principle to evaluate the drama LLM comprehensively.",
}
| Drama is a form of storytelling inspired by human creativity, proceeding with a predefined storyline, carrying emotions and thoughts.This paper introduces LLM-based interactive drama, which endows traditional drama with an unprecedented immersion, where a person is allowed to walk into it and interact with the characters and scenes.We define this new artistic genre by 6 essential elements{---}plot, character, thought, diction, spectacle and interaction{---}and study the entire pipeline to forge a backbone drama LLM to drive the playing process, which is challenged by limited drama resources, uncontrollable narrative development, and complicated instruction following.We propose Narrative Chain to offer finer control over the narrative progression during interaction with players;Auto-Drama to synthesize drama scripts given arbitrary stories;Sparse Instruction Tuning to allow the model to follow sophisticated instructions.We manually craft 3 scripts, Detective Conan, Harry Potter, Romeo and Juliet, and design a 5-dimension principle to evaluate the drama LLM comprehensively. | [
"Wu, Weiqi",
"Wu, Hongqiu",
"Jiang, Lai",
"Liu, Xingyuan",
"Zhao, Hai",
"Zhang, Min"
]
| From Role-Play to Drama-Interaction: An {LLM} Solution | findings-acl.196 | Poster | 2408.02479v1 |
https://aclanthology.org/2024.findings-acl.197.bib | @inproceedings{ahn-etal-2024-timechara,
title = "{T}ime{C}hara: Evaluating Point-in-Time Character Hallucination of Role-Playing Large Language Models",
author = "Ahn, Jaewoo and
Lee, Taehyun and
Lim, Junyoung and
Kim, Jin-Hwa and
Yun, Sangdoo and
Lee, Hwaran and
Kim, Gunhee",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.197",
pages = "3291--3325",
abstract = "While Large Language Models (LLMs) can serve as agents to simulate human behaviors (i.e., role-playing agents), we emphasize the importance of point-in-time role-playing. This situates characters at specific moments in the narrative progression for three main reasons: (i) enhancing users{'} narrative immersion, (ii) avoiding spoilers, and (iii) fostering engagement in fandom role-playing. To accurately represent characters at specific time points, agents must avoid character hallucination, where they display knowledge that contradicts their characters{'} identities and historical timelines. We introduce TimeChara, a new benchmark designed to evaluate point-in-time character hallucination in role-playing LLMs. Comprising 10,895 instances generated through an automated pipeline, this benchmark reveals significant hallucination issues in current state-of-the-art LLMs (e.g., GPT-4o). To counter this challenge, we propose Narrative-Experts, a method that decomposes the reasoning steps and utilizes narrative experts to reduce point-in-time character hallucinations effectively. Still, our findings with TimeChara highlight the ongoing challenges of point-in-time character hallucination, calling for further study.",
}
| While Large Language Models (LLMs) can serve as agents to simulate human behaviors (i.e., role-playing agents), we emphasize the importance of point-in-time role-playing. This situates characters at specific moments in the narrative progression for three main reasons: (i) enhancing users{'} narrative immersion, (ii) avoiding spoilers, and (iii) fostering engagement in fandom role-playing. To accurately represent characters at specific time points, agents must avoid character hallucination, where they display knowledge that contradicts their characters{'} identities and historical timelines. We introduce TimeChara, a new benchmark designed to evaluate point-in-time character hallucination in role-playing LLMs. Comprising 10,895 instances generated through an automated pipeline, this benchmark reveals significant hallucination issues in current state-of-the-art LLMs (e.g., GPT-4o). To counter this challenge, we propose Narrative-Experts, a method that decomposes the reasoning steps and utilizes narrative experts to reduce point-in-time character hallucinations effectively. Still, our findings with TimeChara highlight the ongoing challenges of point-in-time character hallucination, calling for further study. | [
"Ahn, Jaewoo",
"Lee, Taehyun",
"Lim, Junyoung",
"Kim, Jin-Hwa",
"Yun, Sangdoo",
"Lee, Hwaran",
"Kim, Gunhee"
]
| {T}ime{C}hara: Evaluating Point-in-Time Character Hallucination of Role-Playing Large Language Models | findings-acl.197 | Poster | 2312.14504v2 |
https://aclanthology.org/2024.findings-acl.198.bib | @inproceedings{li-etal-2024-red,
title = "Red Teaming Visual Language Models",
author = "Li, Mukai and
Li, Lei and
Yin, Yuwei and
Ahmed, Masood and
Liu, Zhenguang and
Liu, Qi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.198",
pages = "3326--3342",
abstract = "VLMs (Vision-Language Models) extend the capabilities of LLMs (Large Language Models) to accept multimodal inputs. Since it has been verified that LLMs can be induced to generate harmful or inaccurate content through specific test cases (termed as Red Teaming), how VLMs perform in similar scenarios, especially with their combination of textual and visual inputs, remains a question. To explore this problem, we present a novel red teaming dataset RTVLM, which encompasses 12 subtasks (e.g., image misleading, multi-modal jailbreaking, face fairness, etc) under 4 primary aspects (faithfulness, privacy, safety, fairness). Our RTVLM is the first red teaming dataset to benchmark current VLMs in terms of these 4 different aspects. Detailed analysis shows that 10 prominent open-sourced VLMs struggle with the red teaming in different degrees and have up to 31{\%} performance gap with GPT-4V. Additionally, we simply apply red teaming alignment to LLaVA-v1.5 with Supervised Fine-tuning (SFT) using RTVLM, and this bolsters the models{'} performance with 10{\%} in RTVLM test set, 13{\%} in MM-hallu, and without noticeable decline in MM-Bench, overpassing other LLaVA-based models in similar size with regular alignment data. This reveals that current open-sourced VLMs still lack red teaming alignment. Our code and datasets will be open-sourced.",
}
| VLMs (Vision-Language Models) extend the capabilities of LLMs (Large Language Models) to accept multimodal inputs. Since it has been verified that LLMs can be induced to generate harmful or inaccurate content through specific test cases (termed as Red Teaming), how VLMs perform in similar scenarios, especially with their combination of textual and visual inputs, remains a question. To explore this problem, we present a novel red teaming dataset RTVLM, which encompasses 12 subtasks (e.g., image misleading, multi-modal jailbreaking, face fairness, etc) under 4 primary aspects (faithfulness, privacy, safety, fairness). Our RTVLM is the first red teaming dataset to benchmark current VLMs in terms of these 4 different aspects. Detailed analysis shows that 10 prominent open-sourced VLMs struggle with the red teaming in different degrees and have up to 31{\%} performance gap with GPT-4V. Additionally, we simply apply red teaming alignment to LLaVA-v1.5 with Supervised Fine-tuning (SFT) using RTVLM, and this bolsters the models{'} performance with 10{\%} in RTVLM test set, 13{\%} in MM-hallu, and without noticeable decline in MM-Bench, overpassing other LLaVA-based models in similar size with regular alignment data. This reveals that current open-sourced VLMs still lack red teaming alignment. Our code and datasets will be open-sourced. | [
"Li, Mukai",
"Li, Lei",
"Yin, Yuwei",
"Ahmed, Masood",
"Liu, Zhenguang",
"Liu, Qi"
]
| Red Teaming Visual Language Models | findings-acl.198 | Poster | 2401.12915v1 |
https://aclanthology.org/2024.findings-acl.199.bib | @inproceedings{yang-etal-2024-enhancing,
title = "Enhancing Semantic Consistency of Large Language Models through Model Editing: An Interpretability-Oriented Approach",
author = "Yang, Jingyuan and
Chen, Dapeng and
Sun, Yajing and
Li, Rongjun and
Feng, Zhiyong and
Peng, Wei",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.199",
pages = "3343--3353",
abstract = "A Large Language Model (LLM) tends to generate inconsistent and sometimes contradictory outputs when presented with a prompt that has equivalent semantics but is expressed differently from the original prompt. To achieve semantic consistency of an LLM, one of the key approaches is to finetune the model with prompt-output pairs with semantically equivalent meanings. Despite its effectiveness, a data-driven finetuning method incurs substantial computation costs in data preparation and model optimization. In this regime, an LLM is treated as a {``}black box{''}, restricting our ability to gain deeper insights into its internal mechanism. In this paper, we are motivated to enhance the semantic consistency of LLMs through a more interpretable method (i.e., model editing) to this end. We first identify the model components (i.e., attention heads) that have a key impact on the semantic consistency of an LLM. We subsequently inject biases into the output of these model components along the semantic-consistency activation direction. It is noteworthy that these modifications are cost-effective, without reliance on mass manipulations of the original model parameters. Through comprehensive experiments on the constructed NLU and open-source NLG datasets, our method demonstrates significant improvements in the semantic consistency and task performance of LLMs. Additionally, our method exhibits promising generalization capabilities by performing well on tasks beyond the primary tasks.",
}
| A Large Language Model (LLM) tends to generate inconsistent and sometimes contradictory outputs when presented with a prompt that has equivalent semantics but is expressed differently from the original prompt. To achieve semantic consistency of an LLM, one of the key approaches is to finetune the model with prompt-output pairs with semantically equivalent meanings. Despite its effectiveness, a data-driven finetuning method incurs substantial computation costs in data preparation and model optimization. In this regime, an LLM is treated as a {``}black box{''}, restricting our ability to gain deeper insights into its internal mechanism. In this paper, we are motivated to enhance the semantic consistency of LLMs through a more interpretable method (i.e., model editing) to this end. We first identify the model components (i.e., attention heads) that have a key impact on the semantic consistency of an LLM. We subsequently inject biases into the output of these model components along the semantic-consistency activation direction. It is noteworthy that these modifications are cost-effective, without reliance on mass manipulations of the original model parameters. Through comprehensive experiments on the constructed NLU and open-source NLG datasets, our method demonstrates significant improvements in the semantic consistency and task performance of LLMs. Additionally, our method exhibits promising generalization capabilities by performing well on tasks beyond the primary tasks. | [
"Yang, Jingyuan",
"Chen, Dapeng",
"Sun, Yajing",
"Li, Rongjun",
"Feng, Zhiyong",
"Peng, Wei"
]
| Enhancing Semantic Consistency of Large Language Models through Model Editing: An Interpretability-Oriented Approach | findings-acl.199 | Poster | 2405.16011v1 |
https://aclanthology.org/2024.findings-acl.200.bib | @inproceedings{shin-etal-2024-semantic,
title = "Semantic Skill Grounding for Embodied Instruction-Following in Cross-Domain Environments",
author = "Shin, Sangwoo and
Kim, SeungHyun and
Jang, Youngsoo and
Lee, Moontae and
Woo, Honguk",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.200",
pages = "3354--3376",
abstract = "In embodied instruction-following (EIF), the integration of pretrained language models (LMs) as task planners emerges as a significant branch, where tasks are planned at the skill level by prompting LMs with pretrained skills and user instructions. However, grounding these pretrained skills in different domains remains challenging due to their intricate entanglement with the domain-specific knowledge. To address this challenge, we present a semantic skill grounding (SemGro) framework that leverages the hierarchical nature of semantic skills. SemGro recognizes the broad spectrum of these skills, ranging from short-horizon low-semantic skills that are universally applicable across domains to long-horizon rich-semantic skills that are highly specialized and tailored for particular domains. The framework employs an iterative skill decomposition approach, starting from the higher levels of semantic skill hierarchy and then moving downwards, so as to ground each planned skill to an executable level within the target domain. To do so, we use the reasoning capabilities of LMs for composing and decomposing semantic skills, as well as their multi-modal extension for assessing the skill feasibility in the target domain. Our experiments in the VirtualHome benchmark show the efficacy of SemGro in 300 cross-domain EIF scenarios.",
}
| In embodied instruction-following (EIF), the integration of pretrained language models (LMs) as task planners emerges as a significant branch, where tasks are planned at the skill level by prompting LMs with pretrained skills and user instructions. However, grounding these pretrained skills in different domains remains challenging due to their intricate entanglement with the domain-specific knowledge. To address this challenge, we present a semantic skill grounding (SemGro) framework that leverages the hierarchical nature of semantic skills. SemGro recognizes the broad spectrum of these skills, ranging from short-horizon low-semantic skills that are universally applicable across domains to long-horizon rich-semantic skills that are highly specialized and tailored for particular domains. The framework employs an iterative skill decomposition approach, starting from the higher levels of semantic skill hierarchy and then moving downwards, so as to ground each planned skill to an executable level within the target domain. To do so, we use the reasoning capabilities of LMs for composing and decomposing semantic skills, as well as their multi-modal extension for assessing the skill feasibility in the target domain. Our experiments in the VirtualHome benchmark show the efficacy of SemGro in 300 cross-domain EIF scenarios. | [
"Shin, Sangwoo",
"Kim, SeungHyun",
"Jang, Youngsoo",
"Lee, Moontae",
"Woo, Honguk"
]
| Semantic Skill Grounding for Embodied Instruction-Following in Cross-Domain Environments | findings-acl.200 | Poster | 2408.01024v1 |
https://aclanthology.org/2024.findings-acl.201.bib | @inproceedings{zhu-etal-2024-lire,
title = "{LIRE}: listwise reward enhancement for preference alignment",
author = "Zhu, Mingye and
Liu, Yi and
Zhang, Lei and
Guo, Junbo and
Mao, Zhendong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.201",
pages = "3377--3394",
abstract = "Recently, tremendous strides have been made to align the generation of Large Language Models (LLMs) with human values to mitigate toxic or unhelpful content. Leveraging Reinforcement Learning from Human Feedback (RLHF) proves effective and is widely adopted by researchers. However, implementing RLHF is complex, and its sensitivity to hyperparameters renders achieving stable performance and scalability challenging. Furthermore, prevailing approaches to preference alignment primarily concentrate on pairwise comparisons, with limited exploration into multi-response scenarios, thereby overlooking the potential richness within the candidate pool. For the above reasons, we propose a new approach: Listwise Reward Enhancement for Preference Alignment (LIRE), a gradient-based reward optimization approach that incorporates the offline rewards of multiple responses into a streamlined listwise framework, thus eliminating the need for online sampling during training. LIRE is straightforward to implement, requiring minimal parameter tuning, and seamlessly aligns with the pairwise paradigm while naturally extending to multi-response scenarios. Moreover, we introduce a self-enhancement algorithm aimed at iteratively refining the reward during training. Our experiments demonstrate that LIRE consistently outperforms existing methods across several benchmarks on dialogue and summarization tasks, with good transferability to out-of-distribution data, assessed using proxy reward models and human annotators.",
}
| Recently, tremendous strides have been made to align the generation of Large Language Models (LLMs) with human values to mitigate toxic or unhelpful content. Leveraging Reinforcement Learning from Human Feedback (RLHF) proves effective and is widely adopted by researchers. However, implementing RLHF is complex, and its sensitivity to hyperparameters renders achieving stable performance and scalability challenging. Furthermore, prevailing approaches to preference alignment primarily concentrate on pairwise comparisons, with limited exploration into multi-response scenarios, thereby overlooking the potential richness within the candidate pool. For the above reasons, we propose a new approach: Listwise Reward Enhancement for Preference Alignment (LIRE), a gradient-based reward optimization approach that incorporates the offline rewards of multiple responses into a streamlined listwise framework, thus eliminating the need for online sampling during training. LIRE is straightforward to implement, requiring minimal parameter tuning, and seamlessly aligns with the pairwise paradigm while naturally extending to multi-response scenarios. Moreover, we introduce a self-enhancement algorithm aimed at iteratively refining the reward during training. Our experiments demonstrate that LIRE consistently outperforms existing methods across several benchmarks on dialogue and summarization tasks, with good transferability to out-of-distribution data, assessed using proxy reward models and human annotators. | [
"Zhu, Mingye",
"Liu, Yi",
"Zhang, Lei",
"Guo, Junbo",
"Mao, Zhendong"
]
| {LIRE}: listwise reward enhancement for preference alignment | findings-acl.201 | Poster | 2402.09764v3 |
https://aclanthology.org/2024.findings-acl.202.bib | @inproceedings{kim-etal-2024-see,
title = "See It All: Contextualized Late Aggregation for 3{D} Dense Captioning",
author = "Kim, Minjung and
Lim, Hyung and
Kim, Seung Hwan and
Lee, Soonyoung and
Kim, Bumsoo and
Kim, Gunhee",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.202",
pages = "3395--3405",
abstract = "3D dense captioning is a task to localize objects in a 3D scene and generate descriptive sentences for each object. Recent approaches in 3D dense captioning have adopted transformer encoder-decoder frameworks from object detection to build an end-to-end pipeline without hand-crafted components. However, these approaches struggle with contradicting objectives where a single query attention has to simultaneously view both the tightly localized object regions and contextual environment. To overcome this challenge, we introduce SIA (See-It-All), a transformer pipeline that engages in 3D dense captioning with a novel paradigm called late aggregation. SIA simultaneously decodes two sets of queries{---}context query and instance query. The instance query focuses on localization and object attribute descriptions, while the context query versatilely captures the region-of-interest of relationships between multiple objects or with the global scene, then aggregated afterwards (i.e., late aggregation) via simple distance-based measures. To further enhance the quality of contextualized caption generation, we design a novel aggregator to generate a fully informed caption based on the surrounding context, the global environment, and object instances. Extensive experiments on two of the most widely-used 3D dense captioning datasets demonstrate that our proposed method achieves a significant improvement over prior methods.",
}
| 3D dense captioning is a task to localize objects in a 3D scene and generate descriptive sentences for each object. Recent approaches in 3D dense captioning have adopted transformer encoder-decoder frameworks from object detection to build an end-to-end pipeline without hand-crafted components. However, these approaches struggle with contradicting objectives where a single query attention has to simultaneously view both the tightly localized object regions and contextual environment. To overcome this challenge, we introduce SIA (See-It-All), a transformer pipeline that engages in 3D dense captioning with a novel paradigm called late aggregation. SIA simultaneously decodes two sets of queries{---}context query and instance query. The instance query focuses on localization and object attribute descriptions, while the context query versatilely captures the region-of-interest of relationships between multiple objects or with the global scene, then aggregated afterwards (i.e., late aggregation) via simple distance-based measures. To further enhance the quality of contextualized caption generation, we design a novel aggregator to generate a fully informed caption based on the surrounding context, the global environment, and object instances. Extensive experiments on two of the most widely-used 3D dense captioning datasets demonstrate that our proposed method achieves a significant improvement over prior methods. | [
"Kim, Minjung",
"Lim, Hyung",
"Kim, Seung Hwan",
"Lee, Soonyoung",
"Kim, Bumsoo",
"Kim, Gunhee"
]
| See It All: Contextualized Late Aggregation for 3{D} Dense Captioning | findings-acl.202 | Poster | 2408.06662v1 |
https://aclanthology.org/2024.findings-acl.203.bib | @inproceedings{fang-etal-2024-dara,
title = "$\texttt{DARA}$: Decomposition-Alignment-Reasoning Autonomous Language Agent for Question Answering over Knowledge Graphs",
author = "Fang, Haishuo and
Zhu, Xiaodan and
Gurevych, Iryna",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.203",
pages = "3406--3432",
abstract = "Answering Questions over Knowledge Graphs (KGQA) is key to well-functioning autonomous language agents in various real-life applications. To improve the neural-symbolic reasoning capabilities of language agents powered by Large Language Models (LLMs) in KGQA, we propose the Decomposition-Alignment-Reasoning Agent (DARA) framework. DARA effectively parses questions into formal queries through a dual mechanism: high-level iterative task decomposition and low-level task grounding. Importantly, DARA can be efficiently trained with a small number of high-quality reasoning trajectories. Our experimental results demonstrate that DARA fine-tuned on LLMs (e.g. Llama-2-7B, Mistral) outperforms both in-context learning-based agents with GPT-4 and alternative fine-tuned agents, across different benchmarks, making such models more accessible for real-life applications. We also show that DARA attains performance comparable to state-of-the-art enumerating-and-ranking-based methods for KGQA.",
}
| Answering Questions over Knowledge Graphs (KGQA) is key to well-functioning autonomous language agents in various real-life applications. To improve the neural-symbolic reasoning capabilities of language agents powered by Large Language Models (LLMs) in KGQA, we propose the Decomposition-Alignment-Reasoning Agent (DARA) framework. DARA effectively parses questions into formal queries through a dual mechanism: high-level iterative task decomposition and low-level task grounding. Importantly, DARA can be efficiently trained with a small number of high-quality reasoning trajectories. Our experimental results demonstrate that DARA fine-tuned on LLMs (e.g. Llama-2-7B, Mistral) outperforms both in-context learning-based agents with GPT-4 and alternative fine-tuned agents, across different benchmarks, making such models more accessible for real-life applications. We also show that DARA attains performance comparable to state-of-the-art enumerating-and-ranking-based methods for KGQA. | [
"Fang, Haishuo",
"Zhu, Xiaodan",
"Gurevych, Iryna"
]
| $\texttt{DARA}$: Decomposition-Alignment-Reasoning Autonomous Language Agent for Question Answering over Knowledge Graphs | findings-acl.203 | Poster | 2406.07080 |
https://aclanthology.org/2024.findings-acl.204.bib | @inproceedings{yao-etal-2024-gkt,
title = "{GKT}: A Novel Guidance-Based Knowledge Transfer Framework For Efficient Cloud-edge Collaboration {LLM} Deployment",
author = "Yao, Yao and
Li, Zuchao and
Zhao, Hai",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.204",
pages = "3433--3446",
abstract = "The burgeoning size of Large Language Models (LLMs) has led to enhanced capabilities in generating responses, albeit at the expense of increased inference times and elevated resource demands. Existing methods of acceleration, predominantly hinged on knowledge distillation, generally necessitate fine-tuning of considerably large models, such as Llama-7B, posing a challenge for average users. Furthermore, present techniques for expediting inference and reducing costs operate independently. To address these issues, we introduce a novel and intuitive Guidance-based Knowledge Transfer (GKT) framework. This approach leverages a larger LLM as a {''}teacher{''} to create guidance prompts, paired with a smaller {''}student{''} model to finalize responses. Remarkably, GKT requires no fine-tuning and doesn{'}t necessitate the teacher and student models to have the same vocabulary, allowing for extensive batch generation to accelerate the process while ensuring user customization. GKT can be seamlessly integrated into cloud-edge collaboration architectures, and is versatile enough for plug-and-play application across various models. It excels in both efficiency and affordability, epitomizing a {''}cheap and cheerful{''} solution. GKT achieves a maximum accuracy improvement of 14.18{\%}, along with a 10.72 times speed-up on GSM8K and an accuracy improvement of 14.00 {\%} along with a 7.73 times speed-up in CSQA. When utilizing ChatGPT as teacher model and Llama2-70B as the student model, we can achieve 95.00{\%} of ChatGPT{'}s performance at 52{\%} of the cost. The results highlight substantial enhancements in accuracy and processing speed on the GSM8K and CSQA datasets, surpassing the performance of using either the student or teacher models in isolation.",
}
| The burgeoning size of Large Language Models (LLMs) has led to enhanced capabilities in generating responses, albeit at the expense of increased inference times and elevated resource demands. Existing methods of acceleration, predominantly hinged on knowledge distillation, generally necessitate fine-tuning of considerably large models, such as Llama-7B, posing a challenge for average users. Furthermore, present techniques for expediting inference and reducing costs operate independently. To address these issues, we introduce a novel and intuitive Guidance-based Knowledge Transfer (GKT) framework. This approach leverages a larger LLM as a {''}teacher{''} to create guidance prompts, paired with a smaller {''}student{''} model to finalize responses. Remarkably, GKT requires no fine-tuning and doesn{'}t necessitate the teacher and student models to have the same vocabulary, allowing for extensive batch generation to accelerate the process while ensuring user customization. GKT can be seamlessly integrated into cloud-edge collaboration architectures, and is versatile enough for plug-and-play application across various models. It excels in both efficiency and affordability, epitomizing a {''}cheap and cheerful{''} solution. GKT achieves a maximum accuracy improvement of 14.18{\%}, along with a 10.72 times speed-up on GSM8K and an accuracy improvement of 14.00 {\%} along with a 7.73 times speed-up in CSQA. When utilizing ChatGPT as teacher model and Llama2-70B as the student model, we can achieve 95.00{\%} of ChatGPT{'}s performance at 52{\%} of the cost. The results highlight substantial enhancements in accuracy and processing speed on the GSM8K and CSQA datasets, surpassing the performance of using either the student or teacher models in isolation. | [
"Yao, Yao",
"Li, Zuchao",
"Zhao, Hai"
]
| {GKT}: A Novel Guidance-Based Knowledge Transfer Framework For Efficient Cloud-edge Collaboration {LLM} Deployment | findings-acl.204 | Poster | 2405.19635v1 |
https://aclanthology.org/2024.findings-acl.205.bib | @inproceedings{wold-etal-2024-compositional,
title = "Compositional Generalization with Grounded Language Models",
author = "Wold, Sondre and
Simon, {\'E}tienne and
Charpentier, Lucas and
Kostylev, Egor and
Velldal, Erik and
{\O}vrelid, Lilja",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.205",
pages = "3447--3460",
abstract = "Grounded language models use external sources of information, such as knowledge graphs, to meet some of the general challenges associated with pre-training. By extending previous work on compositional generalization in semantic parsing, we allow for a controlled evaluation of the degree to which these models learn and generalize from patterns in knowledge graphs. We develop a procedure for generating natural language questions paired with knowledge graphs that targets different aspects of compositionality and further avoids grounding the language models in information already encoded implicitly in their weights. We evaluate existing methods for combining language models with knowledge graphs and find them to struggle with generalization to sequences of unseen lengths and to novel combinations of seen base components. While our experimental results provide some insight into the expressive power of these models, we hope our work and released datasets motivate future research on how to better combine language models with structured knowledge representations.",
}
| Grounded language models use external sources of information, such as knowledge graphs, to meet some of the general challenges associated with pre-training. By extending previous work on compositional generalization in semantic parsing, we allow for a controlled evaluation of the degree to which these models learn and generalize from patterns in knowledge graphs. We develop a procedure for generating natural language questions paired with knowledge graphs that targets different aspects of compositionality and further avoids grounding the language models in information already encoded implicitly in their weights. We evaluate existing methods for combining language models with knowledge graphs and find them to struggle with generalization to sequences of unseen lengths and to novel combinations of seen base components. While our experimental results provide some insight into the expressive power of these models, we hope our work and released datasets motivate future research on how to better combine language models with structured knowledge representations. | [
"Wold, Sondre",
"Simon, {\\'E}tienne",
"Charpentier, Lucas",
"Kostylev, Egor",
"Velldal, Erik",
"{\\O}vrelid, Lilja"
]
| Compositional Generalization with Grounded Language Models | findings-acl.205 | Poster | 2111.03642v1 |
https://aclanthology.org/2024.findings-acl.206.bib | @inproceedings{ding-etal-2024-rethinking,
title = "Rethinking Negative Instances for Generative Named Entity Recognition",
author = "Ding, Yuyang and
Li, Juntao and
Wang, Pinzheng and
Tang, Zecheng and
Bowen, Yan and
Zhang, Min",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.206",
pages = "3461--3475",
abstract = "Large Language Models (LLMs) have demonstrated impressive capabilities for generalizing in unseen tasks. In the Named Entity Recognition (NER) task, recent advancements have seen the remarkable improvement of LLMs in a broad range of entity domains via instruction tuning, by adopting entity-centric schema. In this work, we explore the potential enhancement of the existing methods by incorporating negative instances into training. Our experiments reveal that negative instances contribute to remarkable improvements by (1) introducing contextual information, and (2) clearly delineating label boundaries. Furthermore, we introduce an efficient longest common subsequence (LCS) matching algorithm, which is tailored to transform unstructured predictions into structured entities. By integrating these components, we present GNER, a Generative NER system that shows improved zero-shot performance across unseen entity domains. Our comprehensive evaluation illustrates our system{'}s superiority, surpassing state-of-the-art (SoTA) methods by 9 $F_1$ score in zero-shot evaluation.",
}
| Large Language Models (LLMs) have demonstrated impressive capabilities for generalizing in unseen tasks. In the Named Entity Recognition (NER) task, recent advancements have seen the remarkable improvement of LLMs in a broad range of entity domains via instruction tuning, by adopting entity-centric schema. In this work, we explore the potential enhancement of the existing methods by incorporating negative instances into training. Our experiments reveal that negative instances contribute to remarkable improvements by (1) introducing contextual information, and (2) clearly delineating label boundaries. Furthermore, we introduce an efficient longest common subsequence (LCS) matching algorithm, which is tailored to transform unstructured predictions into structured entities. By integrating these components, we present GNER, a Generative NER system that shows improved zero-shot performance across unseen entity domains. Our comprehensive evaluation illustrates our system{'}s superiority, surpassing state-of-the-art (SoTA) methods by 9 $F_1$ score in zero-shot evaluation. | [
"Ding, Yuyang",
"Li, Juntao",
"Wang, Pinzheng",
"Tang, Zecheng",
"Bowen, Yan",
"Zhang, Min"
]
| Rethinking Negative Instances for Generative Named Entity Recognition | findings-acl.206 | Poster | 2402.16602v2 |
https://aclanthology.org/2024.findings-acl.207.bib | @inproceedings{hu-etal-2024-wilke,
title = "{W}il{KE}: Wise-Layer Knowledge Editor for Lifelong Knowledge Editing",
author = "Hu, Chenhui and
Cao, Pengfei and
Chen, Yubo and
Liu, Kang and
Zhao, Jun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.207",
pages = "3476--3503",
abstract = "Knowledge editing aims to rectify inaccuracies in large language models (LLMs) without costly retraining for outdated or erroneous knowledge. However, current knowledge editing methods primarily focus on single editing, failing to meet the requirements for lifelong editing. This study reveals a performance degradation encountered by knowledge editing in lifelong editing, characterized by toxicity buildup and toxicity flash, with the primary cause identified as pattern unmatch. We introduce a knowledge editing approach named Wise-Layer Knowledge Editor (WilKE), which selects editing layer based on the pattern matching degree of editing knowledge across different layers in language models. Experimental results demonstrate that, in lifelong editing, WilKE exhibits an average improvement of 46.2{\%} and 67.8{\%} on editing GPT2-XL and GPT-J relative to state-of-the-art knowledge editing methods.",
}
| Knowledge editing aims to rectify inaccuracies in large language models (LLMs) without costly retraining for outdated or erroneous knowledge. However, current knowledge editing methods primarily focus on single editing, failing to meet the requirements for lifelong editing. This study reveals a performance degradation encountered by knowledge editing in lifelong editing, characterized by toxicity buildup and toxicity flash, with the primary cause identified as pattern unmatch. We introduce a knowledge editing approach named Wise-Layer Knowledge Editor (WilKE), which selects editing layer based on the pattern matching degree of editing knowledge across different layers in language models. Experimental results demonstrate that, in lifelong editing, WilKE exhibits an average improvement of 46.2{\%} and 67.8{\%} on editing GPT2-XL and GPT-J relative to state-of-the-art knowledge editing methods. | [
"Hu, Chenhui",
"Cao, Pengfei",
"Chen, Yubo",
"Liu, Kang",
"Zhao, Jun"
]
| {W}il{KE}: Wise-Layer Knowledge Editor for Lifelong Knowledge Editing | findings-acl.207 | Poster | 2402.10987v2 |
https://aclanthology.org/2024.findings-acl.208.bib | @inproceedings{wu-etal-2024-diner,
title = "{DINER}: Debiasing Aspect-based Sentiment Analysis with Multi-variable Causal Inference",
author = "Wu, Jialong and
Zhang, Linhai and
Zhou, Deyu and
Xu, Guoqiang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.208",
pages = "3504--3518",
abstract = "Though notable progress has been made, neural-based aspect-based sentiment analysis (ABSA) models are prone to learn spurious correlations from annotation biases, resulting in poor robustness on adversarial data transformations. Among the debiasing solutions, causal inference-based methods have attracted much research attention, which can be mainly categorized into causal intervention methods and counterfactual reasoning methods. However, most of the present debiasing methods focus on single-variable causal inference, which is not suitable for ABSA with two input variables (the target aspect and the review). In this paper, we propose a novel framework based on multi-variable causal inference for debiasing ABSA. In this framework, different types of biases are tackled based on different causal intervention methods. For the review branch, the bias is modeled as indirect confounding from context, where backdoor adjustment intervention is employed for debiasing. For the aspect branch, the bias is described as a direct correlation with labels, where counterfactual reasoning is adopted for debiasing. Extensive experiments demonstrate the effectiveness of the proposed method compared to various baselines on the two widely used real-world aspect robustness test set datasets.",
}
| Though notable progress has been made, neural-based aspect-based sentiment analysis (ABSA) models are prone to learn spurious correlations from annotation biases, resulting in poor robustness on adversarial data transformations. Among the debiasing solutions, causal inference-based methods have attracted much research attention, which can be mainly categorized into causal intervention methods and counterfactual reasoning methods. However, most of the present debiasing methods focus on single-variable causal inference, which is not suitable for ABSA with two input variables (the target aspect and the review). In this paper, we propose a novel framework based on multi-variable causal inference for debiasing ABSA. In this framework, different types of biases are tackled based on different causal intervention methods. For the review branch, the bias is modeled as indirect confounding from context, where backdoor adjustment intervention is employed for debiasing. For the aspect branch, the bias is described as a direct correlation with labels, where counterfactual reasoning is adopted for debiasing. Extensive experiments demonstrate the effectiveness of the proposed method compared to various baselines on the two widely used real-world aspect robustness test set datasets. | [
"Wu, Jialong",
"Zhang, Linhai",
"Zhou, Deyu",
"Xu, Guoqiang"
]
| {DINER}: Debiasing Aspect-based Sentiment Analysis with Multi-variable Causal Inference | findings-acl.208 | Poster | 2404.11055v1 |
https://aclanthology.org/2024.findings-acl.209.bib | @inproceedings{zhang-etal-2024-star,
title = "{STAR}: Constraint {L}o{RA} with Dynamic Active Learning for Data-Efficient Fine-Tuning of Large Language Models",
author = "Zhang, Linhai and
Wu, Jialong and
Zhou, Deyu and
Xu, Guoqiang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.209",
pages = "3519--3532",
abstract = "Though Large Language Models (LLMs) have demonstrated the powerful capabilities of few-shot learning through prompting methods, supervised training is still necessary for complex reasoning tasks. Because of their extensive parameters and memory consumption, both Parameter-Efficient Fine-Tuning (PEFT) methods and Memory-Efficient Fine-Tuning methods have been proposed for LLMs. Nevertheless, the issue of large annotated data consumption, the aim of Data-Efficient Fine-Tuning, remains unexplored. One obvious way is to combine the PEFT method with active learning. However, the experimental results show that such a combination is not trivial and yields inferior results. Through probe experiments, such observation might be explained by two main reasons: uncertainty gap and poor model calibration. Therefore, in this paper, we propose a novel approach to effectively integrate uncertainty-based active learning and LoRA. Specifically, for the uncertainty gap, we introduce a dynamic uncertainty measurement that combines the uncertainty of the base model and the uncertainty of the full model during the iteration of active learning. For poor model calibration, we incorporate the regularization method during LoRA training to keep the model from being over-confident, and the Monte-Carlo dropout mechanism is employed to enhance the uncertainty estimation. Experimental results show that the proposed approach outperforms existing baseline models on three complex reasoning tasks.",
}
| Though Large Language Models (LLMs) have demonstrated the powerful capabilities of few-shot learning through prompting methods, supervised training is still necessary for complex reasoning tasks. Because of their extensive parameters and memory consumption, both Parameter-Efficient Fine-Tuning (PEFT) methods and Memory-Efficient Fine-Tuning methods have been proposed for LLMs. Nevertheless, the issue of large annotated data consumption, the aim of Data-Efficient Fine-Tuning, remains unexplored. One obvious way is to combine the PEFT method with active learning. However, the experimental results show that such a combination is not trivial and yields inferior results. Through probe experiments, such observation might be explained by two main reasons: uncertainty gap and poor model calibration. Therefore, in this paper, we propose a novel approach to effectively integrate uncertainty-based active learning and LoRA. Specifically, for the uncertainty gap, we introduce a dynamic uncertainty measurement that combines the uncertainty of the base model and the uncertainty of the full model during the iteration of active learning. For poor model calibration, we incorporate the regularization method during LoRA training to keep the model from being over-confident, and the Monte-Carlo dropout mechanism is employed to enhance the uncertainty estimation. Experimental results show that the proposed approach outperforms existing baseline models on three complex reasoning tasks. | [
"Zhang, Linhai",
"Wu, Jialong",
"Zhou, Deyu",
"Xu, Guoqiang"
]
| {STAR}: Constraint {L}o{RA} with Dynamic Active Learning for Data-Efficient Fine-Tuning of Large Language Models | findings-acl.209 | Poster | 2402.11455v1 |
https://aclanthology.org/2024.findings-acl.210.bib | @inproceedings{wang-etal-2024-much,
title = "How Much Does Nonverbal Communication Conform to Entropy Rate Constancy?: A Case Study on Listener Gaze in Interaction",
author = "Wang, Yu and
Xu, Yang and
Skantze, Gabriel and
Buschmeier, Hendrik",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.210",
pages = "3533--3545",
abstract = "According to the Entropy Rate Constancy (ERC) principle, the information density of a text is approximately constant over its length. Whether this principle also applies to nonverbal communication signals is still under investigation. We perform empirical analyses of video-recorded dialogue data and investigate whether listener gaze, as an important nonverbal communication signal, adheres to the ERC principle. Results show (1) that the ERC principle holds for listener gaze; and (2) that the two linguistic factors syntactic complexity and turn transition potential are weakly correlated with local entropy of listener gaze.",
}
| According to the Entropy Rate Constancy (ERC) principle, the information density of a text is approximately constant over its length. Whether this principle also applies to nonverbal communication signals is still under investigation. We perform empirical analyses of video-recorded dialogue data and investigate whether listener gaze, as an important nonverbal communication signal, adheres to the ERC principle. Results show (1) that the ERC principle holds for listener gaze; and (2) that the two linguistic factors syntactic complexity and turn transition potential are weakly correlated with local entropy of listener gaze. | [
"Wang, Yu",
"Xu, Yang",
"Skantze, Gabriel",
"Buschmeier, Hendrik"
]
| How Much Does Nonverbal Communication Conform to Entropy Rate Constancy?: A Case Study on Listener Gaze in Interaction | findings-acl.210 | Poster | null |
https://aclanthology.org/2024.findings-acl.211.bib | @inproceedings{huang-etal-2024-lost,
title = "Lost in the Source Language: How Large Language Models Evaluate the Quality of Machine Translation",
author = "Huang, Xu and
Zhang, Zhirui and
Geng, Xiang and
Du, Yichao and
Chen, Jiajun and
Huang, Shujian",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.211",
pages = "3546--3562",
abstract = "This study investigates how Large Language Models (LLMs) leverage source and reference data in machine translation evaluation task, aiming to better understand the mechanisms behind their remarkable performance in this task.We design the controlled experiments across various input modes and model types, and employ both coarse-grained and fine-grained prompts to discern the utility of source versus reference information.We find that reference information significantly enhances the evaluation accuracy, while surprisingly, source information sometimes is counterproductive, indicating LLMs{'} inability to fully leverage the cross-lingual capability when evaluating translations.Further analysis of the fine-grained evaluation and fine-tuning experiments show similar results.These findings also suggest a potential research direction for LLMs that fully exploits the cross-lingual capability of LLMs to achieve better performance in machine translation evaluation tasks.",
}
| This study investigates how Large Language Models (LLMs) leverage source and reference data in machine translation evaluation task, aiming to better understand the mechanisms behind their remarkable performance in this task.We design the controlled experiments across various input modes and model types, and employ both coarse-grained and fine-grained prompts to discern the utility of source versus reference information.We find that reference information significantly enhances the evaluation accuracy, while surprisingly, source information sometimes is counterproductive, indicating LLMs{'} inability to fully leverage the cross-lingual capability when evaluating translations.Further analysis of the fine-grained evaluation and fine-tuning experiments show similar results.These findings also suggest a potential research direction for LLMs that fully exploits the cross-lingual capability of LLMs to achieve better performance in machine translation evaluation tasks. | [
"Huang, Xu",
"Zhang, Zhirui",
"Geng, Xiang",
"Du, Yichao",
"Chen, Jiajun",
"Huang, Shujian"
]
| Lost in the Source Language: How Large Language Models Evaluate the Quality of Machine Translation | findings-acl.211 | Poster | 2401.06568v2 |
https://aclanthology.org/2024.findings-acl.212.bib | @inproceedings{dhuliawala-etal-2024-chain,
title = "Chain-of-Verification Reduces Hallucination in Large Language Models",
author = "Dhuliawala, Shehzaad and
Komeili, Mojtaba and
Xu, Jing and
Raileanu, Roberta and
Li, Xian and
Celikyilmaz, Asli and
Weston, Jason",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.212",
pages = "3563--3578",
abstract = "Generation of plausible yet incorrect factual information, termed hallucination, is an unsolved issue in large language models. We study the ability of language models to deliberate on the responses they give in order to correct their mistakes. We develop the Chain-of-Verification (CoVe) method whereby the model first (i) drafts an initial response; then (ii) plans verification questions to fact-check its draft; (iii) answers those questions independently so the answers are not biased by other responses; and (iv) generates its final verified response. In experiments, we show CoVe decreases hallucinations across a variety of tasks, from list-based questions from Wikidata, closed book MultiSpanQA and longform text generation.",
}
| Generation of plausible yet incorrect factual information, termed hallucination, is an unsolved issue in large language models. We study the ability of language models to deliberate on the responses they give in order to correct their mistakes. We develop the Chain-of-Verification (CoVe) method whereby the model first (i) drafts an initial response; then (ii) plans verification questions to fact-check its draft; (iii) answers those questions independently so the answers are not biased by other responses; and (iv) generates its final verified response. In experiments, we show CoVe decreases hallucinations across a variety of tasks, from list-based questions from Wikidata, closed book MultiSpanQA and longform text generation. | [
"Dhuliawala, Shehzaad",
"Komeili, Mojtaba",
"Xu, Jing",
"Raileanu, Roberta",
"Li, Xian",
"Celikyilmaz, Asli",
"Weston, Jason"
]
| Chain-of-Verification Reduces Hallucination in Large Language Models | findings-acl.212 | Poster | 2404.01588v1 |
https://aclanthology.org/2024.findings-acl.213.bib | @inproceedings{xia-etal-2024-measuring,
title = "Measuring Bargaining Abilities of {LLM}s: A Benchmark and A Buyer-Enhancement Method",
author = "Xia, Tian and
He, Zhiwei and
Ren, Tong and
Miao, Yibo and
Zhang, Zhuosheng and
Yang, Yang and
Wang, Rui",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.213",
pages = "3579--3602",
abstract = "Bargaining is an important and unique part of negotiation between humans. As LLM-driven agents learn to negotiate and act like real humans, how to evaluate agents{'} bargaining abilities remains an open problem.For the first time, we formally described the Bargaining task as an asymmetric incomplete information game, defining the gains of the Buyer and Seller in multiple bargaining processes. It allows us to quantitatively assess an agent{'}s performance in the Bargain task.We collected a real product price dataset, AmazonHistoryPrice, and conducted evaluations of various LLM agents{'} bargaining abilities. We find that playing a Buyer is much harder than a Seller, and increasing model size can not effectively improve the Buyer{'}s performance.To address the challenge, we propose a novel approach called OG-Narrator that integrates a deterministic Offer Generator to control the price range of Buyer{'}s offers, and an LLM Narrator to create natural language sentences for generated offers.Experimental results show that OG-Narrator improves the buyer{'}s deal rates from 26.67{\%} to 88.88{\%} and brings a ten times multiplication of profits on all baselines, even a model that has not been aligned.",
}
| Bargaining is an important and unique part of negotiation between humans. As LLM-driven agents learn to negotiate and act like real humans, how to evaluate agents{'} bargaining abilities remains an open problem.For the first time, we formally described the Bargaining task as an asymmetric incomplete information game, defining the gains of the Buyer and Seller in multiple bargaining processes. It allows us to quantitatively assess an agent{'}s performance in the Bargain task.We collected a real product price dataset, AmazonHistoryPrice, and conducted evaluations of various LLM agents{'} bargaining abilities. We find that playing a Buyer is much harder than a Seller, and increasing model size can not effectively improve the Buyer{'}s performance.To address the challenge, we propose a novel approach called OG-Narrator that integrates a deterministic Offer Generator to control the price range of Buyer{'}s offers, and an LLM Narrator to create natural language sentences for generated offers.Experimental results show that OG-Narrator improves the buyer{'}s deal rates from 26.67{\%} to 88.88{\%} and brings a ten times multiplication of profits on all baselines, even a model that has not been aligned. | [
"Xia, Tian",
"He, Zhiwei",
"Ren, Tong",
"Miao, Yibo",
"Zhang, Zhuosheng",
"Yang, Yang",
"Wang, Rui"
]
| Measuring Bargaining Abilities of {LLM}s: A Benchmark and A Buyer-Enhancement Method | findings-acl.213 | Poster | 2402.15813v3 |
https://aclanthology.org/2024.findings-acl.214.bib | @inproceedings{li-etal-2024-deveval,
title = "{D}ev{E}val: A Manually-Annotated Code Generation Benchmark Aligned with Real-World Code Repositories",
author = "Li, Jia and
Li, Ge and
Zhao, Yunfei and
Li, Yongmin and
Liu, Huanyu and
Zhu, Hao and
Wang, Lecheng and
Liu, Kaibo and
Fang, Zheng and
Wang, Lanshen and
Ding, Jiazheng and
Zhang, Xuanming and
Zhu, Yuqi and
Dong, Yihong and
Jin, Zhi and
Li, Binhua and
Huang, Fei and
Li, Yongbin and
Gu, Bin and
Yang, Mengfei",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.214",
pages = "3603--3614",
abstract = "How to evaluate the coding abilities of Large Language Models (LLMs) remains an open question. We find that existing benchmarks are poorly aligned with real-world code repositories and are insufficient to evaluate the coding abilities of LLMs.To address the knowledge gap, we propose a new benchmark named DevEval, which has three advances. (1) DevEval aligns with real-world repositories in multiple dimensions, e.g., code and dependency distributions. (2) DevEval is annotated by 13 developers and contains comprehensive annotations (e.g., requirements, original repositories, reference code, and reference dependencies). (3) DevEval comprises 1,825 testing samples from 115 repositories, covering 10 popular domains (e.g., Internet, Database). Based on DevEval, we propose repository-level code generation and evaluate 8 popular LLMs on DevEval (e.g., gpt-4, gpt-3.5, StarCoder 2, DeepSeek Coder, CodeLLaMa). Our experiments reveal these LLMs{'} coding abilities in real-world code repositories. For example, the highest Pass@1 of gpt-4 only is 53.04{\%} in our experiments. We also analyze LLMs{'} failed cases and summarize their shortcomings. We hope DevEval can facilitate the development of LLMs in real code repositories. DevEval, prompts, and LLMs{'} predictions have been released.",
}
| How to evaluate the coding abilities of Large Language Models (LLMs) remains an open question. We find that existing benchmarks are poorly aligned with real-world code repositories and are insufficient to evaluate the coding abilities of LLMs.To address the knowledge gap, we propose a new benchmark named DevEval, which has three advances. (1) DevEval aligns with real-world repositories in multiple dimensions, e.g., code and dependency distributions. (2) DevEval is annotated by 13 developers and contains comprehensive annotations (e.g., requirements, original repositories, reference code, and reference dependencies). (3) DevEval comprises 1,825 testing samples from 115 repositories, covering 10 popular domains (e.g., Internet, Database). Based on DevEval, we propose repository-level code generation and evaluate 8 popular LLMs on DevEval (e.g., gpt-4, gpt-3.5, StarCoder 2, DeepSeek Coder, CodeLLaMa). Our experiments reveal these LLMs{'} coding abilities in real-world code repositories. For example, the highest Pass@1 of gpt-4 only is 53.04{\%} in our experiments. We also analyze LLMs{'} failed cases and summarize their shortcomings. We hope DevEval can facilitate the development of LLMs in real code repositories. DevEval, prompts, and LLMs{'} predictions have been released. | [
"Li, Jia",
"Li, Ge",
"Zhao, Yunfei",
"Li, Yongmin",
"Liu, Huanyu",
"Zhu, Hao",
"Wang, Lecheng",
"Liu, Kaibo",
"Fang, Zheng",
"Wang, Lanshen",
"Ding, Jiazheng",
"Zhang, Xuanming",
"Zhu, Yuqi",
"Dong, Yihong",
"Jin, Zhi",
"Li, Binhua",
"Huang, Fei",
"Li, Yongbin",
"Gu, Bin",
"Yang, Mengfei"
]
| {D}ev{E}val: A Manually-Annotated Code Generation Benchmark Aligned with Real-World Code Repositories | findings-acl.214 | Poster | 2405.19856v1 |
https://aclanthology.org/2024.findings-acl.215.bib | @inproceedings{bi-etal-2024-lpnl,
title = "{LPNL}: Scalable Link Prediction with Large Language Models",
author = "Bi, Baolong and
Liu, Shenghua and
Wang, Yiwei and
Mei, Lingrui and
Cheng, Xueqi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.215",
pages = "3615--3625",
abstract = "Exploring the application of large language models (LLMs) to graph learning is an emerging endeavor. However, the vast amount of information inherent in large graphs poses significant challenges to graph learning with LLMs. This work focuses on the link prediction task and introduces **LPNL** (Link Prediction via Natural Language), a framework based on large language models designed for scalable link prediction on large-scale heterogeneous graphs. We design novel prompts for link prediction that articulate graph details in natural language. We propose a two-stage sampling pipeline to extract crucial information from the graphs, and a divide-and-conquer strategy to control the input tokens within predefined limits, addressing the challenge of overwhelming information. We fine-tune a T5 model based on our self-supervised learning designed for link prediction. Extensive experimental results demonstrate that LPNL outperforms multiple advanced baselines in link prediction tasks on large-scale graphs.",
}
| Exploring the application of large language models (LLMs) to graph learning is an emerging endeavor. However, the vast amount of information inherent in large graphs poses significant challenges to graph learning with LLMs. This work focuses on the link prediction task and introduces **LPNL** (Link Prediction via Natural Language), a framework based on large language models designed for scalable link prediction on large-scale heterogeneous graphs. We design novel prompts for link prediction that articulate graph details in natural language. We propose a two-stage sampling pipeline to extract crucial information from the graphs, and a divide-and-conquer strategy to control the input tokens within predefined limits, addressing the challenge of overwhelming information. We fine-tune a T5 model based on our self-supervised learning designed for link prediction. Extensive experimental results demonstrate that LPNL outperforms multiple advanced baselines in link prediction tasks on large-scale graphs. | [
"Bi, Baolong",
"Liu, Shenghua",
"Wang, Yiwei",
"Mei, Lingrui",
"Cheng, Xueqi"
]
| {LPNL}: Scalable Link Prediction with Large Language Models | findings-acl.215 | Poster | 2401.13227v3 |
https://aclanthology.org/2024.findings-acl.216.bib | @inproceedings{heffernan-etal-2024-aligning,
title = "Aligning Speech Segments Beyond Pure Semantics",
author = "Heffernan, Kevin and
Kozhevnikov, Artyom and
Barrault, Loic and
Mourachko, Alexandre and
Schwenk, Holger",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.216",
pages = "3626--3635",
abstract = "Multilingual parallel data for speech-to-speech translation is scarce and expensive to create from scratch. This is all the more true for expressive speech translation, which aims at preserving not only the semantics, but also the overall prosody (e.g. style, emotion, rate-of-speech). Existing corpora contain speech utterances with the same meaning, yet the overall prosody is typically different, as human annotators are not tasked with reproducing these aspects, or crowed-sourced efforts do not specifically target this kind of alignment in priority. In this paper, we propose a novel alignment algorithm, which automatically forms pairs of speech segments aligned not only in meaning, but also in expressivity. In order to validate our approach, we train an expressive multilingual speech-to-speech translation system on the automatically aligned data. Our experiments show that in comparison to semantic-only approaches, expressively aligned data yields large improvements in source expressivity preservation (e.g. 43{\%} uplift in speech rate preservation on average), while still maintaining content translation quality. In some scenarios, results also indicate that this alignment algorithm can outperform standard, semantic-focused approaches even on content translation quality.",
}
| Multilingual parallel data for speech-to-speech translation is scarce and expensive to create from scratch. This is all the more true for expressive speech translation, which aims at preserving not only the semantics, but also the overall prosody (e.g. style, emotion, rate-of-speech). Existing corpora contain speech utterances with the same meaning, yet the overall prosody is typically different, as human annotators are not tasked with reproducing these aspects, or crowed-sourced efforts do not specifically target this kind of alignment in priority. In this paper, we propose a novel alignment algorithm, which automatically forms pairs of speech segments aligned not only in meaning, but also in expressivity. In order to validate our approach, we train an expressive multilingual speech-to-speech translation system on the automatically aligned data. Our experiments show that in comparison to semantic-only approaches, expressively aligned data yields large improvements in source expressivity preservation (e.g. 43{\%} uplift in speech rate preservation on average), while still maintaining content translation quality. In some scenarios, results also indicate that this alignment algorithm can outperform standard, semantic-focused approaches even on content translation quality. | [
"Heffernan, Kevin",
"Kozhevnikov, Artyom",
"Barrault, Loic",
"Mourachko, Alex",
"re",
"Schwenk, Holger"
]
| Aligning Speech Segments Beyond Pure Semantics | findings-acl.216 | Poster | 2309.09294v1 |
https://aclanthology.org/2024.findings-acl.217.bib | @inproceedings{nguyen-etal-2024-video,
title = "Video-Language Understanding: A Survey from Model Architecture, Model Training, and Data Perspectives",
author = "Nguyen, Thong and
Bin, Yi and
Xiao, Junbin and
Qu, Leigang and
Li, Yicong and
Wu, Jay Zhangjie and
Nguyen, Cong-Duy and
Ng, See-Kiong and
Luu, Anh Tuan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.217",
pages = "3636--3657",
abstract = "Humans use multiple senses to comprehend the environment. Vision and language are two of the most vital senses since they allow us to easily communicate our thoughts and perceive the world around us. There has been a lot of interest in creating video-language understanding systems with human-like senses since a video-language pair can mimic both our linguistic medium and visual environment with temporal dynamics. In this survey, we review the key tasks of these systems and highlight the associated challenges. Based on the challenges, we summarize their methods from model architecture, model training, and data perspectives. We also conduct performance comparison among the methods, and discuss promising directions for future research.",
}
| Humans use multiple senses to comprehend the environment. Vision and language are two of the most vital senses since they allow us to easily communicate our thoughts and perceive the world around us. There has been a lot of interest in creating video-language understanding systems with human-like senses since a video-language pair can mimic both our linguistic medium and visual environment with temporal dynamics. In this survey, we review the key tasks of these systems and highlight the associated challenges. Based on the challenges, we summarize their methods from model architecture, model training, and data perspectives. We also conduct performance comparison among the methods, and discuss promising directions for future research. | [
"Nguyen, Thong",
"Bin, Yi",
"Xiao, Junbin",
"Qu, Leigang",
"Li, Yicong",
"Wu, Jay Zhangjie",
"Nguyen, Cong-Duy",
"Ng, See-Kiong",
"Luu, Anh Tuan"
]
| Video-Language Understanding: A Survey from Model Architecture, Model Training, and Data Perspectives | findings-acl.217 | Poster | 2406.05615v2 |
https://aclanthology.org/2024.findings-acl.218.bib | @inproceedings{ding-etal-2024-generative,
title = "Generative Input: Towards Next-Generation Input Methods Paradigm",
author = "Ding, Keyu and
Wang, Yongcan and
Xu, Zihang and
Jia, Zhenzhen and
Chen, Enhong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.218",
pages = "3658--3669",
abstract = "Since the release of ChatGPT, generative models have achieved tremendous success and become the de facto approach for various NLP tasks. However, its application in the field of input methods remains under-explored. Many neural network approaches have been applied to the construction of Chinese input method engines (IMEs). Previous research often assumed that the input pinyin was correct and focused on Pinyin-to-character (P2C) task, which significantly falls short of meeting users{'} demands. Moreover, previous research could not leverage user feedback to optimize the model and provide personalized results. In this study, we propose a novel Generative Input paradigm named GeneInput. It uses prompts to handle all input scenarios and other intelligent auxiliary input functions, optimizing the model with user feedback. The results demonstrate that we have achieved state-of-the-art performance for the first time in the Full-mode Key-sequence to Characters task. GeneInput also includes RLHF-IME, a novel RLHF application framework for input method, that eliminates the need for manual ranking annotations and the performance surpasses GPT-4. Relevant resources have been open-sourced.",
}
| Since the release of ChatGPT, generative models have achieved tremendous success and become the de facto approach for various NLP tasks. However, its application in the field of input methods remains under-explored. Many neural network approaches have been applied to the construction of Chinese input method engines (IMEs). Previous research often assumed that the input pinyin was correct and focused on Pinyin-to-character (P2C) task, which significantly falls short of meeting users{'} demands. Moreover, previous research could not leverage user feedback to optimize the model and provide personalized results. In this study, we propose a novel Generative Input paradigm named GeneInput. It uses prompts to handle all input scenarios and other intelligent auxiliary input functions, optimizing the model with user feedback. The results demonstrate that we have achieved state-of-the-art performance for the first time in the Full-mode Key-sequence to Characters task. GeneInput also includes RLHF-IME, a novel RLHF application framework for input method, that eliminates the need for manual ranking annotations and the performance surpasses GPT-4. Relevant resources have been open-sourced. | [
"Ding, Keyu",
"Wang, Yongcan",
"Xu, Zihang",
"Jia, Zhenzhen",
"Chen, Enhong"
]
| Generative Input: Towards Next-Generation Input Methods Paradigm | findings-acl.218 | Poster | 2406.01921v1 |
https://aclanthology.org/2024.findings-acl.219.bib | @inproceedings{tang-etal-2024-b,
title = "A + {B}: A General Generator-Reader Framework for Optimizing {LLM}s to Unleash Synergy Potential",
author = "Tang, Wei and
Cao, Yixin and
Ying, Jiahao and
Wang, Bo and
Zhao, Yuyue and
Liao, Yong and
Zhou, Peng",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.219",
pages = "3670--3685",
abstract = "Retrieval-Augmented Generation (RAG) is an effective solution to supplement necessary knowledge to large language models (LLMs). Targeting its bottleneck of retriever performance, {``}generate-then-read{''} pipeline is proposed to replace the retrieval stage with generation from the LLM itself. Although promising, this research direction is underexplored and still cannot work in the scenario when source knowledge is given. In this paper, we formalize a general {``}A + B{''} framework with varying combinations of foundation models and types for systematic investigation. We explore the efficacy of the base and chat versions of LLMs and found their different functionalities suitable for generator A and reader B, respectively. Their combinations consistently outperform single models, especially in complex scenarios. Furthermore, we extend the application of the {``}A + B{''} framework to scenarios involving source documents through continuous learning, enabling the direct integration of external knowledge into LLMs. This approach not only facilitates effective acquisition of new knowledge but also addresses the challenges of safety and helpfulness post-adaptation. The paper underscores the versatility of the {``}A + B{''} framework, demonstrating its potential to enhance the practical application of LLMs across various domains.",
}
| Retrieval-Augmented Generation (RAG) is an effective solution to supplement necessary knowledge to large language models (LLMs). Targeting its bottleneck of retriever performance, {``}generate-then-read{''} pipeline is proposed to replace the retrieval stage with generation from the LLM itself. Although promising, this research direction is underexplored and still cannot work in the scenario when source knowledge is given. In this paper, we formalize a general {``}A + B{''} framework with varying combinations of foundation models and types for systematic investigation. We explore the efficacy of the base and chat versions of LLMs and found their different functionalities suitable for generator A and reader B, respectively. Their combinations consistently outperform single models, especially in complex scenarios. Furthermore, we extend the application of the {``}A + B{''} framework to scenarios involving source documents through continuous learning, enabling the direct integration of external knowledge into LLMs. This approach not only facilitates effective acquisition of new knowledge but also addresses the challenges of safety and helpfulness post-adaptation. The paper underscores the versatility of the {``}A + B{''} framework, demonstrating its potential to enhance the practical application of LLMs across various domains. | [
"Tang, Wei",
"Cao, Yixin",
"Ying, Jiahao",
"Wang, Bo",
"Zhao, Yuyue",
"Liao, Yong",
"Zhou, Peng"
]
| A + {B}: A General Generator-Reader Framework for Optimizing {LLM}s to Unleash Synergy Potential | findings-acl.219 | Poster | 2406.03963 |
https://aclanthology.org/2024.findings-acl.220.bib | @inproceedings{to-etal-2024-functional,
title = "Functional Overlap Reranking for Neural Code Generation",
author = "To, Hung and
Nguyen, Minh and
Bui, Nghi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.220",
pages = "3686--3704",
abstract = "Code Large Language Models (CodeLLMs) have ushered in a new era in code generation advancements. However, selecting the best code solutions from all possible CodeLLM outputs remains a challenge. Previous methods often overlooked the intricate functional similarities and interactions between solution clusters. We introduce SRank, a novel reranking strategy for selecting the best solutions from code generation, focusing on modeling the relationships between clusters of solutions. By quantifying the functional overlap between solution clusters, our approach provides a better ranking strategy for code solutions. Empirical results show that our method achieves remarkable results on the pass@1 score. For instance, on the Human-Eval benchmark, we achieve 69.66{\%} in pass@1 with Codex002, 75.31{\%} with WizardCoder, 53.99{\%} with StarCoder, and 60.55{\%} with CodeGen, surpassing state-of-the-art code generation reranking methods such as CodeT and Coder-Reviewer on the same CodeLLM by a significant margin approx 6.1{\%} improvement on average. Even in scenarios with a limited number of sampled solutions and test cases, our approach demonstrates robustness and superiority, marking a new benchmark in code generation reranking. Our implementation can be found at https://github.com/FSoft-AI4Code/SRank-CodeRanker.",
}
| Code Large Language Models (CodeLLMs) have ushered in a new era in code generation advancements. However, selecting the best code solutions from all possible CodeLLM outputs remains a challenge. Previous methods often overlooked the intricate functional similarities and interactions between solution clusters. We introduce SRank, a novel reranking strategy for selecting the best solutions from code generation, focusing on modeling the relationships between clusters of solutions. By quantifying the functional overlap between solution clusters, our approach provides a better ranking strategy for code solutions. Empirical results show that our method achieves remarkable results on the pass@1 score. For instance, on the Human-Eval benchmark, we achieve 69.66{\%} in pass@1 with Codex002, 75.31{\%} with WizardCoder, 53.99{\%} with StarCoder, and 60.55{\%} with CodeGen, surpassing state-of-the-art code generation reranking methods such as CodeT and Coder-Reviewer on the same CodeLLM by a significant margin approx 6.1{\%} improvement on average. Even in scenarios with a limited number of sampled solutions and test cases, our approach demonstrates robustness and superiority, marking a new benchmark in code generation reranking. Our implementation can be found at https://github.com/FSoft-AI4Code/SRank-CodeRanker. | [
"To, Hung",
"Nguyen, Minh",
"Bui, Nghi"
]
| Functional Overlap Reranking for Neural Code Generation | findings-acl.220 | Poster | 2311.03366v4 |
https://aclanthology.org/2024.findings-acl.221.bib | @inproceedings{cheng-etal-2024-adversarial,
title = "Adversarial Preference Optimization: Enhancing Your Alignment via {RM}-{LLM} Game",
author = "Cheng, Pengyu and
Yang, Yifan and
Li, Jian and
Dai, Yong and
Hu, Tianhao and
Cao, Peixin and
Du, Nan and
Li, Xiaolong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.221",
pages = "3705--3716",
abstract = "Human preference alignment is essential to improve the interaction quality of large language models (LLMs). Existing alignment methods depend on manually annotated preference data to guide the LLM optimization directions. However, continuously updating LLMs for alignment raises a distribution gap between model-generated samples and human-annotated responses, hindering training effectiveness. To mitigate this issue, previous methods require additional preference annotation on newly generated samples to adapt to the shifted distribution, which consumes a large amount of annotation resources. Targeting more efficient human preference optimization, we propose an Adversarial Preference Optimization (APO) framework, in which the LLM and the reward model update alternatively via a min-max game. Through adversarial training, the reward model can adapt to the shifted generation distribution of the LLM without any additional annotation. With comprehensive experiments, we find the proposed adversarial training framework further enhances existing alignment baselines in terms of LLM helpfulness and harmlessness. The code is at https://github.com/Linear95/APO.",
}
| Human preference alignment is essential to improve the interaction quality of large language models (LLMs). Existing alignment methods depend on manually annotated preference data to guide the LLM optimization directions. However, continuously updating LLMs for alignment raises a distribution gap between model-generated samples and human-annotated responses, hindering training effectiveness. To mitigate this issue, previous methods require additional preference annotation on newly generated samples to adapt to the shifted distribution, which consumes a large amount of annotation resources. Targeting more efficient human preference optimization, we propose an Adversarial Preference Optimization (APO) framework, in which the LLM and the reward model update alternatively via a min-max game. Through adversarial training, the reward model can adapt to the shifted generation distribution of the LLM without any additional annotation. With comprehensive experiments, we find the proposed adversarial training framework further enhances existing alignment baselines in terms of LLM helpfulness and harmlessness. The code is at https://github.com/Linear95/APO. | [
"Cheng, Pengyu",
"Yang, Yifan",
"Li, Jian",
"Dai, Yong",
"Hu, Tianhao",
"Cao, Peixin",
"Du, Nan",
"Li, Xiaolong"
]
| Adversarial Preference Optimization: Enhancing Your Alignment via {RM}-{LLM} Game | findings-acl.221 | Poster | 2311.08045v4 |
https://aclanthology.org/2024.findings-acl.222.bib | @inproceedings{zhu-etal-2024-pinpointing,
title = "Pinpointing Diffusion Grid Noise to Enhance Aspect Sentiment Quad Prediction",
author = "Zhu, Linan and
Chen, Xiangfan and
Guo, Xiaolei and
Zhang, Chenwei and
Zhu, Zhechao and
Zhou, Zehai and
Kong, Xiangjie",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.222",
pages = "3717--3726",
abstract = "Aspect sentiment quad prediction (ASQP) has garnered significant attention in aspect-based sentiment analysis (ABSA). Current ASQP research primarily relies on pre-trained generative language models to produce templated sequences, often complemented by grid-based auxiliary methods. Despite these efforts, the persistent challenge of generation instability remains unresolved and the effectiveness of grid methods remains underexplored in current studies. To this end, we introduce \textbf{G}rid Noise \textbf{D}iffusion \textbf{P}inpoint Network (\textbf{GDP}), a T5-based generative model aiming to tackle the issue of generation instability. The model consists of three novel modules, including Diffusion Vague Learning (DVL) to facilitate effective model learning and enhance overall robustness; Consistency Likelihood Learning (CLL) to discern the characteristics and commonalities of sentiment elements and thus reduce the impact of distributed noise; and GDP-FOR, a novel generation template, to enable models to generate outputs in a more natural way. Extensive experiments on four datasets demonstrate the remarkable effectiveness of our approach in addressing ASQP tasks.",
}
| Aspect sentiment quad prediction (ASQP) has garnered significant attention in aspect-based sentiment analysis (ABSA). Current ASQP research primarily relies on pre-trained generative language models to produce templated sequences, often complemented by grid-based auxiliary methods. Despite these efforts, the persistent challenge of generation instability remains unresolved and the effectiveness of grid methods remains underexplored in current studies. To this end, we introduce \textbf{G}rid Noise \textbf{D}iffusion \textbf{P}inpoint Network (\textbf{GDP}), a T5-based generative model aiming to tackle the issue of generation instability. The model consists of three novel modules, including Diffusion Vague Learning (DVL) to facilitate effective model learning and enhance overall robustness; Consistency Likelihood Learning (CLL) to discern the characteristics and commonalities of sentiment elements and thus reduce the impact of distributed noise; and GDP-FOR, a novel generation template, to enable models to generate outputs in a more natural way. Extensive experiments on four datasets demonstrate the remarkable effectiveness of our approach in addressing ASQP tasks. | [
"Zhu, Linan",
"Chen, Xiangfan",
"Guo, Xiaolei",
"Zhang, Chenwei",
"Zhu, Zhechao",
"Zhou, Zehai",
"Kong, Xiangjie"
]
| Pinpointing Diffusion Grid Noise to Enhance Aspect Sentiment Quad Prediction | findings-acl.222 | Poster | 2401.06394v1 |
https://aclanthology.org/2024.findings-acl.223.bib | @inproceedings{cappellazzo-etal-2024-continual,
title = "Continual Contrastive Spoken Language Understanding",
author = "Cappellazzo, Umberto and
Fini, Enrico and
Yang, Muqiao and
Falavigna, Daniele and
Brutti, Alessio and
Raj, Bhiksha",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.223",
pages = "3727--3741",
abstract = "Recently, neural networks have shown impressive progress across diverse fields, with speech processing being no exception. However, recent breakthroughs in this area require extensive offline training using large datasets and tremendous computing resources. Unfortunately, these models struggle to retain their previously acquired knowledge when learning new tasks continually. In this paper, we investigate the problem of learning sequence-to-sequence models for spoken language understanding in a class-incremental learning (CIL) setting and we propose COCONUT, a CIL method that relies on the combination of experience replay and contrastive learning. Through a modified version of the standard supervised contrastive loss, COCONUT preserves the learned representations by pulling closer samples from the same class and pushing away the others. Moreover, we leverage a multimodal contrastive loss that helps the model learn more discriminative representations of the new data by aligning audio and text features. We also investigate different contrastive designs to combine the strengths of the contrastive loss with teacher-student architectures used for distillation. Experiments on two established SLU datasets reveal the effectiveness of our proposed approach and significant improvements over the baselines. We also show that COCONUT can be combined with methods that operate on the decoder side of the model, resulting in further metrics improvements.",
}
| Recently, neural networks have shown impressive progress across diverse fields, with speech processing being no exception. However, recent breakthroughs in this area require extensive offline training using large datasets and tremendous computing resources. Unfortunately, these models struggle to retain their previously acquired knowledge when learning new tasks continually. In this paper, we investigate the problem of learning sequence-to-sequence models for spoken language understanding in a class-incremental learning (CIL) setting and we propose COCONUT, a CIL method that relies on the combination of experience replay and contrastive learning. Through a modified version of the standard supervised contrastive loss, COCONUT preserves the learned representations by pulling closer samples from the same class and pushing away the others. Moreover, we leverage a multimodal contrastive loss that helps the model learn more discriminative representations of the new data by aligning audio and text features. We also investigate different contrastive designs to combine the strengths of the contrastive loss with teacher-student architectures used for distillation. Experiments on two established SLU datasets reveal the effectiveness of our proposed approach and significant improvements over the baselines. We also show that COCONUT can be combined with methods that operate on the decoder side of the model, resulting in further metrics improvements. | [
"Cappellazzo, Umberto",
"Fini, Enrico",
"Yang, Muqiao",
"Falavigna, Daniele",
"Brutti, Alessio",
"Raj, Bhiksha"
]
| Continual Contrastive Spoken Language Understanding | findings-acl.223 | Poster | 2205.00693v2 |
https://aclanthology.org/2024.findings-acl.224.bib | @inproceedings{wang-etal-2024-llm,
title = "{LLM} as Prompter: Low-resource Inductive Reasoning on Arbitrary Knowledge Graphs",
author = "Wang, Kai and
Xu, Yuwei and
Wu, Zhiyong and
Luo, Siqiang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.224",
pages = "3742--3759",
abstract = "Knowledge Graph (KG) inductive reasoning, which aims to infer missing facts from new KGs that are not seen during training, has been widely adopted in various applications. One critical challenge of KG inductive reasoning is handling low-resource scenarios with scarcity in both textual and structural aspects. In this paper, we attempt to address this challenge with Large Language Models (LLMs). Particularly, we utilize the state-of-the-art LLMs to generate a graph-structural prompt to enhance the pre-trained Graph Neural Networks (GNNs), which brings us new methodological insights into the KG inductive reasoning methods, as well as high generalizability in practice. On the methodological side, we introduce a novel pretraining and prompting framework ProLINK, designed for low-resource inductive reasoning across arbitrary KGs without requiring additional training. On the practical side, we experimentally evaluate our approach on 36 low-resource KG datasets and find that ProLINK outperforms previous methods in three-shot, one-shot, and zero-shot reasoning tasks, exhibiting average performance improvements by 20{\%}, 45{\%}, and 147{\%}, respectively. Furthermore, ProLINK demonstrates strong robustness for various LLM promptings as well as full-shot scenarios.",
}
| Knowledge Graph (KG) inductive reasoning, which aims to infer missing facts from new KGs that are not seen during training, has been widely adopted in various applications. One critical challenge of KG inductive reasoning is handling low-resource scenarios with scarcity in both textual and structural aspects. In this paper, we attempt to address this challenge with Large Language Models (LLMs). Particularly, we utilize the state-of-the-art LLMs to generate a graph-structural prompt to enhance the pre-trained Graph Neural Networks (GNNs), which brings us new methodological insights into the KG inductive reasoning methods, as well as high generalizability in practice. On the methodological side, we introduce a novel pretraining and prompting framework ProLINK, designed for low-resource inductive reasoning across arbitrary KGs without requiring additional training. On the practical side, we experimentally evaluate our approach on 36 low-resource KG datasets and find that ProLINK outperforms previous methods in three-shot, one-shot, and zero-shot reasoning tasks, exhibiting average performance improvements by 20{\%}, 45{\%}, and 147{\%}, respectively. Furthermore, ProLINK demonstrates strong robustness for various LLM promptings as well as full-shot scenarios. | [
"Wang, Kai",
"Xu, Yuwei",
"Wu, Zhiyong",
"Luo, Siqiang"
]
| {LLM} as Prompter: Low-resource Inductive Reasoning on Arbitrary Knowledge Graphs | findings-acl.224 | Poster | 2402.11804v3 |
https://aclanthology.org/2024.findings-acl.225.bib | @inproceedings{chen-etal-2024-unsupervised,
title = "Unsupervised Parsing by Searching for Frequent Word Sequences among Sentences with Equivalent Predicate-Argument Structures",
author = "Chen, Junjie and
He, Xiangheng and
Bollegala, Danushka and
Miyao, Yusuke",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.225",
pages = "3760--3772",
abstract = "Unsupervised constituency parsing focuses on identifying word sequences that form a syntactic unit (i.e., constituents) in target sentences. Linguists identify the constituent by evaluating a set of Predicate-Argument Structure (PAS) equivalent sentences where we find the constituent appears more frequently than non-constituents (i.e., the constituent corresponds to a frequent word sequence within the sentence set). However, such frequency information is unavailable in previous parsing methods that identify the constituent by observing sentences with diverse PAS. In this study, we empirically show that constituents correspond to frequent word sequences in the PAS-equivalent sentence set. We propose a frequency-based parser, span-overlap, that (1) computes the span-overlap score as the word sequence{'}s frequency in the PAS-equivalent sentence set and (2) identifies the constituent structure by finding a constituent tree with the maximum span-overlap score. The parser achieves state-of-the-art level parsing accuracy, outperforming existing unsupervised parsers in eight out of ten languages. Additionally, we discover a multilingual phenomenon: participant-denoting constituents tend to have higher span-overlap scores than equal-length event-denoting constituents, meaning that the former tend to appear more frequently in the PAS-equivalent sentence set than the latter. The phenomenon indicates a statistical difference between the two constituent types, laying the foundation for future labeled unsupervised parsing research.",
}
| Unsupervised constituency parsing focuses on identifying word sequences that form a syntactic unit (i.e., constituents) in target sentences. Linguists identify the constituent by evaluating a set of Predicate-Argument Structure (PAS) equivalent sentences where we find the constituent appears more frequently than non-constituents (i.e., the constituent corresponds to a frequent word sequence within the sentence set). However, such frequency information is unavailable in previous parsing methods that identify the constituent by observing sentences with diverse PAS. In this study, we empirically show that constituents correspond to frequent word sequences in the PAS-equivalent sentence set. We propose a frequency-based parser, span-overlap, that (1) computes the span-overlap score as the word sequence{'}s frequency in the PAS-equivalent sentence set and (2) identifies the constituent structure by finding a constituent tree with the maximum span-overlap score. The parser achieves state-of-the-art level parsing accuracy, outperforming existing unsupervised parsers in eight out of ten languages. Additionally, we discover a multilingual phenomenon: participant-denoting constituents tend to have higher span-overlap scores than equal-length event-denoting constituents, meaning that the former tend to appear more frequently in the PAS-equivalent sentence set than the latter. The phenomenon indicates a statistical difference between the two constituent types, laying the foundation for future labeled unsupervised parsing research. | [
"Chen, Junjie",
"He, Xiangheng",
"Bollegala, Danushka",
"Miyao, Yusuke"
]
| Unsupervised Parsing by Searching for Frequent Word Sequences among Sentences with Equivalent Predicate-Argument Structures | findings-acl.225 | Poster | 2404.12059v2 |
https://aclanthology.org/2024.findings-acl.226.bib | @inproceedings{li-etal-2024-data,
title = "Data-Centric Explainable Debiasing for Improving Fairness in Pre-trained Language Models",
author = "Li, Yingji and
Du, Mengnan and
Song, Rui and
Wang, Xin and
Wang, Ying",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.226",
pages = "3773--3786",
abstract = "Human-like social bias of pre-trained language models (PLMs) on downstream tasks have attracted increasing attention. The potential flaws in the training data are the main factor that causes unfairness in PLMs. Existing data-centric debiasing strategies mainly leverage explicit bias words (defined as sensitive attribute words specific to demographic groups) for counterfactual data augmentation to balance the training data. However, they lack consideration of implicit bias words potentially associated with explicit bias words in complex distribution data, which indirectly harms the fairness of PLMs. To this end, we propose a **Data**-Centric **Debias**ing method (named Data-Debias), which uses an explainability method to search for implicit bias words to assist in debiasing PLMs. Specifically, we compute the feature attributions of all tokens using the Integrated Gradients method, and then treat the tokens that have a large impact on the model{'}s decision as implicit bias words. To make the search results more precise, we iteratively train a biased model to amplify the bias with each iteration. Finally, we use the implicit bias words searched in the last iteration to assist in debiasing PLMs. Extensive experimental results on multiple PLMs debiasing on three different classification tasks demonstrate that Data-Debias achieves state-of-the-art debiasing performance and strong generalization while maintaining predictive abilities.",
}
| Human-like social bias of pre-trained language models (PLMs) on downstream tasks have attracted increasing attention. The potential flaws in the training data are the main factor that causes unfairness in PLMs. Existing data-centric debiasing strategies mainly leverage explicit bias words (defined as sensitive attribute words specific to demographic groups) for counterfactual data augmentation to balance the training data. However, they lack consideration of implicit bias words potentially associated with explicit bias words in complex distribution data, which indirectly harms the fairness of PLMs. To this end, we propose a **Data**-Centric **Debias**ing method (named Data-Debias), which uses an explainability method to search for implicit bias words to assist in debiasing PLMs. Specifically, we compute the feature attributions of all tokens using the Integrated Gradients method, and then treat the tokens that have a large impact on the model{'}s decision as implicit bias words. To make the search results more precise, we iteratively train a biased model to amplify the bias with each iteration. Finally, we use the implicit bias words searched in the last iteration to assist in debiasing PLMs. Extensive experimental results on multiple PLMs debiasing on three different classification tasks demonstrate that Data-Debias achieves state-of-the-art debiasing performance and strong generalization while maintaining predictive abilities. | [
"Li, Yingji",
"Du, Mengnan",
"Song, Rui",
"Wang, Xin",
"Wang, Ying"
]
| Data-Centric Explainable Debiasing for Improving Fairness in Pre-trained Language Models | findings-acl.226 | Poster | 2402.11764v1 |
https://aclanthology.org/2024.findings-acl.227.bib | @inproceedings{jain-etal-2024-knowledge,
title = "Knowledge-Driven Cross-Document Relation Extraction",
author = "Jain, Monika and
Mutharaju, Raghava and
Singh, Kuldeep and
Kavuluru, Ramakanth",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.227",
pages = "3787--3797",
abstract = "Relation extraction (RE) is a well-known NLP application often treated as a sentence or document-level task. However, a handful of recent efforts explore it across documents or in the cross-document setting (CrossDocRE). This is distinct from the single document case because different documents often focus on disparate themes, while text within a document tends to have a single goal.Current CrossDocRE efforts do not consider domain knowledge, which are often assumed to be known to the reader when documents are authored. Here, we propose a novel approach, KXDocRE, that embed domain knowledge of entities with input text for cross-document RE. Our proposed framework has three main benefits over baselines: 1) it incorporates domain knowledge of entities along with documents{'} text; 2) it offers interpretability by producing explanatory text for predicted relations between entities 3) it improves performance over the prior methods. Code and models are available at \url{https://github.com/kracr/cross-doc-relation-extraction}.",
}
| Relation extraction (RE) is a well-known NLP application often treated as a sentence or document-level task. However, a handful of recent efforts explore it across documents or in the cross-document setting (CrossDocRE). This is distinct from the single document case because different documents often focus on disparate themes, while text within a document tends to have a single goal.Current CrossDocRE efforts do not consider domain knowledge, which are often assumed to be known to the reader when documents are authored. Here, we propose a novel approach, KXDocRE, that embed domain knowledge of entities with input text for cross-document RE. Our proposed framework has three main benefits over baselines: 1) it incorporates domain knowledge of entities along with documents{'} text; 2) it offers interpretability by producing explanatory text for predicted relations between entities 3) it improves performance over the prior methods. Code and models are available at \url{https://github.com/kracr/cross-doc-relation-extraction}. | [
"Jain, Monika",
"Mutharaju, Raghava",
"Singh, Kuldeep",
"Kavuluru, Ramakanth"
]
| Knowledge-Driven Cross-Document Relation Extraction | findings-acl.227 | Poster | 2310.09265v1 |
https://aclanthology.org/2024.findings-acl.228.bib | @inproceedings{chang-chen-2024-injecting,
title = "Injecting Salesperson{'}s Dialogue Strategies in Large Language Models with Chain-of-Thought Reasoning",
author = "Chang, Wen and
Chen, Yun-Nung",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.228",
pages = "3798--3812",
abstract = "Recent research in dialogue systems focuses on two main categories: task-oriented (TOD) and open-domain (chit-chat) dialogues. TOD systems help users complete specific tasks, while open-domain systems aim to create engaging conversations. However, user intents often emerge during interactions. A recent study introduced SalesBot, simulating dialogues that transition from chit-chat to task-oriented scenarios to train sales agents. Unfortunately, the initial data lacked smooth transitions and coherent long dialogues, resulting in unnatural interactions. This paper presents SalesBot 2.0, an improved dataset leveraging commonsense knowledge from large language models (LLMs) through strategic prompting. Additionally, we introduce SalesAgent, a novel model trained on salesperson interactions using chain-of-thought (CoT) reasoning. This model excels in transitioning topics, understanding user intents, and selecting appropriate strategies.Experiments with diverse user simulations validate our method{'}s effectiveness in controlling dialogue strategies in LLMs. SalesBot 2.0 enhances coherence and reduces aggression, improving model learning for sales-customer interactions.",
}
| Recent research in dialogue systems focuses on two main categories: task-oriented (TOD) and open-domain (chit-chat) dialogues. TOD systems help users complete specific tasks, while open-domain systems aim to create engaging conversations. However, user intents often emerge during interactions. A recent study introduced SalesBot, simulating dialogues that transition from chit-chat to task-oriented scenarios to train sales agents. Unfortunately, the initial data lacked smooth transitions and coherent long dialogues, resulting in unnatural interactions. This paper presents SalesBot 2.0, an improved dataset leveraging commonsense knowledge from large language models (LLMs) through strategic prompting. Additionally, we introduce SalesAgent, a novel model trained on salesperson interactions using chain-of-thought (CoT) reasoning. This model excels in transitioning topics, understanding user intents, and selecting appropriate strategies.Experiments with diverse user simulations validate our method{'}s effectiveness in controlling dialogue strategies in LLMs. SalesBot 2.0 enhances coherence and reduces aggression, improving model learning for sales-customer interactions. | [
"Chang, Wen",
"Chen, Yun-Nung"
]
| Injecting Salesperson{'}s Dialogue Strategies in Large Language Models with Chain-of-Thought Reasoning | findings-acl.228 | Poster | 2404.18564v1 |
https://aclanthology.org/2024.findings-acl.229.bib | @inproceedings{tian-etal-2024-kg,
title = "{KG}-Adapter: Enabling Knowledge Graph Integration in Large Language Models through Parameter-Efficient Fine-Tuning",
author = "Tian, Shiyu and
Luo, Yangyang and
Xu, Tianze and
Yuan, Caixia and
Jiang, Huixing and
Wei, Chen and
Wang, Xiaojie",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.229",
pages = "3813--3828",
abstract = "Although large language models (LLMs) show remarkable capabilities and generalizability across various tasks, they are criticized for lack of expertise. One promising solution is to combine knowledge graphs (KGs) with LLMs, and recent studies focus on integrating KGs into LLMs through prompt-based methods. However, these approaches fail to use the structural information of the KGs, suffer from the problem of knowledge conflict, and over-reliance on super LLMs. To address these challenges, we propose KG-Adapter, a parameter-level KG integration method based on parameter-efficient fine-tuning (PEFT). Specifically, we introduce a novel adapter structure designed for decoder-only LLMs, which can encode KGs from both node-centered and relation-centered perspectives, and then perform joint reasoning with LLMs to generate responses end-to-end. Experiments with diverse models on four datasets for two different tasks all demonstrate significant improvements. With only 28M parameters trained, we make the 7B-parameter LLM outperform the previous full-parameter fine-tuned state-of-the-art method and comparable to the prompt-based ChatGPT methods.",
}
| Although large language models (LLMs) show remarkable capabilities and generalizability across various tasks, they are criticized for lack of expertise. One promising solution is to combine knowledge graphs (KGs) with LLMs, and recent studies focus on integrating KGs into LLMs through prompt-based methods. However, these approaches fail to use the structural information of the KGs, suffer from the problem of knowledge conflict, and over-reliance on super LLMs. To address these challenges, we propose KG-Adapter, a parameter-level KG integration method based on parameter-efficient fine-tuning (PEFT). Specifically, we introduce a novel adapter structure designed for decoder-only LLMs, which can encode KGs from both node-centered and relation-centered perspectives, and then perform joint reasoning with LLMs to generate responses end-to-end. Experiments with diverse models on four datasets for two different tasks all demonstrate significant improvements. With only 28M parameters trained, we make the 7B-parameter LLM outperform the previous full-parameter fine-tuned state-of-the-art method and comparable to the prompt-based ChatGPT methods. | [
"Tian, Shiyu",
"Luo, Yangyang",
"Xu, Tianze",
"Yuan, Caixia",
"Jiang, Huixing",
"Wei, Chen",
"Wang, Xiaojie"
]
| {KG}-Adapter: Enabling Knowledge Graph Integration in Large Language Models through Parameter-Efficient Fine-Tuning | findings-acl.229 | Poster | 2402.01469v1 |
https://aclanthology.org/2024.findings-acl.230.bib | @inproceedings{lin-etal-2024-just,
title = "Just Ask One More Time! Self-Agreement Improves Reasoning of Language Models in (Almost) All Scenarios",
author = "Lin, Lei and
Fu, Jiayi and
Liu, Pengli and
Li, Qingyang and
Gong, Yan and
Wan, Junchen and
Zhang, Fuzheng and
Wang, Zhongyuan and
Zhang, Di and
Gai, Kun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.230",
pages = "3829--3852",
abstract = "Although chain-of-thought (CoT) prompting combined with language models has achieved encouraging results on complex reasoning tasks, the naive greedy decoding used in CoT prompting usually causes the repetitiveness and local optimality. To address this shortcoming, ensemble-optimization tries to obtain multiple reasoning paths to get the final answer assembly. However, current ensemble-optimization methods either simply employ rule-based post-processing such as self-consistency, or train an additional model based on several task-related human annotations to select the best one among multiple reasoning paths, yet fail to generalize to realistic settings where the type of input questions is unknown or the answer format of reasoning paths is unknown. To avoid their limitations, we propose Self-Agreement, a generalizable ensemble-optimization method applying in almost all scenarios where the type of input questions and the answer format of reasoning paths may be known or unknown. Self-agreement firstly samples from language model{'}s decoder to generate a diverse set of reasoning paths, and subsequently prompts the language model one more time to determine the optimal answer by selecting the most agreed answer among the sampled reasoning paths. Self-agreement simultaneously achieves remarkable performance on six public reasoning benchmarks and superior generalization capabilities.",
}
| Although chain-of-thought (CoT) prompting combined with language models has achieved encouraging results on complex reasoning tasks, the naive greedy decoding used in CoT prompting usually causes the repetitiveness and local optimality. To address this shortcoming, ensemble-optimization tries to obtain multiple reasoning paths to get the final answer assembly. However, current ensemble-optimization methods either simply employ rule-based post-processing such as self-consistency, or train an additional model based on several task-related human annotations to select the best one among multiple reasoning paths, yet fail to generalize to realistic settings where the type of input questions is unknown or the answer format of reasoning paths is unknown. To avoid their limitations, we propose Self-Agreement, a generalizable ensemble-optimization method applying in almost all scenarios where the type of input questions and the answer format of reasoning paths may be known or unknown. Self-agreement firstly samples from language model{'}s decoder to generate a diverse set of reasoning paths, and subsequently prompts the language model one more time to determine the optimal answer by selecting the most agreed answer among the sampled reasoning paths. Self-agreement simultaneously achieves remarkable performance on six public reasoning benchmarks and superior generalization capabilities. | [
"Lin, Lei",
"Fu, Jiayi",
"Liu, Pengli",
"Li, Qingyang",
"Gong, Yan",
"Wan, Junchen",
"Zhang, Fuzheng",
"Wang, Zhongyuan",
"Zhang, Di",
"Gai, Kun"
]
| Just Ask One More Time! Self-Agreement Improves Reasoning of Language Models in (Almost) All Scenarios | findings-acl.230 | Poster | 2311.08154v3 |
https://aclanthology.org/2024.findings-acl.231.bib | @inproceedings{srivastava-etal-2024-evaluating,
title = "Evaluating {LLM}s{'} Mathematical Reasoning in Financial Document Question Answering",
author = "Srivastava, Pragya and
Malik, Manuj and
Gupta, Vivek and
Ganu, Tanuja and
Roth, Dan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.231",
pages = "3853--3878",
abstract = "Large Language Models (LLMs), excel in natural language understanding, but their capability for complex mathematical reasoning with a hybrid of structured tables and unstructured text remain uncertain. This study explores LLMs{'} mathematical reasoning on four financial tabular question-answering datasets: TATQA, FinQA, ConvFinQA, and Multihiertt. Through extensive experiments with various models and prompting techniques, we assess how LLMs adapt to complex tables and mathematical tasks. We focus on sensitivity to table complexity and performance variations with an increasing number of arithmetic reasoning steps. The results provide insights into LLMs{'} capabilities and limitations in handling complex mathematical scenarios for semi-structured tables. Ultimately, we introduce a novel prompting technique EEDP tailored to semi-structured documents, matching or outperforming baselines performance while providing a nuanced understanding of LLMs abilities.",
}
| Large Language Models (LLMs), excel in natural language understanding, but their capability for complex mathematical reasoning with a hybrid of structured tables and unstructured text remain uncertain. This study explores LLMs{'} mathematical reasoning on four financial tabular question-answering datasets: TATQA, FinQA, ConvFinQA, and Multihiertt. Through extensive experiments with various models and prompting techniques, we assess how LLMs adapt to complex tables and mathematical tasks. We focus on sensitivity to table complexity and performance variations with an increasing number of arithmetic reasoning steps. The results provide insights into LLMs{'} capabilities and limitations in handling complex mathematical scenarios for semi-structured tables. Ultimately, we introduce a novel prompting technique EEDP tailored to semi-structured documents, matching or outperforming baselines performance while providing a nuanced understanding of LLMs abilities. | [
"Srivastava, Pragya",
"Malik, Manuj",
"Gupta, Vivek",
"Ganu, Tanuja",
"Roth, Dan"
]
| Evaluating {LLM}s{'} Mathematical Reasoning in Financial Document Question Answering | findings-acl.231 | Poster | 2402.11194v2 |
https://aclanthology.org/2024.findings-acl.232.bib | @inproceedings{xu-etal-2024-improving,
title = "Improving In-Context Learning with Prediction Feedback for Sentiment Analysis",
author = "Xu, Hongling and
Wang, Qianlong and
Zhang, Yice and
Yang, Min and
Zeng, Xi and
Qin, Bing and
Xu, Ruifeng",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.232",
pages = "3879--3890",
abstract = "Large language models (LLMs) have achieved promising results in sentiment analysis through the in-context learning (ICL) paradigm. However, their ability to distinguish subtle sentiments still remains a challenge. Inspired by the human ability to adjust understanding via feedback, this paper enhances ICL by incorporating prior predictions and feedback, aiming to rectify sentiment misinterpretation of LLMs. Specifically, the proposed framework consists of three steps: (1) acquiring prior predictions of LLMs, (2) devising predictive feedback based on correctness, and (3) leveraging a feedback-driven prompt to refine sentiment understanding. Experimental results across nine sentiment analysis datasets demonstrate the superiority of our framework over conventional ICL methods, with an average F1 improvement of 5.95{\%}.",
}
| Large language models (LLMs) have achieved promising results in sentiment analysis through the in-context learning (ICL) paradigm. However, their ability to distinguish subtle sentiments still remains a challenge. Inspired by the human ability to adjust understanding via feedback, this paper enhances ICL by incorporating prior predictions and feedback, aiming to rectify sentiment misinterpretation of LLMs. Specifically, the proposed framework consists of three steps: (1) acquiring prior predictions of LLMs, (2) devising predictive feedback based on correctness, and (3) leveraging a feedback-driven prompt to refine sentiment understanding. Experimental results across nine sentiment analysis datasets demonstrate the superiority of our framework over conventional ICL methods, with an average F1 improvement of 5.95{\%}. | [
"Xu, Hongling",
"Wang, Qianlong",
"Zhang, Yice",
"Yang, Min",
"Zeng, Xi",
"Qin, Bing",
"Xu, Ruifeng"
]
| Improving In-Context Learning with Prediction Feedback for Sentiment Analysis | findings-acl.232 | Poster | 2406.02911v1 |
https://aclanthology.org/2024.findings-acl.233.bib | @inproceedings{li-etal-2024-large-language,
title = "Can Large Language Models Mine Interpretable Financial Factors More Effectively? A Neural-Symbolic Factor Mining Agent Model",
author = "Li, Zhiwei and
Song, Ran and
Sun, Caihong and
Xu, Wei and
Yu, Zhengtao and
Wen, Ji-Rong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.233",
pages = "3891--3902",
abstract = "Finding interpretable factors for stock returns is the most vital issue in the empirical asset pricing domain. As data-driven methods, existing factor mining models can be categorized into symbol-based and neural-based models. Symbol-based models are interpretable but inefficient, while neural-based approaches are efficient but lack interpretability. Hence, mining interpretable factors effectively presents a significant challenge. Inspired by the success of Large Language Models (LLMs) in various tasks, we propose a FActor Mining Agent (FAMA) model that enables LLMs to integrate the strengths of both neural and symbolic models for factor mining. In this paper, FAMA consists of two main components: Cross-Sample Selection (CSS) and Chain-of-Experience (CoE). CSS addresses the homogeneity challenges in LLMs during factor mining by assimilating diverse factors as in-context samples, whereas CoE enables LLMs to leverage past successful mining experiences, expediting the mining of effective factors. Experimental evaluations on real-world stock market data demonstrate the effectiveness of our approach by surpassing the SOTA RankIC by 0.006 and RankICIR by 0.105 in predicting S{\&}P 500 returns. Furthermore, the investment simulation shows that our model can achieve superior performance with an annualized return of 38.4{\%} and a Sharpe ratio of 667.2{\%}.",
}
| Finding interpretable factors for stock returns is the most vital issue in the empirical asset pricing domain. As data-driven methods, existing factor mining models can be categorized into symbol-based and neural-based models. Symbol-based models are interpretable but inefficient, while neural-based approaches are efficient but lack interpretability. Hence, mining interpretable factors effectively presents a significant challenge. Inspired by the success of Large Language Models (LLMs) in various tasks, we propose a FActor Mining Agent (FAMA) model that enables LLMs to integrate the strengths of both neural and symbolic models for factor mining. In this paper, FAMA consists of two main components: Cross-Sample Selection (CSS) and Chain-of-Experience (CoE). CSS addresses the homogeneity challenges in LLMs during factor mining by assimilating diverse factors as in-context samples, whereas CoE enables LLMs to leverage past successful mining experiences, expediting the mining of effective factors. Experimental evaluations on real-world stock market data demonstrate the effectiveness of our approach by surpassing the SOTA RankIC by 0.006 and RankICIR by 0.105 in predicting S{\&}P 500 returns. Furthermore, the investment simulation shows that our model can achieve superior performance with an annualized return of 38.4{\%} and a Sharpe ratio of 667.2{\%}. | [
"Li, Zhiwei",
"Song, Ran",
"Sun, Caihong",
"Xu, Wei",
"Yu, Zhengtao",
"Wen, Ji-Rong"
]
| Can Large Language Models Mine Interpretable Financial Factors More Effectively? A Neural-Symbolic Factor Mining Agent Model | findings-acl.233 | Poster | 2406.18394v1 |
https://aclanthology.org/2024.findings-acl.234.bib | @inproceedings{yuan-etal-2024-discerning,
title = "Discerning and Resolving Knowledge Conflicts through Adaptive Decoding with Contextual Information-Entropy Constraint",
author = "Yuan, Xiaowei and
Yang, Zhao and
Wang, Yequan and
Liu, Shengping and
Zhao, Jun and
Liu, Kang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.234",
pages = "3903--3922",
abstract = "Large language models (LLMs) internalize enormous \textit{parametric knowledge} during pre-training. Concurrently, realistic applications necessitate external \textit{contextual knowledge} to aid models on the underlying tasks. This raises a crucial dilemma known as \textit{knowledge conflicts}, where the contextual knowledge clashes with the parametric knowledge. However, existing decoding works are specialized in resolving knowledge conflicts and could inadvertently deteriorate performance in absence of conflicts. In this paper, we propose an adaptive decoding method, termed as contextual information-entropy constraint decoding (COIECD), to discern whether the knowledge conflicts occur and resolve them. It can improve the model{'}s faithfulness to conflicting context, and simultaneously maintain high performance among non-conflicting context. Our experiments show that COIECD exhibits strong performance and robustness over knowledge conflicts in realistic datasets.",
}
| Large language models (LLMs) internalize enormous \textit{parametric knowledge} during pre-training. Concurrently, realistic applications necessitate external \textit{contextual knowledge} to aid models on the underlying tasks. This raises a crucial dilemma known as \textit{knowledge conflicts}, where the contextual knowledge clashes with the parametric knowledge. However, existing decoding works are specialized in resolving knowledge conflicts and could inadvertently deteriorate performance in absence of conflicts. In this paper, we propose an adaptive decoding method, termed as contextual information-entropy constraint decoding (COIECD), to discern whether the knowledge conflicts occur and resolve them. It can improve the model{'}s faithfulness to conflicting context, and simultaneously maintain high performance among non-conflicting context. Our experiments show that COIECD exhibits strong performance and robustness over knowledge conflicts in realistic datasets. | [
"Yuan, Xiaowei",
"Yang, Zhao",
"Wang, Yequan",
"Liu, Shengping",
"Zhao, Jun",
"Liu, Kang"
]
| Discerning and Resolving Knowledge Conflicts through Adaptive Decoding with Contextual Information-Entropy Constraint | findings-acl.234 | Poster | 2402.11893v3 |
https://aclanthology.org/2024.findings-acl.235.bib | @inproceedings{li-etal-2024-salad,
title = "{SALAD}-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Models",
author = "Li, Lijun and
Dong, Bowen and
Wang, Ruohui and
Hu, Xuhao and
Zuo, Wangmeng and
Lin, Dahua and
Qiao, Yu and
Shao, Jing",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.235",
pages = "3923--3954",
abstract = "In the rapidly evolving landscape of Large Language Models (LLMs), ensuring robust safety measures is paramount. To meet this crucial need, we propose SALAD-Bench, a safety benchmark specifically designed for evaluating LLMs, attack, and defense methods. Distinguished by its breadth, SALAD-Bench transcends conventional benchmarks through its large scale, rich diversity, intricate taxonomy spanning three levels, and versatile functionalities.SALAD-Bench is crafted with a meticulous array of questions, from standard queries to complex ones enriched with attack, defense modifications and multiple-choice. To effectively manage the inherent complexity, we introduce an innovative evaluators: the LLM-based MD-Judge for QA pairs with a particular focus on attack-enhanced queries, ensuring a seamless, and reliable evaluation. Above components extend SALAD-Bench from standard LLM safety evaluation to both LLM attack and defense methods evaluation, ensuring the joint-purpose utility. Our extensive experiments shed light on the resilience of LLMs against emerging threats and the efficacy of contemporary defense tactics. Data and evaluator are released under https://github.com/OpenSafetyLab/SALAD-BENCH",
}
| In the rapidly evolving landscape of Large Language Models (LLMs), ensuring robust safety measures is paramount. To meet this crucial need, we propose SALAD-Bench, a safety benchmark specifically designed for evaluating LLMs, attack, and defense methods. Distinguished by its breadth, SALAD-Bench transcends conventional benchmarks through its large scale, rich diversity, intricate taxonomy spanning three levels, and versatile functionalities.SALAD-Bench is crafted with a meticulous array of questions, from standard queries to complex ones enriched with attack, defense modifications and multiple-choice. To effectively manage the inherent complexity, we introduce an innovative evaluators: the LLM-based MD-Judge for QA pairs with a particular focus on attack-enhanced queries, ensuring a seamless, and reliable evaluation. Above components extend SALAD-Bench from standard LLM safety evaluation to both LLM attack and defense methods evaluation, ensuring the joint-purpose utility. Our extensive experiments shed light on the resilience of LLMs against emerging threats and the efficacy of contemporary defense tactics. Data and evaluator are released under https://github.com/OpenSafetyLab/SALAD-BENCH | [
"Li, Lijun",
"Dong, Bowen",
"Wang, Ruohui",
"Hu, Xuhao",
"Zuo, Wangmeng",
"Lin, Dahua",
"Qiao, Yu",
"Shao, Jing"
]
| {SALAD}-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Models | findings-acl.235 | Poster | 2406.10311v1 |
https://aclanthology.org/2024.findings-acl.236.bib | @inproceedings{messina-etal-2024-extracting,
title = "Extracting and Encoding: Leveraging Large Language Models and Medical Knowledge to Enhance Radiological Text Representation",
author = "Messina, Pablo and
Vidal, Rene and
Parra, Denis and
Soto, Alvaro and
Araujo, Vladimir",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.236",
pages = "3955--3986",
abstract = "Advancing representation learning in specialized fields like medicine remains challenging due to the scarcity of expert annotations for text and images. To tackle this issue, we present a novel two-stage framework designed to extract high-quality factual statements from free-text radiology reports in order to improve the representations of text encoders and, consequently, their performance on various downstream tasks.In the first stage, we propose a \textit{Fact Extractor} that leverages large language models (LLMs) to identify factual statements from well-curated domain-specific datasets. In the second stage, we introduce a \textit{Fact Encoder} (CXRFE) based on a BERT model fine-tuned with objective functions designed to improve its representations using the extracted factual data. Our framework also includes a new embedding-based metric (CXRFEScore) for evaluating chest X-ray text generation systems, leveraging both stages of our approach. Extensive evaluations show that our fact extractor and encoder outperform current state-of-the-art methods in tasks such as sentence ranking, natural language inference, and label extraction from radiology reports. Additionally, our metric proves to be more robust and effective than existing metrics commonly used in the radiology report generation literature. The code of this project is available at \url{https://github.com/PabloMessina/CXR-Fact-Encoder}.",
}
| Advancing representation learning in specialized fields like medicine remains challenging due to the scarcity of expert annotations for text and images. To tackle this issue, we present a novel two-stage framework designed to extract high-quality factual statements from free-text radiology reports in order to improve the representations of text encoders and, consequently, their performance on various downstream tasks.In the first stage, we propose a \textit{Fact Extractor} that leverages large language models (LLMs) to identify factual statements from well-curated domain-specific datasets. In the second stage, we introduce a \textit{Fact Encoder} (CXRFE) based on a BERT model fine-tuned with objective functions designed to improve its representations using the extracted factual data. Our framework also includes a new embedding-based metric (CXRFEScore) for evaluating chest X-ray text generation systems, leveraging both stages of our approach. Extensive evaluations show that our fact extractor and encoder outperform current state-of-the-art methods in tasks such as sentence ranking, natural language inference, and label extraction from radiology reports. Additionally, our metric proves to be more robust and effective than existing metrics commonly used in the radiology report generation literature. The code of this project is available at \url{https://github.com/PabloMessina/CXR-Fact-Encoder}. | [
"Messina, Pablo",
"Vidal, Rene",
"Parra, Denis",
"Soto, Alvaro",
"Araujo, Vladimir"
]
| Extracting and Encoding: Leveraging Large Language Models and Medical Knowledge to Enhance Radiological Text Representation | findings-acl.236 | Poster | 2407.01948v1 |
https://aclanthology.org/2024.findings-acl.237.bib | @inproceedings{yuan-etal-2024-gnnavi,
title = "{GNN}avi: Navigating the Information Flow in Large Language Models by Graph Neural Network",
author = {Yuan, Shuzhou and
Nie, Ercong and
F{\"a}rber, Michael and
Schmid, Helmut and
Schuetze, Hinrich},
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.237",
pages = "3987--4001",
abstract = "Large Language Models (LLMs) exhibit strong In-Context Learning (ICL) capabilities when prompts with demonstrations are used. However, fine-tuning still remains crucial to further enhance their adaptability. Prompt-based fine-tuning proves to be an effective fine-tuning method in low-data scenarios, but high demands on computing resources limit its practicality. We address this issue by introducing a prompt-based parameter-efficient fine-tuning (PEFT) approach. GNNavi leverages insights into ICL{'}s information flow dynamics, which indicates that label words act in prompts as anchors for information propagation. GNNavi employs a Graph Neural Network (GNN) layer to precisely guide the aggregation and distribution of information flow during the processing of prompts by hardwiring the desired information flow into the GNN. Our experiments on text classification tasks with GPT-2 and Llama2 show GNNavi surpasses standard prompt-based fine-tuning methods in few-shot settings by updating just 0.2{\%} to 0.5{\%} of parameters. We compare GNNavi with prevalent PEFT approaches, such as prefix tuning, LoRA and Adapter in terms of performance and efficiency. Our analysis reveals that GNNavi enhances information flow and ensures a clear aggregation process.",
}
| Large Language Models (LLMs) exhibit strong In-Context Learning (ICL) capabilities when prompts with demonstrations are used. However, fine-tuning still remains crucial to further enhance their adaptability. Prompt-based fine-tuning proves to be an effective fine-tuning method in low-data scenarios, but high demands on computing resources limit its practicality. We address this issue by introducing a prompt-based parameter-efficient fine-tuning (PEFT) approach. GNNavi leverages insights into ICL{'}s information flow dynamics, which indicates that label words act in prompts as anchors for information propagation. GNNavi employs a Graph Neural Network (GNN) layer to precisely guide the aggregation and distribution of information flow during the processing of prompts by hardwiring the desired information flow into the GNN. Our experiments on text classification tasks with GPT-2 and Llama2 show GNNavi surpasses standard prompt-based fine-tuning methods in few-shot settings by updating just 0.2{\%} to 0.5{\%} of parameters. We compare GNNavi with prevalent PEFT approaches, such as prefix tuning, LoRA and Adapter in terms of performance and efficiency. Our analysis reveals that GNNavi enhances information flow and ensures a clear aggregation process. | [
"Yuan, Shuzhou",
"Nie, Ercong",
"F{\\\"a}rber, Michael",
"Schmid, Helmut",
"Schuetze, Hinrich"
]
| {GNN}avi: Navigating the Information Flow in Large Language Models by Graph Neural Network | findings-acl.237 | Poster | 2402.11709v2 |
https://aclanthology.org/2024.findings-acl.238.bib | @inproceedings{subramanian-etal-2024-qalm,
title = "{M}-{QALM}: A Benchmark to Assess Clinical Reading Comprehension and Knowledge Recall in Large Language Models via Question Answering",
author = "Subramanian, Anand and
Schlegel, Viktor and
Ramesh Kashyap, Abhinav and
Nguyen, Thanh-Tung and
Dwivedi, Vijay Prakash and
Winkler, Stefan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.238",
pages = "4002--4042",
abstract = "There is vivid research on adapting Large Language Models (LLMs) to perform a variety of tasks in high-stakes domains such as healthcare. Despite their popularity, there is a lack of understanding of the extent and contributing factors that allow LLMs to recall relevant knowledge and combine it with presented information in the clinical and biomedical domain: a fundamental pre-requisite for success on down-stream tasks.Addressing this gap, we use Multiple Choice and Abstractive Question Answering to conduct a large-scale empirical study on 22 datasets in three generalist and three specialist biomedical sub-domains. Our multifaceted analysis of the performance of 15 LLMs, further broken down by sub-domain, source of knowledge and model architecture, uncovers success factors such as instruction tuning that lead to improved recall and comprehension. We further show that while recently proposed domain-adapted models may lack adequate knowledge, directly fine-tuning on our collected medical knowledge datasets shows encouraging results, even generalising to unseen specialist sub-domains. We complement the quantitative results with a skill-oriented manual error analysis, which reveals a significant gap between the models{'} capabilities to simply recall necessary knowledge and to integrate it with the presented context.To foster research and collaboration in this field we share M-QALM, our resources, standardised methodology, and evaluation results, with the research community to facilitate further advancements in clinical knowledge representation learning within language models.",
}
| There is vivid research on adapting Large Language Models (LLMs) to perform a variety of tasks in high-stakes domains such as healthcare. Despite their popularity, there is a lack of understanding of the extent and contributing factors that allow LLMs to recall relevant knowledge and combine it with presented information in the clinical and biomedical domain: a fundamental pre-requisite for success on down-stream tasks.Addressing this gap, we use Multiple Choice and Abstractive Question Answering to conduct a large-scale empirical study on 22 datasets in three generalist and three specialist biomedical sub-domains. Our multifaceted analysis of the performance of 15 LLMs, further broken down by sub-domain, source of knowledge and model architecture, uncovers success factors such as instruction tuning that lead to improved recall and comprehension. We further show that while recently proposed domain-adapted models may lack adequate knowledge, directly fine-tuning on our collected medical knowledge datasets shows encouraging results, even generalising to unseen specialist sub-domains. We complement the quantitative results with a skill-oriented manual error analysis, which reveals a significant gap between the models{'} capabilities to simply recall necessary knowledge and to integrate it with the presented context.To foster research and collaboration in this field we share M-QALM, our resources, standardised methodology, and evaluation results, with the research community to facilitate further advancements in clinical knowledge representation learning within language models. | [
"Subramanian, An",
"",
"Schlegel, Viktor",
"Ramesh Kashyap, Abhinav",
"Nguyen, Thanh-Tung",
"Dwivedi, Vijay Prakash",
"Winkler, Stefan"
]
| {M}-{QALM}: A Benchmark to Assess Clinical Reading Comprehension and Knowledge Recall in Large Language Models via Question Answering | findings-acl.238 | Poster | 2406.03699v1 |
https://aclanthology.org/2024.findings-acl.239.bib | @inproceedings{saxena-keller-2024-moviesum,
title = "{M}ovie{S}um: An Abstractive Summarization Dataset for Movie Screenplays",
author = "Saxena, Rohit and
Keller, Frank",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.239",
pages = "4043--4050",
abstract = "Movie screenplay summarization is challenging, as it requires an understanding of long input contexts and various elements unique to movies. Large language models have shown significant advancements in document summarization, but they often struggle with processing long input contexts. Furthermore, while television transcripts have received attention in recent studies, movie screenplay summarization remains underexplored. To stimulate research in this area, we present a new dataset, MovieSum, for abstractive summarization of movie screenplays. This dataset comprises 2200 movie screenplays accompanied by their Wikipedia plot summaries. We manually formatted the movie screenplays to represent their structural elements. Compared to existing datasets, MovieSum possesses several distinctive features: 1) It includes movie screenplays which are longer than scripts of TV episodes. 2) It is twice the size of previous movie screenplay datasets. 3) It provides metadata with IMDb IDs to facilitate access to additional external knowledge. We also show the results of recently released large language models applied to summarization on our dataset to provide a detailed baseline.",
}
| Movie screenplay summarization is challenging, as it requires an understanding of long input contexts and various elements unique to movies. Large language models have shown significant advancements in document summarization, but they often struggle with processing long input contexts. Furthermore, while television transcripts have received attention in recent studies, movie screenplay summarization remains underexplored. To stimulate research in this area, we present a new dataset, MovieSum, for abstractive summarization of movie screenplays. This dataset comprises 2200 movie screenplays accompanied by their Wikipedia plot summaries. We manually formatted the movie screenplays to represent their structural elements. Compared to existing datasets, MovieSum possesses several distinctive features: 1) It includes movie screenplays which are longer than scripts of TV episodes. 2) It is twice the size of previous movie screenplay datasets. 3) It provides metadata with IMDb IDs to facilitate access to additional external knowledge. We also show the results of recently released large language models applied to summarization on our dataset to provide a detailed baseline. | [
"Saxena, Rohit",
"Keller, Frank"
]
| {M}ovie{S}um: An Abstractive Summarization Dataset for Movie Screenplays | findings-acl.239 | Poster | 2408.06281v1 |
https://aclanthology.org/2024.findings-acl.240.bib | @inproceedings{pei-etal-2024-autonomous,
title = "Autonomous Workflow for Multimodal Fine-Grained Training Assistants Towards Mixed Reality",
author = "Pei, Jiahuan and
Viola, Irene and
Huang, Haochen and
Wang, Junxiao and
Ahsan, Moonisa and
Ye, Fanghua and
Yiming, Jiang and
Sai, Yao and
Wang, Di and
Chen, Zhumin and
Ren, Pengjie and
Cesar, Pablo",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.240",
pages = "4051--4066",
abstract = "Autonomous artificial intelligence (AI) agents have emerged as promising protocols for automatically understanding the language-based environment, particularly with the exponential development of large language models (LLMs). However, a fine-grained, comprehensive understanding of multimodal environments remains under-explored. This work designs an autonomous workflow tailored for integrating AI agents seamlessly into extended reality (XR) applications for fine-grained training. We present a demonstration of a multimodal fine-grained training assistant for LEGO brick assembly in a pilot XR environment. Specifically, we design a cerebral language agent that integrates LLM with memory, planning, and interaction with XR tools and a vision-language agent, enabling agents to decide their actions based on past experiences. Furthermore, we introduce LEGO-MRTA, a multimodal fine-grained assembly dialogue dataset synthesized automatically in the workflow served by a commercial LLM. This dataset comprises multimodal instruction manuals, conversations, XR responses, and vision question answering. Last, we present several prevailing open-resource LLMs as benchmarks, assessing their performance with and without fine-tuning on the proposed dataset. We anticipate that the broader impact of this workflow will advance the development of smarter assistants for seamless user interaction in XR environments, fostering research in both AI and HCI communities.",
}
| Autonomous artificial intelligence (AI) agents have emerged as promising protocols for automatically understanding the language-based environment, particularly with the exponential development of large language models (LLMs). However, a fine-grained, comprehensive understanding of multimodal environments remains under-explored. This work designs an autonomous workflow tailored for integrating AI agents seamlessly into extended reality (XR) applications for fine-grained training. We present a demonstration of a multimodal fine-grained training assistant for LEGO brick assembly in a pilot XR environment. Specifically, we design a cerebral language agent that integrates LLM with memory, planning, and interaction with XR tools and a vision-language agent, enabling agents to decide their actions based on past experiences. Furthermore, we introduce LEGO-MRTA, a multimodal fine-grained assembly dialogue dataset synthesized automatically in the workflow served by a commercial LLM. This dataset comprises multimodal instruction manuals, conversations, XR responses, and vision question answering. Last, we present several prevailing open-resource LLMs as benchmarks, assessing their performance with and without fine-tuning on the proposed dataset. We anticipate that the broader impact of this workflow will advance the development of smarter assistants for seamless user interaction in XR environments, fostering research in both AI and HCI communities. | [
"Pei, Jiahuan",
"Viola, Irene",
"Huang, Haochen",
"Wang, Junxiao",
"Ahsan, Moonisa",
"Ye, Fanghua",
"Yiming, Jiang",
"Sai, Yao",
"Wang, Di",
"Chen, Zhumin",
"Ren, Pengjie",
"Cesar, Pablo"
]
| Autonomous Workflow for Multimodal Fine-Grained Training Assistants Towards Mixed Reality | findings-acl.240 | Poster | 2405.13034v2 |
https://aclanthology.org/2024.findings-acl.241.bib | @inproceedings{holt-etal-2024-perceptions,
title = "Perceptions of Language Technology Failures from {S}outh {A}sian {E}nglish Speakers",
author = "Holt, Faye and
Held, William and
Yang, Diyi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.241",
pages = "4067--4081",
abstract = "English NLP systems have empirically worse performance for dialects other than Standard American English (SAmE). However, how these discrepancies impact use of language technology by speakers of non-SAmE global Englishes is not well understood. We focus on reducing this gap for South Asian Englishes (SAsE), a macro-group of regional varieties with cumulatively more speakers than SAmE, by surveying SAsE speakers about their interactions with language technology and compare their responses to a control survey of SAmE speakers. SAsE speakers are more likely to recall failures with language technology and more likely to reference specific issues with written language technology than their SAmE counterparts. Furthermore, SAsE speakers indicate that they modify both their lexicon and syntax to make technology work better, but that lexical issues are perceived as the most salient challenge. We then assess whether these issues are pervasive in more recently developed Large Language Models (LLMs), introducing two benchmarks for broader SAsE Lexical and Indian English Syntactic understanding and evaluating 11 families of LLMs on them.",
}
| English NLP systems have empirically worse performance for dialects other than Standard American English (SAmE). However, how these discrepancies impact use of language technology by speakers of non-SAmE global Englishes is not well understood. We focus on reducing this gap for South Asian Englishes (SAsE), a macro-group of regional varieties with cumulatively more speakers than SAmE, by surveying SAsE speakers about their interactions with language technology and compare their responses to a control survey of SAmE speakers. SAsE speakers are more likely to recall failures with language technology and more likely to reference specific issues with written language technology than their SAmE counterparts. Furthermore, SAsE speakers indicate that they modify both their lexicon and syntax to make technology work better, but that lexical issues are perceived as the most salient challenge. We then assess whether these issues are pervasive in more recently developed Large Language Models (LLMs), introducing two benchmarks for broader SAsE Lexical and Indian English Syntactic understanding and evaluating 11 families of LLMs on them. | [
"Holt, Faye",
"Held, William",
"Yang, Diyi"
]
| Perceptions of Language Technology Failures from {S}outh {A}sian {E}nglish Speakers | findings-acl.241 | Poster | 2310.05381v1 |
https://aclanthology.org/2024.findings-acl.242.bib | @inproceedings{brinkmann-etal-2024-mechanistic,
title = "A Mechanistic Analysis of a Transformer Trained on a Symbolic Multi-Step Reasoning Task",
author = "Brinkmann, Jannik and
Sheshadri, Abhay and
Levoso, Victor and
Swoboda, Paul and
Bartelt, Christian",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.242",
pages = "4082--4102",
abstract = "Transformers demonstrate impressive performance on a range of reasoning benchmarks. To evaluate the degree to which these abilities are a result of actual reasoning, existing work has focused on developing sophisticated benchmarks for behavioral studies. However, these studies do not provide insights into the internal mechanisms driving the observed capabilities. To improve our understanding of the internal mechanisms of transformers, we present a comprehensive mechanistic analysis of a transformer trained on a synthetic reasoning task. We identify a set of interpretable mechanisms the model uses to solve the task, and validate our findings using correlational and causal evidence. Our results suggest that it implements a depth-bounded recurrent mechanisms that operates in parallel and stores intermediate results in selected token positions. We anticipate that the motifs we identified in our synthetic setting can provide valuable insights into the broader operating principles of transformers and thus provide a basis for understanding more complex models.",
}
| Transformers demonstrate impressive performance on a range of reasoning benchmarks. To evaluate the degree to which these abilities are a result of actual reasoning, existing work has focused on developing sophisticated benchmarks for behavioral studies. However, these studies do not provide insights into the internal mechanisms driving the observed capabilities. To improve our understanding of the internal mechanisms of transformers, we present a comprehensive mechanistic analysis of a transformer trained on a synthetic reasoning task. We identify a set of interpretable mechanisms the model uses to solve the task, and validate our findings using correlational and causal evidence. Our results suggest that it implements a depth-bounded recurrent mechanisms that operates in parallel and stores intermediate results in selected token positions. We anticipate that the motifs we identified in our synthetic setting can provide valuable insights into the broader operating principles of transformers and thus provide a basis for understanding more complex models. | [
"Brinkmann, Jannik",
"Sheshadri, Abhay",
"Levoso, Victor",
"Swoboda, Paul",
"Bartelt, Christian"
]
| A Mechanistic Analysis of a Transformer Trained on a Symbolic Multi-Step Reasoning Task | findings-acl.242 | Poster | 2402.11917v3 |
https://aclanthology.org/2024.findings-acl.243.bib | @inproceedings{zhang-etal-2024-optimal,
title = "Optimal Transport Guided Correlation Assignment for Multimodal Entity Linking",
author = "Zhang, Zefeng and
Sheng, Jiawei and
Chuang, Zhang and
Liangyunzhi, Liangyunzhi and
Zhang, Wenyuan and
Wang, Siqi and
Liu, Tingwen",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.243",
pages = "4103--4117",
abstract = "Multimodal entity linking (MEL) aims to link ambiguous mentions in multimodal contexts to entities in a multimodal knowledge graph. A pivotal challenge is to fully leverage multi-element correlations between mentions and entities to bridge modality gap and enable fine-grained semantic matching. Existing methods attempt several local correlative mechanisms, relying heavily on the automatically learned attention weights, which may over-concentrate on partial correlations. To mitigate this issue, we formulate the correlation assignment problem as an optimal transport (OT) problem, and propose a novel MEL framework, namely OT-MEL, with OT-guided correlation assignment. Thereby, we exploit the correlation between multimodal features to enhance multimodal fusion, and the correlation between mentions and entities to enhance fine-grained matching. To accelerate model prediction, we further leverage knowledge distillation to transfer OT assignment knowledge to attention mechanism. Experimental results show that our model significantly outperforms previous state-of-the-art baselines and confirm the effectiveness of the OT-guided correlation assignment.",
}
| Multimodal entity linking (MEL) aims to link ambiguous mentions in multimodal contexts to entities in a multimodal knowledge graph. A pivotal challenge is to fully leverage multi-element correlations between mentions and entities to bridge modality gap and enable fine-grained semantic matching. Existing methods attempt several local correlative mechanisms, relying heavily on the automatically learned attention weights, which may over-concentrate on partial correlations. To mitigate this issue, we formulate the correlation assignment problem as an optimal transport (OT) problem, and propose a novel MEL framework, namely OT-MEL, with OT-guided correlation assignment. Thereby, we exploit the correlation between multimodal features to enhance multimodal fusion, and the correlation between mentions and entities to enhance fine-grained matching. To accelerate model prediction, we further leverage knowledge distillation to transfer OT assignment knowledge to attention mechanism. Experimental results show that our model significantly outperforms previous state-of-the-art baselines and confirm the effectiveness of the OT-guided correlation assignment. | [
"Zhang, Zefeng",
"Sheng, Jiawei",
"Chuang, Zhang",
"Liangyunzhi, Liangyunzhi",
"Zhang, Wenyuan",
"Wang, Siqi",
"Liu, Tingwen"
]
| Optimal Transport Guided Correlation Assignment for Multimodal Entity Linking | findings-acl.243 | Poster | 2406.01934v2 |
https://aclanthology.org/2024.findings-acl.244.bib | @inproceedings{svete-etal-2024-efficiently,
title = "On Efficiently Representing Regular Languages as {RNN}s",
author = "Svete, Anej and
Chan, Robin and
Cotterell, Ryan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.244",
pages = "4118--4135",
abstract = "Recent work by Hewitt et al. (2020) provides an interpretation of the empirical success of recurrent neural networks (RNNs) as language models (LMs). It shows that RNNs can efficiently represent bounded hierarchical structures that are prevalent in human language.This suggests that RNNs{'} success might be linked to their ability to model hierarchy. However, a closer inspection of hewitt-etal-2020-rnns construction shows that it is not inherently limited to hierarchical structures. This poses a natural question: What other classes of LMs RNNs can efficiently represent? To this end, we generalize Hewitt et al.{'}s (2020) construction and show that RNNs can efficiently represent a larger class of LMs than previously claimed{---}specifically, those that can be represented by a pushdown automaton with a bounded stack and a specific stack update function. Altogether, the efficiency of representing this diverse class of LMs with RNN LMs suggests novel interpretations of their inductive bias.",
}
| Recent work by Hewitt et al. (2020) provides an interpretation of the empirical success of recurrent neural networks (RNNs) as language models (LMs). It shows that RNNs can efficiently represent bounded hierarchical structures that are prevalent in human language.This suggests that RNNs{'} success might be linked to their ability to model hierarchy. However, a closer inspection of hewitt-etal-2020-rnns construction shows that it is not inherently limited to hierarchical structures. This poses a natural question: What other classes of LMs RNNs can efficiently represent? To this end, we generalize Hewitt et al.{'}s (2020) construction and show that RNNs can efficiently represent a larger class of LMs than previously claimed{---}specifically, those that can be represented by a pushdown automaton with a bounded stack and a specific stack update function. Altogether, the efficiency of representing this diverse class of LMs with RNN LMs suggests novel interpretations of their inductive bias. | [
"Svete, Anej",
"Chan, Robin",
"Cotterell, Ryan"
]
| On Efficiently Representing Regular Languages as {RNN}s | findings-acl.244 | Poster | 2402.15814v2 |
https://aclanthology.org/2024.findings-acl.245.bib | @inproceedings{reinig-etal-2024-survey,
title = "A Survey on Modelling Morality for Text Analysis",
author = "Reinig, Ines and
Becker, Maria and
Rehbein, Ines and
Ponzetto, Simone",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.245",
pages = "4136--4155",
abstract = "In this survey, we provide a systematic review of recent work on modelling morality in text, an area of research that has garnered increasing attention in recent years. Our survey is motivated by the importance of modelling decisions on the created resources, the models trained on these resources and the analyses that result from the models{'} predictions. We review work at the interface of NLP, Computational Social Science and Psychology and give an overview of the different goals and research questions addressed in the papers, their underlying theoretical backgrounds and the methods that have been applied to pursue these goals. We then identify and discuss challenges and research gaps, such as the lack of a theoretical framework underlying the operationalisation of morality in text, the low IAA reported for manyhuman-annotated resulting resources and the lack of validation of newly proposed resources and analyses.",
}
| In this survey, we provide a systematic review of recent work on modelling morality in text, an area of research that has garnered increasing attention in recent years. Our survey is motivated by the importance of modelling decisions on the created resources, the models trained on these resources and the analyses that result from the models{'} predictions. We review work at the interface of NLP, Computational Social Science and Psychology and give an overview of the different goals and research questions addressed in the papers, their underlying theoretical backgrounds and the methods that have been applied to pursue these goals. We then identify and discuss challenges and research gaps, such as the lack of a theoretical framework underlying the operationalisation of morality in text, the low IAA reported for manyhuman-annotated resulting resources and the lack of validation of newly proposed resources and analyses. | [
"Reinig, Ines",
"Becker, Maria",
"Rehbein, Ines",
"Ponzetto, Simone"
]
| A Survey on Modelling Morality for Text Analysis | findings-acl.245 | Poster | 2404.01715v1 |
https://aclanthology.org/2024.findings-acl.246.bib | @inproceedings{chen-etal-2024-vision,
title = "Your Vision-Language Model Itself Is a Strong Filter: Towards High-Quality Instruction Tuning with Data Selection",
author = "Chen, Ruibo and
Wu, Yihan and
Chen, Lichang and
Liu, Guodong and
He, Qi and
Xiong, Tianyi and
Liu, Chenxi and
Guo, Junfeng and
Huang, Heng",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.246",
pages = "4156--4172",
abstract = "Data selection in instruction tuning emerges as a pivotal process for acquiring high-quality data and training instruction-following large language models (LLMs), but it is still a new and unexplored research area for vision-language models (VLMs). Existing data selection approaches on LLMs either rely on single unreliable scores, or use downstream tasks for selection, which is time-consuming and can lead to potential over-fitting on the chosen evaluation datasets. To address this challenge, we introduce a novel dataset selection method, Self-Filter, that utilizes the VLM itself as a filter. This approach is inspired by the observation that VLMs benefit from training with the most challenging instructions. Self-Filter operates in two stages. In the first stage, we devise a scoring network to evaluate the difficulty of training instructions, which is co-trained with the VLM. In the second stage, we use the trained score net to measure the difficulty of each instruction, select the most challenging samples, and penalize similar samples to encourage diversity. Comprehensive experiments on LLaVA and MiniGPT-4 show that Self-Filter can reach better results compared to full data settings with merely about 15{\%} samples, and can achieve superior performance against competitive baselines.",
}
| Data selection in instruction tuning emerges as a pivotal process for acquiring high-quality data and training instruction-following large language models (LLMs), but it is still a new and unexplored research area for vision-language models (VLMs). Existing data selection approaches on LLMs either rely on single unreliable scores, or use downstream tasks for selection, which is time-consuming and can lead to potential over-fitting on the chosen evaluation datasets. To address this challenge, we introduce a novel dataset selection method, Self-Filter, that utilizes the VLM itself as a filter. This approach is inspired by the observation that VLMs benefit from training with the most challenging instructions. Self-Filter operates in two stages. In the first stage, we devise a scoring network to evaluate the difficulty of training instructions, which is co-trained with the VLM. In the second stage, we use the trained score net to measure the difficulty of each instruction, select the most challenging samples, and penalize similar samples to encourage diversity. Comprehensive experiments on LLaVA and MiniGPT-4 show that Self-Filter can reach better results compared to full data settings with merely about 15{\%} samples, and can achieve superior performance against competitive baselines. | [
"Chen, Ruibo",
"Wu, Yihan",
"Chen, Lichang",
"Liu, Guodong",
"He, Qi",
"Xiong, Tianyi",
"Liu, Chenxi",
"Guo, Junfeng",
"Huang, Heng"
]
| Your Vision-Language Model Itself Is a Strong Filter: Towards High-Quality Instruction Tuning with Data Selection | findings-acl.246 | Poster | 2402.12501v1 |
https://aclanthology.org/2024.findings-acl.247.bib | @inproceedings{tian-etal-2024-debugbench,
title = "{D}ebug{B}ench: Evaluating Debugging Capability of Large Language Models",
author = "Tian, Runchu and
Ye, Yining and
Qin, Yujia and
Cong, Xin and
Lin, Yankai and
Pan, Yinxu and
Wu, Yesai and
Haotian, Hui and
Weichuan, Liu and
Liu, Zhiyuan and
Sun, Maosong",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.247",
pages = "4173--4198",
abstract = "Large Language Models (LLMs) have demonstrated exceptional coding capability. However, as another critical component of programming proficiency, the debugging capability of LLMs remains relatively unexplored. Previous evaluations of LLMs{'} debugging ability are significantly limited by the risk of data leakage, the scale of the dataset, and the variety of tested bugs. To overcome these deficiencies, we introduce {`}DebugBench{'}, an LLM debugging benchmark consisting of 4,253 instances. It covers four major bug categories and 18 minor types in C++, Java, and Python. To construct DebugBench, we collect code snippets from the LeetCode community, implant bugs into source data with GPT-4, and assure rigorous quality checks. We evaluate two commercial and four open-source models in a zero-shot scenario. We find that (1) while closed-source models exhibit inferior debugging performance compared to humans, open-source models relatively lower pass rate scores; (2) the complexity of debugging notably fluctuates depending on the bug category; (3) incorporating runtime feedback has a clear impact on debugging performance which is not always helpful. As an extension, we also compare LLM debugging and code generation, revealing a strong correlation between them for closed-source models. These findings will benefit the development of LLMs in debugging.",
}
| Large Language Models (LLMs) have demonstrated exceptional coding capability. However, as another critical component of programming proficiency, the debugging capability of LLMs remains relatively unexplored. Previous evaluations of LLMs{'} debugging ability are significantly limited by the risk of data leakage, the scale of the dataset, and the variety of tested bugs. To overcome these deficiencies, we introduce {`}DebugBench{'}, an LLM debugging benchmark consisting of 4,253 instances. It covers four major bug categories and 18 minor types in C++, Java, and Python. To construct DebugBench, we collect code snippets from the LeetCode community, implant bugs into source data with GPT-4, and assure rigorous quality checks. We evaluate two commercial and four open-source models in a zero-shot scenario. We find that (1) while closed-source models exhibit inferior debugging performance compared to humans, open-source models relatively lower pass rate scores; (2) the complexity of debugging notably fluctuates depending on the bug category; (3) incorporating runtime feedback has a clear impact on debugging performance which is not always helpful. As an extension, we also compare LLM debugging and code generation, revealing a strong correlation between them for closed-source models. These findings will benefit the development of LLMs in debugging. | [
"Tian, Runchu",
"Ye, Yining",
"Qin, Yujia",
"Cong, Xin",
"Lin, Yankai",
"Pan, Yinxu",
"Wu, Yesai",
"Haotian, Hui",
"Weichuan, Liu",
"Liu, Zhiyuan",
"Sun, Maosong"
]
| {D}ebug{B}ench: Evaluating Debugging Capability of Large Language Models | findings-acl.247 | Poster | 2408.05006v1 |
https://aclanthology.org/2024.findings-acl.248.bib | @inproceedings{zhou-etal-2024-pop,
title = "{POP}-{CEE}: Position-oriented Prompt-tuning Model for Causal Emotion Entailment",
author = "Zhou, Zhihan and
Gu, Xue and
Zhao, Yujie and
Xu, Hao",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.248",
pages = "4199--4210",
abstract = "The objective of the Causal Emotion Entailment (CEE) task is to identify the causes of the target emotional utterances in a given conversation. Most existing studies have focused on a fine-tuning paradigm based on a pretrained model, e.g., the BERT model. However, there are gaps between the pretrained task and the CEE task. Although a pretrained model enhances contextual comprehension to some extent, it cannot acquire specific knowledge that is relevant to the CEE task. In addition, in a typical CEE task, there are peculiarities in the distribution of the positions with different emotion types of emotion utterances and cause utterances in conversations. Existing methods employ a fixed-size window to capture the relationship between neighboring conversations; however, these methods ignore the specific semantic associations between emotions and cause utterances. To address these issues, we propose the Position-oriented Prompt-tuning (POP-CEE) model to solve the CEE task in an end-to-end manner. Specifically, we can model the CEE task by designing prompts with multiple unified goals and by exploring the positional relationship between emotion and cause utterances using a position constraint module. Experimental results demonstrate that the proposed POP-CEE model achieves state-of-the-art performance on a benchmark dataset. Ourcode and data can be found at: https://github.com/Zh0uzh/POP-CEE.",
}
| The objective of the Causal Emotion Entailment (CEE) task is to identify the causes of the target emotional utterances in a given conversation. Most existing studies have focused on a fine-tuning paradigm based on a pretrained model, e.g., the BERT model. However, there are gaps between the pretrained task and the CEE task. Although a pretrained model enhances contextual comprehension to some extent, it cannot acquire specific knowledge that is relevant to the CEE task. In addition, in a typical CEE task, there are peculiarities in the distribution of the positions with different emotion types of emotion utterances and cause utterances in conversations. Existing methods employ a fixed-size window to capture the relationship between neighboring conversations; however, these methods ignore the specific semantic associations between emotions and cause utterances. To address these issues, we propose the Position-oriented Prompt-tuning (POP-CEE) model to solve the CEE task in an end-to-end manner. Specifically, we can model the CEE task by designing prompts with multiple unified goals and by exploring the positional relationship between emotion and cause utterances using a position constraint module. Experimental results demonstrate that the proposed POP-CEE model achieves state-of-the-art performance on a benchmark dataset. Ourcode and data can be found at: https://github.com/Zh0uzh/POP-CEE. | [
"Zhou, Zhihan",
"Gu, Xue",
"Zhao, Yujie",
"Xu, Hao"
]
| {POP}-{CEE}: Position-oriented Prompt-tuning Model for Causal Emotion Entailment | findings-acl.248 | Poster | 2205.00759v2 |
https://aclanthology.org/2024.findings-acl.249.bib | @inproceedings{li-huaping-2024-context,
title = "Context Length Extension via Generalized Extrapolation Scale",
author = "Li, Linhan and
Huaping, Zhang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.249",
pages = "4211--4218",
abstract = "Context length expansion of transformer models is considered a key challenge, especially when handling context beyond the training length during inference stage. In this paper, we propose $\textbf{Ge}$eneralized extrapolatio$\textbf{N}$ scal$\textbf{E}$ (GeNE), a set of parameterized extrapolation functions applied to each layer and attention head to adaptively adjust its extrapolation scales. Experimental results show that GeNE provides a significant improvement on long context language modeling. By randomly scaling the extrapolation ratio during the finetuning, GeNE achieves stable extrapolation on 64k contexts by training on 16k length text. Further, the instruction following Llama2 model based on GeNE achieved competitive results compared with other open-source models of the same parameter scale.",
}
| Context length expansion of transformer models is considered a key challenge, especially when handling context beyond the training length during inference stage. In this paper, we propose $\textbf{Ge}$eneralized extrapolatio$\textbf{N}$ scal$\textbf{E}$ (GeNE), a set of parameterized extrapolation functions applied to each layer and attention head to adaptively adjust its extrapolation scales. Experimental results show that GeNE provides a significant improvement on long context language modeling. By randomly scaling the extrapolation ratio during the finetuning, GeNE achieves stable extrapolation on 64k contexts by training on 16k length text. Further, the instruction following Llama2 model based on GeNE achieved competitive results compared with other open-source models of the same parameter scale. | [
"Li, Linhan",
"Huaping, Zhang"
]
| Context Length Extension via Generalized Extrapolation Scale | findings-acl.249 | Poster | 2405.18009v1 |
https://aclanthology.org/2024.findings-acl.250.bib | @inproceedings{eisenschlos-etal-2024-selectively,
title = "Selectively Answering Visual Questions",
author = "Eisenschlos, Julian and
Maina, Hern{\'a}n and
Ivetta, Guido and
Benotti, Luciana",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.250",
pages = "4219--4229",
abstract = "Recently, large multi-modal models (LMMs) have emerged with the capacity to perform vision tasks such as captioning and visual question answering (VQA) with unprecedented accuracy. Applications such as helping the blind or visually impaired have a critical need for precise answers. It is specially important for models to be well calibrated and be able to quantify their uncertainty in order to selectively decide when to answer and when to abstain or ask for clarifications. We perform the first in-depth analysis of calibration methods and metrics for VQA with in-context learning LMMs. Studying VQA on two answerability benchmarks, we show that the likelihood score of visually grounded models is better calibrated than in their text-only counterparts for in-context learning, where sampling based methods are generally superior, but no clear winner arises. We propose Avg BLEU, a calibration score combining the benefits of both sampling and likelihood methods across modalities.",
}
| Recently, large multi-modal models (LMMs) have emerged with the capacity to perform vision tasks such as captioning and visual question answering (VQA) with unprecedented accuracy. Applications such as helping the blind or visually impaired have a critical need for precise answers. It is specially important for models to be well calibrated and be able to quantify their uncertainty in order to selectively decide when to answer and when to abstain or ask for clarifications. We perform the first in-depth analysis of calibration methods and metrics for VQA with in-context learning LMMs. Studying VQA on two answerability benchmarks, we show that the likelihood score of visually grounded models is better calibrated than in their text-only counterparts for in-context learning, where sampling based methods are generally superior, but no clear winner arises. We propose Avg BLEU, a calibration score combining the benefits of both sampling and likelihood methods across modalities. | [
"Eisenschlos, Julian",
"Maina, Hern{\\'a}n",
"Ivetta, Guido",
"Benotti, Luciana"
]
| Selectively Answering Visual Questions | findings-acl.250 | Poster | 2106.04605v1 |
https://aclanthology.org/2024.findings-acl.251.bib | @inproceedings{liu-etal-2024-wav2sql,
title = "{W}av2{SQL}: Direct Generalizable Speech-To-{SQL} Parsing",
author = "Liu, Huadai and
Huang, Rongjie and
He, Jinzheng and
Sun, Gang and
Shen, Ran and
Cheng, Xize and
Zhao, Zhou",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.251",
pages = "4230--4242",
abstract = "We release a multi-accent dataset and propose speech-programming and gradient reversal classifier to improve the generalization.Abstract: Speech-to-SQL (S2SQL) aims to convert spoken questions into SQL queries given relational databases, which has been traditionally implemented in a cascaded manner while facing the following challenges: 1) model training is faced with the major issue of data scarcity, where limited parallel data is available; and 2) the systems should be robust enough to handle diverse out-of-domain speech samples that differ from the source data. In this work, we propose the direct generalizable speech-to-SQL parsing model Wav2SQL which avoids error compounding across cascaded systems. Specifically, 1) to accelerate speech-driven SQL parsing research in the community, we release a large-scale and multi-accent dataset MASpider; 2) leveraging the recent progress in the large-scale pre-training, we show that it alleviates the data scarcity issue and allow for direct speech-to-SQL parsing; and 3) we include the speech re-programming and gradient reversal classifier techniques to reduce acoustic variance and learned style-agnostic representation, improving generalization to unseen out-of-domain custom data. Experimental results demonstrate that Wav2SQL avoids error compounding and achieves state-of-the-art results by up to 4.7{\%} accuracy improvement over the baseline.",
}
| We release a multi-accent dataset and propose speech-programming and gradient reversal classifier to improve the generalization.Abstract: Speech-to-SQL (S2SQL) aims to convert spoken questions into SQL queries given relational databases, which has been traditionally implemented in a cascaded manner while facing the following challenges: 1) model training is faced with the major issue of data scarcity, where limited parallel data is available; and 2) the systems should be robust enough to handle diverse out-of-domain speech samples that differ from the source data. In this work, we propose the direct generalizable speech-to-SQL parsing model Wav2SQL which avoids error compounding across cascaded systems. Specifically, 1) to accelerate speech-driven SQL parsing research in the community, we release a large-scale and multi-accent dataset MASpider; 2) leveraging the recent progress in the large-scale pre-training, we show that it alleviates the data scarcity issue and allow for direct speech-to-SQL parsing; and 3) we include the speech re-programming and gradient reversal classifier techniques to reduce acoustic variance and learned style-agnostic representation, improving generalization to unseen out-of-domain custom data. Experimental results demonstrate that Wav2SQL avoids error compounding and achieves state-of-the-art results by up to 4.7{\%} accuracy improvement over the baseline. | [
"Liu, Huadai",
"Huang, Rongjie",
"He, Jinzheng",
"Sun, Gang",
"Shen, Ran",
"Cheng, Xize",
"Zhao, Zhou"
]
| {W}av2{SQL}: Direct Generalizable Speech-To-{SQL} Parsing | findings-acl.251 | Poster | 2210.12674v1 |
https://aclanthology.org/2024.findings-acl.252.bib | @inproceedings{liu-etal-2024-e2,
title = "E2-{LLM}: Efficient and Extreme Length Extension of Large Language Models",
author = "Liu, Jiaheng and
ZhiqiBai, ZhiqiBai and
Zhang, Yuanxing and
Zhang, Chenchen and
YuangZh, YuangZh and
Zhang, Ge and
JiakaiWang, JiakaiWang and
Que, Haoran and
Chen, Yukang and
Su, Wenbo and
Ge, Tiezheng and
Fu, Jie and
Chen, Wenhu and
Zheng, Bo",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.252",
pages = "4243--4253",
abstract = "Training Large Language Models (LLMs) to process extensive context lengths incurs prohibitive computational costs. Prevailing techniques for extending context capabilities in LLMs typically require not only additional training procedures but also access to datasets with long context (e.g., sequences of 32K tokens), presupposing substantial GPU expenditures. To address the aforementioned issues, we introduce a novel solution named Efficient and Extreme length extension for Large Language Models (E2-LLM). E2-LLM entails a singular training process over considerably short sequences (e.g., 4K tokens), which greatly mitigates the cost of continual-pretraining or fine-tuning. Within the training phase, we incorporate a dual augmentation strategy with Rotary Position Embeddings (RoPE) that adjusts the scale and position indices across distinct training samples. E 2 -LLM is meticulously designed to enhance the model{'}s robustness to diverse relative positions. The experimental results on multiple benchmark datasets demonstrate the superior performance of E 2 -LLM on demanding tasks of processing long contexts.",
}
| Training Large Language Models (LLMs) to process extensive context lengths incurs prohibitive computational costs. Prevailing techniques for extending context capabilities in LLMs typically require not only additional training procedures but also access to datasets with long context (e.g., sequences of 32K tokens), presupposing substantial GPU expenditures. To address the aforementioned issues, we introduce a novel solution named Efficient and Extreme length extension for Large Language Models (E2-LLM). E2-LLM entails a singular training process over considerably short sequences (e.g., 4K tokens), which greatly mitigates the cost of continual-pretraining or fine-tuning. Within the training phase, we incorporate a dual augmentation strategy with Rotary Position Embeddings (RoPE) that adjusts the scale and position indices across distinct training samples. E 2 -LLM is meticulously designed to enhance the model{'}s robustness to diverse relative positions. The experimental results on multiple benchmark datasets demonstrate the superior performance of E 2 -LLM on demanding tasks of processing long contexts. | [
"Liu, Jiaheng",
"ZhiqiBai, ZhiqiBai",
"Zhang, Yuanxing",
"Zhang, Chenchen",
"YuangZh, YuangZh",
"Zhang, Ge",
"JiakaiWang, JiakaiWang",
"Que, Haoran",
"Chen, Yukang",
"Su, Wenbo",
"Ge, Tiezheng",
"Fu, Jie",
"Chen, Wenhu",
"Zheng, Bo"
]
| E2-{LLM}: Efficient and Extreme Length Extension of Large Language Models | findings-acl.252 | Poster | 2401.06951v3 |
https://aclanthology.org/2024.findings-acl.253.bib | @inproceedings{ju-etal-2024-female,
title = "Are Female Carpenters like Blue Bananas? A Corpus Investigation of Occupation Gender Typicality",
author = "Ju, Da and
Ullrich, Karen and
Williams, Adina",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.253",
pages = "4254--4274",
abstract = "People tend to use language to mention surprising properties of events: for example, when a banana is blue, we are more likely to mention color than when it is yellow. This fact is taken to suggest that yellowness is somehow a typical feature of bananas, and blueness is exceptional. Similar to how a yellow color is typical of bananas, there may also be genders that are typical of occupations. In this work, we explore this question using information theoretic techniques coupled with corpus statistic analysis. In two distinct large corpora, we do not find strong evidence that occupations and gender display the same patterns of mentioning as do bananas and color. Instead, we find that gender mentioning is correlated with femaleness of occupation in particular, suggesting perhaps that woman-dominated occupations are seen as somehow {``}more gendered{''} than male-dominated ones, and thereby they encourage more gender mentioning overall.",
}
| People tend to use language to mention surprising properties of events: for example, when a banana is blue, we are more likely to mention color than when it is yellow. This fact is taken to suggest that yellowness is somehow a typical feature of bananas, and blueness is exceptional. Similar to how a yellow color is typical of bananas, there may also be genders that are typical of occupations. In this work, we explore this question using information theoretic techniques coupled with corpus statistic analysis. In two distinct large corpora, we do not find strong evidence that occupations and gender display the same patterns of mentioning as do bananas and color. Instead, we find that gender mentioning is correlated with femaleness of occupation in particular, suggesting perhaps that woman-dominated occupations are seen as somehow {``}more gendered{''} than male-dominated ones, and thereby they encourage more gender mentioning overall. | [
"Ju, Da",
"Ullrich, Karen",
"Williams, Adina"
]
| Are Female Carpenters like Blue Bananas? A Corpus Investigation of Occupation Gender Typicality | findings-acl.253 | Poster | 2408.02948v1 |
https://aclanthology.org/2024.findings-acl.254.bib | @inproceedings{cheng-etal-2024-call,
title = "Call Me When Necessary: {LLM}s can Efficiently and Faithfully Reason over Structured Environments",
author = "Cheng, Sitao and
Zhuang, Ziyuan and
Xu, Yong and
Yang, Fangkai and
Zhang, Chaoyun and
Qin, Xiaoting and
Huang, Xiang and
Chen, Ling and
Lin, Qingwei and
Zhang, Dongmei and
Rajmohan, Saravan and
Zhang, Qi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.254",
pages = "4275--4295",
abstract = "Large Language Models (LLMs) have shown potential in reasoning over structured environments, e.g., knowledge graphs and tables. Such tasks typically require multi-hop reasoning, i.e., match natural language utterance with instances in the environment. Previous works adopt LLMs to incrementally build a reasoning path, where LLMs either invoke tools or pick up items by step-by-step interacting with the environment. We propose Reasoning-Path-Editing (Readi), a novel framework where LLMs can efficiently and faithfully reason over structured environments. In Readi, LLMs initially generate a reasoning path given a query, and edit the path only when necessary. We instantiate the path on structured environments and provide feedback to edit the path if anything goes wrong. Experimental results on three KGQA and two TableQA datasets show the effectiveness of Readi, significantly surpassing previous LLM-based methods (by 9.1{\%} Hit@1 on WebQSP, 12.4{\%} on MQA-3H and 9.5{\%} on WTQ), comparable with state-of-the-art fine-tuned methods (67{\%} on CWQ and 74.7{\%} on WebQSP) and substantially boosting the vanilla LLMs (by 14.9{\%} on CWQ). Our code will be available on \url{https://aka.ms/readi}.",
}
| Large Language Models (LLMs) have shown potential in reasoning over structured environments, e.g., knowledge graphs and tables. Such tasks typically require multi-hop reasoning, i.e., match natural language utterance with instances in the environment. Previous works adopt LLMs to incrementally build a reasoning path, where LLMs either invoke tools or pick up items by step-by-step interacting with the environment. We propose Reasoning-Path-Editing (Readi), a novel framework where LLMs can efficiently and faithfully reason over structured environments. In Readi, LLMs initially generate a reasoning path given a query, and edit the path only when necessary. We instantiate the path on structured environments and provide feedback to edit the path if anything goes wrong. Experimental results on three KGQA and two TableQA datasets show the effectiveness of Readi, significantly surpassing previous LLM-based methods (by 9.1{\%} Hit@1 on WebQSP, 12.4{\%} on MQA-3H and 9.5{\%} on WTQ), comparable with state-of-the-art fine-tuned methods (67{\%} on CWQ and 74.7{\%} on WebQSP) and substantially boosting the vanilla LLMs (by 14.9{\%} on CWQ). Our code will be available on \url{https://aka.ms/readi}. | [
"Cheng, Sitao",
"Zhuang, Ziyuan",
"Xu, Yong",
"Yang, Fangkai",
"Zhang, Chaoyun",
"Qin, Xiaoting",
"Huang, Xiang",
"Chen, Ling",
"Lin, Qingwei",
"Zhang, Dongmei",
"Rajmohan, Saravan",
"Zhang, Qi"
]
| Call Me When Necessary: {LLM}s can Efficiently and Faithfully Reason over Structured Environments | findings-acl.254 | Poster | 2403.08593v2 |
https://aclanthology.org/2024.findings-acl.255.bib | @inproceedings{nigam-etal-2024-legal,
title = "Legal Judgment Reimagined: {P}red{E}x and the Rise of Intelligent {AI} Interpretation in {I}ndian Courts",
author = "Nigam, Shubham and
Sharma, Anurag and
Khanna, Danush and
Shallum, Noel and
Ghosh, Kripabandhu and
Bhattacharya, Arnab",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.255",
pages = "4296--4315",
abstract = "In the era of Large Language Models (LLMs), predicting judicial outcomes poses significant challenges due to the complexity of legal proceedings and the scarcity of expert-annotated datasets. Addressing this, we introduce \textbf{Pred}iction with \textbf{Ex}planation (PredEx), the largest expert-annotated dataset for legal judgment prediction and explanation in the Indian context, featuring over 15,000 annotations. This groundbreaking corpus significantly enhances the training and evaluation of AI models in legal analysis, with innovations including the application of instruction tuning to LLMs. This method has markedly improved the predictive accuracy and explanatory depth of these models for legal judgments. We employed various transformer-based models, tailored for both general and Indian legal contexts. Through rigorous lexical, semantic, and expert assessments, our models effectively leverage PredEx to provide precise predictions and meaningful explanations, establishing it as a valuable benchmark for both the legal profession and the NLP community.",
}
| In the era of Large Language Models (LLMs), predicting judicial outcomes poses significant challenges due to the complexity of legal proceedings and the scarcity of expert-annotated datasets. Addressing this, we introduce \textbf{Pred}iction with \textbf{Ex}planation (PredEx), the largest expert-annotated dataset for legal judgment prediction and explanation in the Indian context, featuring over 15,000 annotations. This groundbreaking corpus significantly enhances the training and evaluation of AI models in legal analysis, with innovations including the application of instruction tuning to LLMs. This method has markedly improved the predictive accuracy and explanatory depth of these models for legal judgments. We employed various transformer-based models, tailored for both general and Indian legal contexts. Through rigorous lexical, semantic, and expert assessments, our models effectively leverage PredEx to provide precise predictions and meaningful explanations, establishing it as a valuable benchmark for both the legal profession and the NLP community. | [
"Nigam, Shubham",
"Sharma, Anurag",
"Khanna, Danush",
"Shallum, Noel",
"Ghosh, Kripab",
"hu",
"Bhattacharya, Arnab"
]
| Legal Judgment Reimagined: {P}red{E}x and the Rise of Intelligent {AI} Interpretation in {I}ndian Courts | findings-acl.255 | Poster | 2406.04136v1 |
https://aclanthology.org/2024.findings-acl.256.bib | @inproceedings{tang-etal-2024-rule,
title = "{R}ul{E}: Knowledge Graph Reasoning with Rule Embedding",
author = "Tang, Xiaojuan and
Zhu, Song-Chun and
Liang, Yitao and
Zhang, Muhan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.256",
pages = "4316--4335",
abstract = "Knowledge graph reasoning is an important problem for knowledge graphs. In this paper, we propose a novel and principled framework called \textbf{RulE} (stands for Rule Embedding) to effectively leverage logical rules to enhance KG reasoning. Unlike knowledge graph embedding methods, RulE learns rule embeddings from existing triplets and first-order rules by jointly representing \textbf{entities}, \textbf{relations} and \textbf{logical rules} in a unified embedding space. Based on the learned rule embeddings, a confidence score can be calculated for each rule, reflecting its consistency with the observed triplets. This allows us to perform logical rule inference in a soft way, thus alleviating the brittleness of logic. On the other hand, RulE injects prior logical rule information into the embedding space, enriching and regularizing the entity/relation embeddings. This makes KGE alone perform better too. RulE is conceptually simple and empirically effective. We conduct extensive experiments to verify each component of RulE.Results on multiple benchmarks reveal that our model outperforms the majority of existing embedding-based and rule-based approaches.",
}
| Knowledge graph reasoning is an important problem for knowledge graphs. In this paper, we propose a novel and principled framework called \textbf{RulE} (stands for Rule Embedding) to effectively leverage logical rules to enhance KG reasoning. Unlike knowledge graph embedding methods, RulE learns rule embeddings from existing triplets and first-order rules by jointly representing \textbf{entities}, \textbf{relations} and \textbf{logical rules} in a unified embedding space. Based on the learned rule embeddings, a confidence score can be calculated for each rule, reflecting its consistency with the observed triplets. This allows us to perform logical rule inference in a soft way, thus alleviating the brittleness of logic. On the other hand, RulE injects prior logical rule information into the embedding space, enriching and regularizing the entity/relation embeddings. This makes KGE alone perform better too. RulE is conceptually simple and empirically effective. We conduct extensive experiments to verify each component of RulE.Results on multiple benchmarks reveal that our model outperforms the majority of existing embedding-based and rule-based approaches. | [
"Tang, Xiaojuan",
"Zhu, Song-Chun",
"Liang, Yitao",
"Zhang, Muhan"
]
| {R}ul{E}: Knowledge Graph Reasoning with Rule Embedding | findings-acl.256 | Poster | 2210.14905v3 |
https://aclanthology.org/2024.findings-acl.257.bib | @inproceedings{nguyen-etal-2024-multi,
title = "Multi-Objective Linguistic Control of Large Language Models",
author = "Nguyen, Dang and
Chen, Jiuhai and
Zhou, Tianyi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.257",
pages = "4336--4347",
abstract = "Large language models (LLMs), despite their breakthroughs on many challenging benchmark tasks, prefer to generate verbose responses and lack the controllability of output complexity, which is usually preferred by human users in practice. In this paper, we study how to precisely control multiple linguistic complexities of LLM output by finetuning using off-the-shelf data. To this end, we propose multi-control tuning (MCTune), which includes multiple linguistic complexity values of ground-truth responses as controls in the input for instruction tuning. We finetune LLaMA2-7B on Alpaca-GPT4 and WizardLM datasets. Evaluations on widely used benchmarks demonstrate that our method does not only improve LLMs{'} multi-complexity controllability substantially but also retains or even enhances the quality of the responses as a side benefit.",
}
| Large language models (LLMs), despite their breakthroughs on many challenging benchmark tasks, prefer to generate verbose responses and lack the controllability of output complexity, which is usually preferred by human users in practice. In this paper, we study how to precisely control multiple linguistic complexities of LLM output by finetuning using off-the-shelf data. To this end, we propose multi-control tuning (MCTune), which includes multiple linguistic complexity values of ground-truth responses as controls in the input for instruction tuning. We finetune LLaMA2-7B on Alpaca-GPT4 and WizardLM datasets. Evaluations on widely used benchmarks demonstrate that our method does not only improve LLMs{'} multi-complexity controllability substantially but also retains or even enhances the quality of the responses as a side benefit. | [
"Nguyen, Dang",
"Chen, Jiuhai",
"Zhou, Tianyi"
]
| Multi-Objective Linguistic Control of Large Language Models | findings-acl.257 | Poster | 2106.01207v1 |
https://aclanthology.org/2024.findings-acl.258.bib | @inproceedings{zhou-etal-2024-evaluating,
title = "Evaluating the Smooth Control of Attribute Intensity in Text Generation with {LLM}s",
author = "Zhou, Shang and
Yao, Feng and
Dong, Chengyu and
Wang, Zihan and
Shang, Jingbo",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.258",
pages = "4348--4362",
abstract = "Controlling the attribute intensity of text generation is crucial across scenarios (e.g., writing conciseness, chatting emotion, and explanation clarity). The remarkable capabilities of large language models (LLMs) have revolutionized text generation, prompting us to explore such \textit{smooth control} of LLM generation. Specifically, we propose metrics to assess the range, calibration, and consistency of the generated text{'}s attribute intensity in response to varying control values, as well as its relevance to the intended context. To quantify the attribute intensity and context relevance, we leverage an Elo rating system and GPT4, respectively, both renowned for their robust alignment with human judgment. We look into two viable training-free methods for achieving smooth control of LLMs: (1) Prompting with semantic shifters, and (2) Modifying internal model representations. The evaluations of these two methods are conducted on 5 different attributes with various models.",
}
| Controlling the attribute intensity of text generation is crucial across scenarios (e.g., writing conciseness, chatting emotion, and explanation clarity). The remarkable capabilities of large language models (LLMs) have revolutionized text generation, prompting us to explore such \textit{smooth control} of LLM generation. Specifically, we propose metrics to assess the range, calibration, and consistency of the generated text{'}s attribute intensity in response to varying control values, as well as its relevance to the intended context. To quantify the attribute intensity and context relevance, we leverage an Elo rating system and GPT4, respectively, both renowned for their robust alignment with human judgment. We look into two viable training-free methods for achieving smooth control of LLMs: (1) Prompting with semantic shifters, and (2) Modifying internal model representations. The evaluations of these two methods are conducted on 5 different attributes with various models. | [
"Zhou, Shang",
"Yao, Feng",
"Dong, Chengyu",
"Wang, Zihan",
"Shang, Jingbo"
]
| Evaluating the Smooth Control of Attribute Intensity in Text Generation with {LLM}s | findings-acl.258 | Poster | 2406.04460v1 |
https://aclanthology.org/2024.findings-acl.259.bib | @inproceedings{huang-etal-2024-planning-creation,
title = "Planning, Creation, Usage: Benchmarking {LLM}s for Comprehensive Tool Utilization in Real-World Complex Scenarios",
author = "Huang, Shijue and
Zhong, Wanjun and
Lu, Jianqiao and
Zhu, Qi and
Gao, Jiahui and
Liu, Weiwen and
Hou, Yutai and
Zeng, Xingshan and
Wang, Yasheng and
Shang, Lifeng and
Jiang, Xin and
Xu, Ruifeng and
Liu, Qun",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.259",
pages = "4363--4400",
abstract = "The recent trend of using Large Language Models (LLMs) as tool agents in real-world applications underscores the necessity for comprehensive evaluations of their capabilities, particularly in complex scenarios involving planning, creating, and using tools. However, existing benchmarks typically focus on simple synthesized queries that do not reflect real-world complexity, thereby offering limited perspectives in evaluating tool utilization. To address this issue, we present UltraTool, a novel benchmark designed to improve and evaluate LLMs{'} ability in tool utilization within real-world scenarios. UltraTool focuses on the entire process of using tools - from planning and creating to applying them in complex tasks. It emphasizes real-world complexities, demanding accurate, multi-step planning for effective problem-solving. A key feature of UltraTool is its independent evaluation of planning with natural language, which happens before tool usage and simplifies the task solving by mapping out the intermediate steps. Thus, unlike previous work, it eliminates the restriction of pre-defined toolset. Through extensive experiments on various LLMs, we offer novel insights into the evaluation of capabilities of LLMs in tool utilization, thereby contributing a fresh perspective to this rapidly evolving field. The benchmark is publicly available at https://github.com/JoeYing1019/UltraTool.",
}
| The recent trend of using Large Language Models (LLMs) as tool agents in real-world applications underscores the necessity for comprehensive evaluations of their capabilities, particularly in complex scenarios involving planning, creating, and using tools. However, existing benchmarks typically focus on simple synthesized queries that do not reflect real-world complexity, thereby offering limited perspectives in evaluating tool utilization. To address this issue, we present UltraTool, a novel benchmark designed to improve and evaluate LLMs{'} ability in tool utilization within real-world scenarios. UltraTool focuses on the entire process of using tools - from planning and creating to applying them in complex tasks. It emphasizes real-world complexities, demanding accurate, multi-step planning for effective problem-solving. A key feature of UltraTool is its independent evaluation of planning with natural language, which happens before tool usage and simplifies the task solving by mapping out the intermediate steps. Thus, unlike previous work, it eliminates the restriction of pre-defined toolset. Through extensive experiments on various LLMs, we offer novel insights into the evaluation of capabilities of LLMs in tool utilization, thereby contributing a fresh perspective to this rapidly evolving field. The benchmark is publicly available at https://github.com/JoeYing1019/UltraTool. | [
"Huang, Shijue",
"Zhong, Wanjun",
"Lu, Jianqiao",
"Zhu, Qi",
"Gao, Jiahui",
"Liu, Weiwen",
"Hou, Yutai",
"Zeng, Xingshan",
"Wang, Yasheng",
"Shang, Lifeng",
"Jiang, Xin",
"Xu, Ruifeng",
"Liu, Qun"
]
| Planning, Creation, Usage: Benchmarking {LLM}s for Comprehensive Tool Utilization in Real-World Complex Scenarios | findings-acl.259 | Poster | 2401.17167v3 |
https://aclanthology.org/2024.findings-acl.260.bib | @inproceedings{ch-wang-etal-2024-androids,
title = "Do Androids Know They{'}re Only Dreaming of Electric Sheep?",
author = "CH-Wang, Sky and
Van Durme, Benjamin and
Eisner, Jason and
Kedzie, Chris",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.260",
pages = "4401--4420",
abstract = "We design probes trained on the internal representations of a transformer language model to predict its hallucinatory behavior on three grounded generation tasks. To train the probes, we annotate for span-level hallucination on both sampled (organic) and manually edited (synthetic) reference outputs. Our probes are narrowly trained and we find that they are sensitive to their training domain: they generalize poorly from one task to another or from synthetic to organic hallucinations. However, on in-domain data, they can reliably detect hallucinations at many transformer layers, achieving 95{\%} of their peak performance as early as layer 4. Here, probing proves accurate for evaluating hallucination, outperforming several contemporary baselines and even surpassing an expert human annotator in response-level detection F1. Similarly, on span-level labeling, probes are on par or better than the expert annotator on two out of three generation tasks. Overall, we find that probing is a feasible and efficient alternative to language model hallucination evaluation when model states are available.",
}
| We design probes trained on the internal representations of a transformer language model to predict its hallucinatory behavior on three grounded generation tasks. To train the probes, we annotate for span-level hallucination on both sampled (organic) and manually edited (synthetic) reference outputs. Our probes are narrowly trained and we find that they are sensitive to their training domain: they generalize poorly from one task to another or from synthetic to organic hallucinations. However, on in-domain data, they can reliably detect hallucinations at many transformer layers, achieving 95{\%} of their peak performance as early as layer 4. Here, probing proves accurate for evaluating hallucination, outperforming several contemporary baselines and even surpassing an expert human annotator in response-level detection F1. Similarly, on span-level labeling, probes are on par or better than the expert annotator on two out of three generation tasks. Overall, we find that probing is a feasible and efficient alternative to language model hallucination evaluation when model states are available. | [
"CH-Wang, Sky",
"Van Durme, Benjamin",
"Eisner, Jason",
"Kedzie, Chris"
]
| Do Androids Know They{'}re Only Dreaming of Electric Sheep? | findings-acl.260 | Poster | 2312.17249v2 |
https://aclanthology.org/2024.findings-acl.261.bib | @inproceedings{lv-etal-2024-urg,
title = "{URG}: A Unified Ranking and Generation Method for Ensembling Language Models",
author = "Lv, Bo and
Tang, Chen and
Zhang, Yanan and
Liu, Xin and
Luo, Ping and
Yu, Yue",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.261",
pages = "4421--4434",
abstract = "Prior research endeavors of the ensemble Large Language Models (LLMs) achieved great success by employing an individual language model (LM) rank before the text generation. However, the use of an individual LM ranker faces two primary challenges: (1) The time-intensive nature of the ranking process, stemming from the comparisons between models; (2) The issue of error propagation arising from the separate ranking and generation models within the framework. In order to overcome these challenges, we propose a novel ensemble framework, namely Unified Ranking and Generation (URG). URG represents an end-to-end framework that jointly ranks the outputs of LLMs and generates fine-grained fusion results, via utilizing a dedicated cross-attention-based module and noise mitigation training against irrelevant information stemming from bad ranking results. Through extensive experimentation and evaluation, we demonstrate the efficiency and effectiveness of our framework in both the ranking and generation tasks. With the close coordination of the ranking and generation modules, our end-to-end framework achieves the state-of-the-art (SOTA) performance on these tasks, and exhibits substantial enhancements to any of the ensembled models.",
}
| Prior research endeavors of the ensemble Large Language Models (LLMs) achieved great success by employing an individual language model (LM) rank before the text generation. However, the use of an individual LM ranker faces two primary challenges: (1) The time-intensive nature of the ranking process, stemming from the comparisons between models; (2) The issue of error propagation arising from the separate ranking and generation models within the framework. In order to overcome these challenges, we propose a novel ensemble framework, namely Unified Ranking and Generation (URG). URG represents an end-to-end framework that jointly ranks the outputs of LLMs and generates fine-grained fusion results, via utilizing a dedicated cross-attention-based module and noise mitigation training against irrelevant information stemming from bad ranking results. Through extensive experimentation and evaluation, we demonstrate the efficiency and effectiveness of our framework in both the ranking and generation tasks. With the close coordination of the ranking and generation modules, our end-to-end framework achieves the state-of-the-art (SOTA) performance on these tasks, and exhibits substantial enhancements to any of the ensembled models. | [
"Lv, Bo",
"Tang, Chen",
"Zhang, Yanan",
"Liu, Xin",
"Luo, Ping",
"Yu, Yue"
]
| {URG}: A Unified Ranking and Generation Method for Ensembling Language Models | findings-acl.261 | Poster | 1312.7077v2 |
https://aclanthology.org/2024.findings-acl.262.bib | @inproceedings{gourav-etal-2024-multi,
title = "Multi-Modal Retrieval For Large Language Model Based Speech Recognition",
author = "Gourav, Aditya and
Kolehmainen, Jari and
Shivakumar, Prashanth and
Gu, Yile and
Strimel, Grant and
Gandhe, Ankur and
Rastrow, Ariya and
Bulyko, Ivan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.262",
pages = "4435--4446",
abstract = "Retrieval is a widely adopted approach for improving language models leveraging external information. As the field moves towards multi-modal large language models, it is important to extend the pure text based methods to incorporate other modalities in retrieval as well for applications across the wide spectrum of machine learning tasks and data types. In this work, we propose multi-modal retrieval with two approaches: kNN-LM and cross-attention techniques. We demonstrate the effectiveness of our retrieval approaches empirically by applying them to automatic speech recognition tasks with access to external information. Under this setting, we show that speech-based multi-modal retrieval outperforms text based retrieval, and yields up to improvement in word error rate over the multi-modal language model baseline. Furthermore, we achieve state-of-the-art recognition results on the Spoken-Squad question answering dataset.",
}
| Retrieval is a widely adopted approach for improving language models leveraging external information. As the field moves towards multi-modal large language models, it is important to extend the pure text based methods to incorporate other modalities in retrieval as well for applications across the wide spectrum of machine learning tasks and data types. In this work, we propose multi-modal retrieval with two approaches: kNN-LM and cross-attention techniques. We demonstrate the effectiveness of our retrieval approaches empirically by applying them to automatic speech recognition tasks with access to external information. Under this setting, we show that speech-based multi-modal retrieval outperforms text based retrieval, and yields up to improvement in word error rate over the multi-modal language model baseline. Furthermore, we achieve state-of-the-art recognition results on the Spoken-Squad question answering dataset. | [
"Gourav, Aditya",
"Kolehmainen, Jari",
"Shivakumar, Prashanth",
"Gu, Yile",
"Strimel, Grant",
"G",
"he, Ankur",
"Rastrow, Ariya",
"Bulyko, Ivan"
]
| Multi-Modal Retrieval For Large Language Model Based Speech Recognition | findings-acl.262 | Poster | 0206037v1 |
https://aclanthology.org/2024.findings-acl.263.bib | @inproceedings{zhao-etal-2024-loraretriever,
title = "{L}ora{R}etriever: Input-Aware {L}o{RA} Retrieval and Composition for Mixed Tasks in the Wild",
author = "Zhao, Ziyu and
Gan, Leilei and
Wang, Guoyin and
Zhou, Wangchunshu and
Yang, Hongxia and
Kuang, Kun and
Wu, Fei",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.263",
pages = "4447--4462",
abstract = "Low-Rank Adaptation (LoRA) provides an effective yet efficient solution for fine-tuning large language models (LLMs). The modular and plug-and-play nature of LoRA enables the integration of diverse domain-specific LoRAs to enhance the capabilities of LLMs. Previous research on exploiting multiple LoRAs either focuses on specific isolated downstream tasks or fixes the selection of LoRAs during training. However, in real-world scenarios, LLMs receive diverse prompts covering different tasks, and the pool of candidate LoRAs is often dynamically updated. To bridge this gap, we propose LoraRetriever, a retrieve-then-compose framework that adaptively retrieves and composes multiple LoRAs according to the input prompts. LoraRetriever contains three main components: firstly, identifying and retrieving LoRAs relevant to the given input; secondly, formulating strategies for effectively integrating the retrieved LoRAs; and thirdly, developing efficient batch inference to accommodate heterogeneous requests. Experimental results indicate that LoraRetriever consistently outperforms the baselines, highlighting its practical effectiveness and versatility. Our code is available at https://github.com/StyxXuan/LoraRetriever.",
}
| Low-Rank Adaptation (LoRA) provides an effective yet efficient solution for fine-tuning large language models (LLMs). The modular and plug-and-play nature of LoRA enables the integration of diverse domain-specific LoRAs to enhance the capabilities of LLMs. Previous research on exploiting multiple LoRAs either focuses on specific isolated downstream tasks or fixes the selection of LoRAs during training. However, in real-world scenarios, LLMs receive diverse prompts covering different tasks, and the pool of candidate LoRAs is often dynamically updated. To bridge this gap, we propose LoraRetriever, a retrieve-then-compose framework that adaptively retrieves and composes multiple LoRAs according to the input prompts. LoraRetriever contains three main components: firstly, identifying and retrieving LoRAs relevant to the given input; secondly, formulating strategies for effectively integrating the retrieved LoRAs; and thirdly, developing efficient batch inference to accommodate heterogeneous requests. Experimental results indicate that LoraRetriever consistently outperforms the baselines, highlighting its practical effectiveness and versatility. Our code is available at https://github.com/StyxXuan/LoraRetriever. | [
"Zhao, Ziyu",
"Gan, Leilei",
"Wang, Guoyin",
"Zhou, Wangchunshu",
"Yang, Hongxia",
"Kuang, Kun",
"Wu, Fei"
]
| {L}ora{R}etriever: Input-Aware {L}o{RA} Retrieval and Composition for Mixed Tasks in the Wild | findings-acl.263 | Poster | 2402.09997v1 |
https://aclanthology.org/2024.findings-acl.264.bib | @inproceedings{zhang-etal-2024-elad,
title = "{ELAD}: Explanation-Guided Large Language Models Active Distillation",
author = "Zhang, Yifei and
Pan, Bo and
Ling, Chen and
Hu, Yuntong and
Zhao, Liang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.264",
pages = "4463--4475",
abstract = "The deployment and application of Large Language Models (LLMs) is hindered by their memory inefficiency, computational demands, and the high costs of API inferences. Traditional distillation methods, which transfer the capabilities of LLMs to smaller models, often fail to determine whether the knowledge has been sufficiently transferred, potentially resulting in high costs or incomplete distillation. In this paper, we propose an Explanation-Guided LLMs Active Distillation (ELAD) framework that employs an active learning strategy to optimize the balance between annotation costs and model performance. To improve the efficiency of sample selection, we introduce an explanation-guided sample selection method that identifies samples challenging its reasoning by exploiting uncertainties in reasoning explanation steps. Additionally, we present a customized LLM-annotated explanation revision technique where the teacher model detects and corrects flaws in the student model{'}s reasoning. Our experiments across various reasoning datasets demonstrate that our framework significantly enhances the efficiency of LLMs knowledge distillation.",
}
| The deployment and application of Large Language Models (LLMs) is hindered by their memory inefficiency, computational demands, and the high costs of API inferences. Traditional distillation methods, which transfer the capabilities of LLMs to smaller models, often fail to determine whether the knowledge has been sufficiently transferred, potentially resulting in high costs or incomplete distillation. In this paper, we propose an Explanation-Guided LLMs Active Distillation (ELAD) framework that employs an active learning strategy to optimize the balance between annotation costs and model performance. To improve the efficiency of sample selection, we introduce an explanation-guided sample selection method that identifies samples challenging its reasoning by exploiting uncertainties in reasoning explanation steps. Additionally, we present a customized LLM-annotated explanation revision technique where the teacher model detects and corrects flaws in the student model{'}s reasoning. Our experiments across various reasoning datasets demonstrate that our framework significantly enhances the efficiency of LLMs knowledge distillation. | [
"Zhang, Yifei",
"Pan, Bo",
"Ling, Chen",
"Hu, Yuntong",
"Zhao, Liang"
]
| {ELAD}: Explanation-Guided Large Language Models Active Distillation | findings-acl.264 | Poster | 2401.11365v1 |
https://aclanthology.org/2024.findings-acl.265.bib | @inproceedings{holtermann-etal-2024-evaluating,
title = "Evaluating the Elementary Multilingual Capabilities of Large Language Models with {M}ulti{Q}",
author = {Holtermann, Carolin and
R{\"o}ttger, Paul and
Dill, Timm and
Lauscher, Anne},
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.265",
pages = "4476--4494",
abstract = "Large language models (LLMs) need to serve everyone, including a global majority of non-English speakers. However, most LLMs today, and open LLMs in particular, are often intended for use in just English (e.g. Llama2, Mistral) or a small handful of high-resource languages (e.g. Mixtral, Qwen). Recent research shows that, despite limits in their intended use, people prompt LLMs in many different languages.Therefore, in this paper, we investigate the basic multilingual capabilities of state-of-the-art open LLMs beyond their intended use.For this purpose, we introduce MultiQ, a new silver standard benchmark for basic open-ended question answering with 27.4k test questions across a typologically diverse set of 137 languages. With MultiQ, we evaluate language fidelity, i.e. whether models respond in the prompted language, and question answering accuracy. All LLMs we test respond faithfully and/or accurately for at least some languages beyond their intended use. Most models are more accurate when they respond faithfully. However, differences across models are large, and there is a long tail of languages where models are neither accurate nor faithful. We explore differences in tokenization as a potential explanation for our findings, identifying possible correlations that warrant further investigation.",
}
| Large language models (LLMs) need to serve everyone, including a global majority of non-English speakers. However, most LLMs today, and open LLMs in particular, are often intended for use in just English (e.g. Llama2, Mistral) or a small handful of high-resource languages (e.g. Mixtral, Qwen). Recent research shows that, despite limits in their intended use, people prompt LLMs in many different languages.Therefore, in this paper, we investigate the basic multilingual capabilities of state-of-the-art open LLMs beyond their intended use.For this purpose, we introduce MultiQ, a new silver standard benchmark for basic open-ended question answering with 27.4k test questions across a typologically diverse set of 137 languages. With MultiQ, we evaluate language fidelity, i.e. whether models respond in the prompted language, and question answering accuracy. All LLMs we test respond faithfully and/or accurately for at least some languages beyond their intended use. Most models are more accurate when they respond faithfully. However, differences across models are large, and there is a long tail of languages where models are neither accurate nor faithful. We explore differences in tokenization as a potential explanation for our findings, identifying possible correlations that warrant further investigation. | [
"Holtermann, Carolin",
"R{\\\"o}ttger, Paul",
"Dill, Timm",
"Lauscher, Anne"
]
| Evaluating the Elementary Multilingual Capabilities of Large Language Models with {M}ulti{Q} | findings-acl.265 | Poster | 2403.03814v2 |
https://aclanthology.org/2024.findings-acl.266.bib | @inproceedings{matthews-etal-2024-semantics,
title = "Semantics or spelling? Probing contextual word embeddings with orthographic noise",
author = "Matthews, Jacob and
Starr, John and
Schijndel, Marten",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.266",
pages = "4495--4504",
abstract = "Pretrained language model (PLM) hidden states are frequently employed as contextual word embeddings (CWE): high-dimensional representations that encode semantic information given linguistic context. Across many areas of computational linguistics research, similarity between CWEs is interpreted as semantic similarity. However, it remains unclear exactly what information is encoded in PLM hidden states. We investigate this practice by probing PLM representations using minimal orthographic noise. We expect that if CWEs primarily encode semantic information, a single character swap in the input word will not drastically affect the resulting representation, given sufficient linguistic context. Surprisingly, we find that CWEs generated by popular PLMs are highly sensitive to noise in input data, and that this sensitivity is related to subword tokenization: the fewer tokens used to represent a word at input, the more sensitive its corresponding CWE. This suggests that CWEs capture information unrelated to word-level meaning and can be manipulated through trivial modifications of input data. We conclude that these PLM-derived CWEs may not be reliable semantic proxies, and that caution is warranted when interpreting representational similarity.",
}
| Pretrained language model (PLM) hidden states are frequently employed as contextual word embeddings (CWE): high-dimensional representations that encode semantic information given linguistic context. Across many areas of computational linguistics research, similarity between CWEs is interpreted as semantic similarity. However, it remains unclear exactly what information is encoded in PLM hidden states. We investigate this practice by probing PLM representations using minimal orthographic noise. We expect that if CWEs primarily encode semantic information, a single character swap in the input word will not drastically affect the resulting representation, given sufficient linguistic context. Surprisingly, we find that CWEs generated by popular PLMs are highly sensitive to noise in input data, and that this sensitivity is related to subword tokenization: the fewer tokens used to represent a word at input, the more sensitive its corresponding CWE. This suggests that CWEs capture information unrelated to word-level meaning and can be manipulated through trivial modifications of input data. We conclude that these PLM-derived CWEs may not be reliable semantic proxies, and that caution is warranted when interpreting representational similarity. | [
"Matthews, Jacob",
"Starr, John",
"Schijndel, Marten"
]
| Semantics or spelling? Probing contextual word embeddings with orthographic noise | findings-acl.266 | Poster | 2408.04162v1 |
https://aclanthology.org/2024.findings-acl.267.bib | @inproceedings{zeng-etal-2024-good,
title = "The Good and The Bad: Exploring Privacy Issues in Retrieval-Augmented Generation ({RAG})",
author = "Zeng, Shenglai and
Zhang, Jiankun and
He, Pengfei and
Liu, Yiding and
Xing, Yue and
Xu, Han and
Ren, Jie and
Chang, Yi and
Wang, Shuaiqiang and
Yin, Dawei and
Tang, Jiliang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.267",
pages = "4505--4524",
abstract = "Retrieval-augmented generation (RAG) is a powerful technique to facilitate language model generation with proprietary and private data, where data privacy is a pivotal concern. Whereas extensive research has demonstrated the privacy risks of large language models (LLMs), the RAG technique could potentially reshape the inherent behaviors of LLM generation, posing new privacy issues that are currently under-explored. To this end, we conduct extensive empirical studies with novel attack methods, which demonstrate the vulnerability of RAG systems on leaking the private retrieval database. Despite the new risks brought by RAG on the retrieval data, we further discover that RAG can be used to mitigate the old risks, i.e., the leakage of the LLMs{'} training data. In general, we reveal many new insights in this paper for privacy protection of retrieval-augmented LLMs, which could benefit both LLMs and RAG systems builders.",
}
| Retrieval-augmented generation (RAG) is a powerful technique to facilitate language model generation with proprietary and private data, where data privacy is a pivotal concern. Whereas extensive research has demonstrated the privacy risks of large language models (LLMs), the RAG technique could potentially reshape the inherent behaviors of LLM generation, posing new privacy issues that are currently under-explored. To this end, we conduct extensive empirical studies with novel attack methods, which demonstrate the vulnerability of RAG systems on leaking the private retrieval database. Despite the new risks brought by RAG on the retrieval data, we further discover that RAG can be used to mitigate the old risks, i.e., the leakage of the LLMs{'} training data. In general, we reveal many new insights in this paper for privacy protection of retrieval-augmented LLMs, which could benefit both LLMs and RAG systems builders. | [
"Zeng, Shenglai",
"Zhang, Jiankun",
"He, Pengfei",
"Liu, Yiding",
"Xing, Yue",
"Xu, Han",
"Ren, Jie",
"Chang, Yi",
"Wang, Shuaiqiang",
"Yin, Dawei",
"Tang, Jiliang"
]
| The Good and The Bad: Exploring Privacy Issues in Retrieval-Augmented Generation ({RAG}) | findings-acl.267 | Poster | 2402.16893v1 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.