bibtex_url
stringlengths 41
53
| proceedings
stringlengths 38
50
| bibtext
stringlengths 566
3.75k
| abstract
stringlengths 4
3.1k
| authors
sequencelengths 1
66
| title
stringlengths 12
172
| id
stringlengths 7
19
| type
stringclasses 2
values | arxiv_id
stringlengths 0
10
| GitHub
sequencelengths 1
1
| paper_page
stringlengths 0
40
| n_linked_authors
int64 -1
21
| upvotes
int64 -1
116
| num_comments
int64 -1
11
| n_authors
int64 -1
61
| Models
sequencelengths 0
100
| Datasets
sequencelengths 0
100
| Spaces
sequencelengths 0
100
| old_Models
sequencelengths 0
100
| old_Datasets
sequencelengths 0
100
| old_Spaces
sequencelengths 0
100
| paper_page_exists_pre_conf
int64 0
1
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
https://aclanthology.org/2024.emnlp-main.501.bib | https://aclanthology.org/2024.emnlp-main.501/ | @inproceedings{hu-etal-2024-mosel,
title = "{MOSEL}: Inference Serving Using Dynamic Modality Selection",
author = "Hu, Bodun and
Xu, Le and
Moon, Jeongyoon and
Yadwadkar, Neeraja J and
Akella, Aditya",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.501",
pages = "8872--8886",
abstract = "Rapid advancements over the years have helped machine learning models reach previously hard-to-achieve goals, sometimes even exceeding human capabilities. However, achieving desired accuracy comes at the cost of larger model sizes and increased computational demands. Thus, serving predictions from these models to meet any latency and cost requirements of applications remains a key challenge, despite recent work in building inference serving systems as well as algorithmic approaches that dynamically adapt models based on inputs. Our paper introduces a new form of dynamism, modality selection, where we adaptively choose modalities from inference inputs while maintaining the model quality. We introduce MOSEL, an automated inference serving system for multi-modal ML models that carefully picks input modalities per request based on user-defined performance and accuracy requirements. MOSEL exploits modality configurations extensively, improving system throughput by 3.6 $\times$ with an accuracy guarantee. It also reduces job completion times by 11$\times$ compared to modality-agnostic approaches.",
}
| Rapid advancements over the years have helped machine learning models reach previously hard-to-achieve goals, sometimes even exceeding human capabilities. However, achieving desired accuracy comes at the cost of larger model sizes and increased computational demands. Thus, serving predictions from these models to meet any latency and cost requirements of applications remains a key challenge, despite recent work in building inference serving systems as well as algorithmic approaches that dynamically adapt models based on inputs. Our paper introduces a new form of dynamism, modality selection, where we adaptively choose modalities from inference inputs while maintaining the model quality. We introduce MOSEL, an automated inference serving system for multi-modal ML models that carefully picks input modalities per request based on user-defined performance and accuracy requirements. MOSEL exploits modality configurations extensively, improving system throughput by 3.6 $\times$ with an accuracy guarantee. It also reduces job completion times by 11$\times$ compared to modality-agnostic approaches. | [
"Hu, Bodun",
"Xu, Le",
"Moon, Jeongyoon",
"Yadwadkar, Neeraja J",
"Akella, Aditya"
] | MOSEL: Inference Serving Using Dynamic Modality Selection | emnlp-main.501 | Poster | 2310.18481 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.502.bib | https://aclanthology.org/2024.emnlp-main.502/ | @inproceedings{jain-etal-2024-rag,
title = "From {RAG} to Riches: Retrieval Interlaced with Sequence Generation",
author = "Jain, Palak and
Baldini Soares, Livio and
Kwiatkowski, Tom",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.502",
pages = "8887--8904",
abstract = "We present RICHES, a novel approach that interleaves retrieval with sequence generation tasks. RICHES offers an alternative to conventional RAG systems by eliminating the need for separate retriever and generator. It retrieves documents by directly decoding their contents, constrained on the corpus. Unifying retrieval with generation allows us to adapt to diverse new tasks via prompting alone. RICHES can work with any Instruction-tuned model, without additional training. It provides attributed evidence, supports multi-hop retrievals and interleaves thoughts to plan on what to retrieve next, all within a single decoding pass of the LLM. We demonstrate the strong performance of RICHES across ODQA tasks including attributed and multi-hop QA.",
}
| We present RICHES, a novel approach that interleaves retrieval with sequence generation tasks. RICHES offers an alternative to conventional RAG systems by eliminating the need for separate retriever and generator. It retrieves documents by directly decoding their contents, constrained on the corpus. Unifying retrieval with generation allows us to adapt to diverse new tasks via prompting alone. RICHES can work with any Instruction-tuned model, without additional training. It provides attributed evidence, supports multi-hop retrievals and interleaves thoughts to plan on what to retrieve next, all within a single decoding pass of the LLM. We demonstrate the strong performance of RICHES across ODQA tasks including attributed and multi-hop QA. | [
"Jain, Palak",
"Baldini Soares, Livio",
"Kwiatkowski, Tom"
] | From RAG to Riches: Retrieval Interlaced with Sequence Generation | emnlp-main.502 | Poster | 2407.00361 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.503.bib | https://aclanthology.org/2024.emnlp-main.503/ | @inproceedings{su-etal-2024-task,
title = "Task Arithmetic can Mitigate Synthetic-to-Real Gap in Automatic Speech Recognition",
author = "Su, Hsuan and
Farn, Hua and
Sun, Fan-Yun and
Chen, Shang-Tse and
Lee, Hung-yi",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.503",
pages = "8905--8915",
abstract = "Synthetic data is widely used in speech recognition due to the availability of text-to-speech models, which facilitate adapting models to previously unseen text domains. However, existing methods suffer in performance when they fine-tune an automatic speech recognition (ASR) model on synthetic data as they suffer from the distributional shift commonly referred to as the synthetic-to-real gap. In this paper, we find that task arithmetic is effective at mitigating this gap. Our proposed method, $SYN2REAL$ task vector, shows an average improvement of 10.03{\%} improvement in word error rate over baselines on the SLURP dataset. Additionally, we show that an average of $SYN2REAL$ task vectors, when we have real speeches from multiple different domains, can further adapt the original ASR model to perform better on the target text domain.",
}
| Synthetic data is widely used in speech recognition due to the availability of text-to-speech models, which facilitate adapting models to previously unseen text domains. However, existing methods suffer in performance when they fine-tune an automatic speech recognition (ASR) model on synthetic data as they suffer from the distributional shift commonly referred to as the synthetic-to-real gap. In this paper, we find that task arithmetic is effective at mitigating this gap. Our proposed method, $SYN2REAL$ task vector, shows an average improvement of 10.03{\%} improvement in word error rate over baselines on the SLURP dataset. Additionally, we show that an average of $SYN2REAL$ task vectors, when we have real speeches from multiple different domains, can further adapt the original ASR model to perform better on the target text domain. | [
"Su, Hsuan",
"Farn, Hua",
"Sun, Fan-Yun",
"Chen, Shang-Tse",
"Lee, Hung-yi"
] | Task Arithmetic can Mitigate Synthetic-to-Real Gap in Automatic Speech Recognition | emnlp-main.503 | Poster | 2406.02925 | [
""
] | https://huggingface.co/papers/2406.02925 | 1 | 0 | 0 | 4 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.504.bib | https://aclanthology.org/2024.emnlp-main.504/ | @inproceedings{kim-etal-2024-learning,
title = "Learning to Correct for {QA} Reasoning with Black-box {LLM}s",
author = "Kim, Jaehyung and
Kim, Dongyoung and
Yang, Yiming",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.504",
pages = "8916--8937",
abstract = "An open challenge in recent machine learning is about how to improve the reasoning capability of large language models (LLMs) in a black-box setting, i.e., without access to detailed information such as output token probabilities. Existing approaches either rely on accessibility (which is often unrealistic) or involve significantly increased train- and inference-time costs. This paper addresses those limitations or shortcomings by proposing a novel approach, namely CoBB (Correct for improving QA reasoning of Black-Box LLMs). It uses a trained adaptation model to perform a seq2seq mapping from the often-imperfect reasonings of the original black-box LLM to the correct or improved reasonings. Specifically, the adaptation model is initialized with a relatively small open-source LLM and adapted over a collection of sub-sampled training pairs. To select the representative pairs of correct and incorrect reasonings, we formulated the dataset construction as an optimization problem that minimizes the statistical divergence between the sampled subset and the entire collection, and solved it via a genetic algorithm. We then train the adaptation model over the sampled pairs by contrasting the likelihoods of correct and incorrect reasonings. Our experimental results demonstrate that CoBB significantly improves reasoning accuracy across various QA benchmarks, compared to the best-performing adaptation baselines.",
}
| An open challenge in recent machine learning is about how to improve the reasoning capability of large language models (LLMs) in a black-box setting, i.e., without access to detailed information such as output token probabilities. Existing approaches either rely on accessibility (which is often unrealistic) or involve significantly increased train- and inference-time costs. This paper addresses those limitations or shortcomings by proposing a novel approach, namely CoBB (Correct for improving QA reasoning of Black-Box LLMs). It uses a trained adaptation model to perform a seq2seq mapping from the often-imperfect reasonings of the original black-box LLM to the correct or improved reasonings. Specifically, the adaptation model is initialized with a relatively small open-source LLM and adapted over a collection of sub-sampled training pairs. To select the representative pairs of correct and incorrect reasonings, we formulated the dataset construction as an optimization problem that minimizes the statistical divergence between the sampled subset and the entire collection, and solved it via a genetic algorithm. We then train the adaptation model over the sampled pairs by contrasting the likelihoods of correct and incorrect reasonings. Our experimental results demonstrate that CoBB significantly improves reasoning accuracy across various QA benchmarks, compared to the best-performing adaptation baselines. | [
"Kim, Jaehyung",
"Kim, Dongyoung",
"Yang, Yiming"
] | Learning to Correct for QA Reasoning with Black-box LLMs | emnlp-main.504 | Poster | 2406.18695 | [
"https://github.com/bbuing9/cobb"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.505.bib | https://aclanthology.org/2024.emnlp-main.505/ | @inproceedings{yoran-etal-2024-assistantbench,
title = "{A}ssistant{B}ench: Can Web Agents Solve Realistic and Time-Consuming Tasks?",
author = "Yoran, Ori and
Amouyal, Samuel Joseph and
Malaviya, Chaitanya and
Bogin, Ben and
Press, Ofir and
Berant, Jonathan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.505",
pages = "8938--8968",
abstract = "Language agents, built on top of language models (LMs), are systems that can interact with complex environments, such as the open web. In this work, we examine whether such agents can perform realistic and time-consuming tasks on the web, e.g., monitoring real-estate markets or locating relevant nearby businesses. We introduce AssistantBench, a challenging new benchmark consisting of 214 realistic tasks that can be automatically evaluated, covering different scenarios and domains. We find that AssistantBench exposes the limitations of current systems, including language models and retrieval-augmented language models, as no model reaches an accuracy of more than 25 points. While closed-book LMs perform well in terms of accuracy, they exhibit low precision and tend to hallucinate facts. State-of-the-art web agents reach a score of near zero. Additionally, we introduce SeePlanAct (SPA), a new web agent that significantly outperforms previous agents, and an ensemble of SPA and closed-book models reaches the best overall performance. Moreover, we analyze failures of current systems and highlight that open web navigation remains a major challenge.",
}
| Language agents, built on top of language models (LMs), are systems that can interact with complex environments, such as the open web. In this work, we examine whether such agents can perform realistic and time-consuming tasks on the web, e.g., monitoring real-estate markets or locating relevant nearby businesses. We introduce AssistantBench, a challenging new benchmark consisting of 214 realistic tasks that can be automatically evaluated, covering different scenarios and domains. We find that AssistantBench exposes the limitations of current systems, including language models and retrieval-augmented language models, as no model reaches an accuracy of more than 25 points. While closed-book LMs perform well in terms of accuracy, they exhibit low precision and tend to hallucinate facts. State-of-the-art web agents reach a score of near zero. Additionally, we introduce SeePlanAct (SPA), a new web agent that significantly outperforms previous agents, and an ensemble of SPA and closed-book models reaches the best overall performance. Moreover, we analyze failures of current systems and highlight that open web navigation remains a major challenge. | [
"Yoran, Ori",
"Amouyal, Samuel Joseph",
"Malaviya, Chaitanya",
"Bogin, Ben",
"Press, Ofir",
"Berant, Jonathan"
] | AssistantBench: Can Web Agents Solve Realistic and Time-Consuming Tasks? | emnlp-main.505 | Poster | 2407.15711 | [
""
] | https://huggingface.co/papers/2407.15711 | 2 | 9 | 2 | 6 | [] | [
"AssistantBench/AssistantBench"
] | [
"AssistantBench/leaderboard"
] | [] | [
"AssistantBench/AssistantBench"
] | [
"AssistantBench/leaderboard"
] | 1 |
https://aclanthology.org/2024.emnlp-main.506.bib | https://aclanthology.org/2024.emnlp-main.506/ | @inproceedings{chang-etal-2024-postmark,
title = "{P}ost{M}ark: A Robust Blackbox Watermark for Large Language Models",
author = "Chang, Yapei and
Krishna, Kalpesh and
Houmansadr, Amir and
Wieting, John Frederick and
Iyyer, Mohit",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.506",
pages = "8969--8987",
abstract = "The most effective techniques to detect LLM-generated text rely on inserting a detectable signature{---}or watermark{---}during the model{'}s decoding process. Most existing watermarking methods require access to the underlying LLM{'}s logits, which LLM API providers are loath to share due to fears of model distillation. As such, these watermarks must be implemented independently by each LLM provider. In this paper, we develop PostMark, a modular post-hoc watermarking procedure in which an input-dependent set of words (determined via a semantic embedding) is inserted into the text after the decoding process has completed. Critically, PostMark does not require logit access, which means it can be implemented by a third party. We also show that PostMark is more robust to paraphrasing attacks than existing watermarking methods: our experiments cover eight baseline algorithms, five base LLMs, and three datasets. Finally, we evaluate the impact of PostMark on text quality using both automated and human assessments, highlighting the trade-off between quality and robustness to paraphrasing. We release our code, outputs, and annotations at https://github.com/lilakk/PostMark.",
}
| The most effective techniques to detect LLM-generated text rely on inserting a detectable signature{---}or watermark{---}during the model{'}s decoding process. Most existing watermarking methods require access to the underlying LLM{'}s logits, which LLM API providers are loath to share due to fears of model distillation. As such, these watermarks must be implemented independently by each LLM provider. In this paper, we develop PostMark, a modular post-hoc watermarking procedure in which an input-dependent set of words (determined via a semantic embedding) is inserted into the text after the decoding process has completed. Critically, PostMark does not require logit access, which means it can be implemented by a third party. We also show that PostMark is more robust to paraphrasing attacks than existing watermarking methods: our experiments cover eight baseline algorithms, five base LLMs, and three datasets. Finally, we evaluate the impact of PostMark on text quality using both automated and human assessments, highlighting the trade-off between quality and robustness to paraphrasing. We release our code, outputs, and annotations at https://github.com/lilakk/PostMark. | [
"Chang, Yapei",
"Krishna, Kalpesh",
"Houmansadr, Amir",
"Wieting, John Frederick",
"Iyyer, Mohit"
] | PostMark: A Robust Blackbox Watermark for Large Language Models | emnlp-main.506 | Poster | 2406.14517 | [
"https://github.com/lilakk/postmark"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.507.bib | https://aclanthology.org/2024.emnlp-main.507/ | @inproceedings{shen-etal-2024-assessing,
title = "Assessing {``}Implicit{''} Retrieval Robustness of Large Language Models",
author = "Shen, Xiaoyu and
Blloshmi, Rexhina and
Zhu, Dawei and
Pei, Jiahuan and
Zhang, Wei",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.507",
pages = "8988--9003",
abstract = "Retrieval-augmented generation has gained popularity as a framework to enhance large language models with external knowledge. However, its effectiveness hinges on the retrieval robustness of the model. If the model lacks retrieval robustness, its performance is constrained by the accuracy of the retriever, resulting in significant compromises when the retrieved context is irrelevant. In this paper, we evaluate the {``}implicit{''} retrieval robustness of various large language models, instructing them to directly output the final answer without explicitly judging the relevance of the retrieved context. Our findings reveal that fine-tuning on a mix of gold and distracting context significantly enhances the model{'}s robustness to retrieval inaccuracies, while still maintaining its ability to extract correct answers when retrieval is accurate. This suggests that large language models can implicitly handle relevant or irrelevant retrieved context by learning solely from the supervision of the final answer in an end-to-end manner. Introducing an additional process for explicit relevance judgment can be unnecessary and disrupts the end-to-end approach.",
}
| Retrieval-augmented generation has gained popularity as a framework to enhance large language models with external knowledge. However, its effectiveness hinges on the retrieval robustness of the model. If the model lacks retrieval robustness, its performance is constrained by the accuracy of the retriever, resulting in significant compromises when the retrieved context is irrelevant. In this paper, we evaluate the {``}implicit{''} retrieval robustness of various large language models, instructing them to directly output the final answer without explicitly judging the relevance of the retrieved context. Our findings reveal that fine-tuning on a mix of gold and distracting context significantly enhances the model{'}s robustness to retrieval inaccuracies, while still maintaining its ability to extract correct answers when retrieval is accurate. This suggests that large language models can implicitly handle relevant or irrelevant retrieved context by learning solely from the supervision of the final answer in an end-to-end manner. Introducing an additional process for explicit relevance judgment can be unnecessary and disrupts the end-to-end approach. | [
"Shen, Xiaoyu",
"Blloshmi, Rexhina",
"Zhu, Dawei",
"Pei, Jiahuan",
"Zhang, Wei"
] | Assessing “Implicit” Retrieval Robustness of Large Language Models | emnlp-main.507 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.508.bib | https://aclanthology.org/2024.emnlp-main.508/ | @inproceedings{fulay-etal-2024-relationship,
title = "On the Relationship between Truth and Political Bias in Language Models",
author = "Fulay, Suyash and
Brannon, William and
Mohanty, Shrestha and
Overney, Cassandra and
Poole-Dayan, Elinor and
Roy, Deb and
Kabbara, Jad",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.508",
pages = "9004--9018",
abstract = "Language model alignment research often attempts to ensure that models are not only helpful and harmless, but also truthful and unbiased. However, optimizing these objectives simultaneously can obscure how improving one aspect might impact the others. In this work, we focus on analyzing the relationship between two concepts essential in both language model alignment and political science: truthfulness and political bias. We train reward models on various popular truthfulness datasets and subsequently evaluate their political bias. Our findings reveal that optimizing reward models for truthfulness on these datasets tends to result in a left-leaning political bias. We also find that existing open-source reward models (i.e., those trained on standard human preference datasets) already show a similar bias and that the bias is larger for larger models. These results raise important questions about the datasets used to represent truthfulness, potential limitations of aligning models to be both truthful and politically unbiased, and what language models capture about the relationship between truth and politics.",
}
| Language model alignment research often attempts to ensure that models are not only helpful and harmless, but also truthful and unbiased. However, optimizing these objectives simultaneously can obscure how improving one aspect might impact the others. In this work, we focus on analyzing the relationship between two concepts essential in both language model alignment and political science: truthfulness and political bias. We train reward models on various popular truthfulness datasets and subsequently evaluate their political bias. Our findings reveal that optimizing reward models for truthfulness on these datasets tends to result in a left-leaning political bias. We also find that existing open-source reward models (i.e., those trained on standard human preference datasets) already show a similar bias and that the bias is larger for larger models. These results raise important questions about the datasets used to represent truthfulness, potential limitations of aligning models to be both truthful and politically unbiased, and what language models capture about the relationship between truth and politics. | [
"Fulay, Suyash",
"Brannon, William",
"Mohanty, Shrestha",
"Overney, Cass",
"ra",
"Poole-Dayan, Elinor",
"Roy, Deb",
"Kabbara, Jad"
] | On the Relationship between Truth and Political Bias in Language Models | emnlp-main.508 | Oral | 2409.05283 | [
"https://github.com/sfulay/truth_politics"
] | https://huggingface.co/papers/2409.05283 | 2 | 0 | 0 | 7 | [] | [
"wwbrannon/twinviews-13k",
"wwbrannon/TruthGen"
] | [] | [] | [
"wwbrannon/twinviews-13k",
"wwbrannon/TruthGen"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.509.bib | https://aclanthology.org/2024.emnlp-main.509/ | @inproceedings{taneja-goel-2024-active,
title = "Can Active Label Correction Improve {LLM}-based Modular {AI} Systems?",
author = "Taneja, Karan and
Goel, Ashok",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.509",
pages = "9019--9031",
abstract = "Modular AI systems can be developed using LLM-prompts-based modules to minimize deployment time even for complex tasks. However, these systems do not always perform well and improving them using the data traces collected from a deployment remains an open challenge. The data traces contain LLM inputs and outputs, but the annotations from LLMs are noisy. We hypothesize that Active Label Correction (ALC) can be use on the collected data to train smaller task-specific improved models that can replace LLM-based modules. In this paper, we study the noise in three GPT-3.5-annotated datasets and their denoising with human feedback. We also propose a novel method ALC3 that iteratively applies three updates to the training dataset: auto-correction, correction using human feedback and filtering. Our results show that ALC3 can lead to oracle performance with feedback on 17-24{\%} fewer examples than the number of noisy examples in the dataset across three different NLP tasks.",
}
| Modular AI systems can be developed using LLM-prompts-based modules to minimize deployment time even for complex tasks. However, these systems do not always perform well and improving them using the data traces collected from a deployment remains an open challenge. The data traces contain LLM inputs and outputs, but the annotations from LLMs are noisy. We hypothesize that Active Label Correction (ALC) can be use on the collected data to train smaller task-specific improved models that can replace LLM-based modules. In this paper, we study the noise in three GPT-3.5-annotated datasets and their denoising with human feedback. We also propose a novel method ALC3 that iteratively applies three updates to the training dataset: auto-correction, correction using human feedback and filtering. Our results show that ALC3 can lead to oracle performance with feedback on 17-24{\%} fewer examples than the number of noisy examples in the dataset across three different NLP tasks. | [
"Taneja, Karan",
"Goel, Ashok"
] | Can Active Label Correction Improve LLM-based Modular AI Systems? | emnlp-main.509 | Poster | 2401.05467 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.510.bib | https://aclanthology.org/2024.emnlp-main.510/ | @inproceedings{vallebueno-etal-2024-statistical,
title = "Statistical Uncertainty in Word Embeddings: {G}lo{V}e-{V}",
author = "Vallebueno, Andrea and
Handan-Nader, Cassandra and
Manning, Christopher D and
Ho, Daniel E.",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.510",
pages = "9032--9047",
abstract = "Static word embeddings are ubiquitous in computational social science applications and contribute to practical decision-making in a variety of fields including law and healthcare. However, assessing the statistical uncertainty in downstream conclusions drawn from word embedding statistics has remained challenging. When using only point estimates for embeddings, researchers have no streamlined way of assessing the degree to which their model selection criteria or scientific conclusions are subject to noise due to sparsity in the underlying data used to generate the embeddings. We introduce a method to obtain approximate, easy-to-use, and scalable reconstruction error variance estimates for GloVe, one of the most widely used word embedding models, using an analytical approximation to a multivariate normal model. To demonstrate the value of embeddings with variance (GloVe-V), we illustrate how our approach enables principled hypothesis testing in core word embedding tasks, such as comparing the similarity between different word pairs in vector space, assessing the performance of different models, and analyzing the relative degree of ethnic or gender bias in a corpus using different word lists.",
}
| Static word embeddings are ubiquitous in computational social science applications and contribute to practical decision-making in a variety of fields including law and healthcare. However, assessing the statistical uncertainty in downstream conclusions drawn from word embedding statistics has remained challenging. When using only point estimates for embeddings, researchers have no streamlined way of assessing the degree to which their model selection criteria or scientific conclusions are subject to noise due to sparsity in the underlying data used to generate the embeddings. We introduce a method to obtain approximate, easy-to-use, and scalable reconstruction error variance estimates for GloVe, one of the most widely used word embedding models, using an analytical approximation to a multivariate normal model. To demonstrate the value of embeddings with variance (GloVe-V), we illustrate how our approach enables principled hypothesis testing in core word embedding tasks, such as comparing the similarity between different word pairs in vector space, assessing the performance of different models, and analyzing the relative degree of ethnic or gender bias in a corpus using different word lists. | [
"Vallebueno, Andrea",
"H",
"an-Nader, Cass",
"ra",
"Manning, Christopher D",
"Ho, Daniel E."
] | Statistical Uncertainty in Word Embeddings: GloVe-V | emnlp-main.510 | Poster | 2406.12165 | [
"https://github.com/reglab/glove-v"
] | https://huggingface.co/papers/2406.12165 | 1 | 0 | 0 | 4 | [] | [
"reglab/glove-v"
] | [] | [] | [
"reglab/glove-v"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.511.bib | https://aclanthology.org/2024.emnlp-main.511/ | @inproceedings{movva-etal-2024-annotation,
title = "Annotation alignment: Comparing {LLM} and human annotations of conversational safety",
author = "Movva, Rajiv and
Koh, Pang Wei and
Pierson, Emma",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.511",
pages = "9048--9062",
abstract = "Do LLMs align with human perceptions of safety? We study this question via *annotation alignment*, the extent to which LLMs and humans agree when annotating the safety of user-chatbot conversations. We leverage the recent DICES dataset (Aroyo et al. 2023), in which 350 conversations are each rated for safety by 112 annotators spanning 10 race-gender groups. GPT-4 achieves a Pearson correlation of $r=0.59$ with the average annotator rating, higher than the median annotator{'}s correlation with the average ($r=0.51$). We show that larger datasets are needed to resolve whether GPT-4 exhibits disparities in how well it correlates with different demographic groups. Also, there is substantial idiosyncratic variation in correlation within groups, suggesting that race {\&} gender do not fully capture differences in alignment. Finally, we find that GPT-4 cannot predict when one demographic group finds a conversation more unsafe than another.",
}
| Do LLMs align with human perceptions of safety? We study this question via *annotation alignment*, the extent to which LLMs and humans agree when annotating the safety of user-chatbot conversations. We leverage the recent DICES dataset (Aroyo et al. 2023), in which 350 conversations are each rated for safety by 112 annotators spanning 10 race-gender groups. GPT-4 achieves a Pearson correlation of $r=0.59$ with the average annotator rating, higher than the median annotator{'}s correlation with the average ($r=0.51$). We show that larger datasets are needed to resolve whether GPT-4 exhibits disparities in how well it correlates with different demographic groups. Also, there is substantial idiosyncratic variation in correlation within groups, suggesting that race {\&} gender do not fully capture differences in alignment. Finally, we find that GPT-4 cannot predict when one demographic group finds a conversation more unsafe than another. | [
"Movva, Rajiv",
"Koh, Pang Wei",
"Pierson, Emma"
] | Annotation alignment: Comparing LLM and human annotations of conversational safety | emnlp-main.511 | Poster | 2406.06369 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.512.bib | https://aclanthology.org/2024.emnlp-main.512/ | @inproceedings{fernandez-etal-2024-divert,
title = "{D}i{VERT}: Distractor Generation with Variational Errors Represented as Text for Math Multiple-choice Questions",
author = "Fernandez, Nigel and
Scarlatos, Alexander and
Feng, Wanyong and
Woodhead, Simon and
Lan, Andrew",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.512",
pages = "9063--9081",
abstract = "High-quality distractors are crucial to both the assessment and pedagogical value of multiple-choice questions (MCQs), where manually crafting ones that anticipate knowledge deficiencies or misconceptions among real students is difficult. Meanwhile, automated distractor generation, even with the help of large language models (LLMs), remains challenging for subjects like math. It is crucial to not only identify plausible distractors but also understand the error behind them. In this paper, we introduce DiVERT (Distractor Generation with Variational Errors Represented as Text), a novel variational approach that learns an interpretable representation of errors behind distractors in math MCQs. Through experiments on a real-world math MCQ dataset with 1,434 questions used by hundreds of thousands of students, we show that DiVERT, despite using a base open-source LLM with 7B parameters, outperforms state-of-the-art approaches using GPT-4o on downstream distractor generation. We also conduct a human evaluation with math educators and find that DiVERT leads to error labels that are of comparable quality to human-authored ones.",
}
| High-quality distractors are crucial to both the assessment and pedagogical value of multiple-choice questions (MCQs), where manually crafting ones that anticipate knowledge deficiencies or misconceptions among real students is difficult. Meanwhile, automated distractor generation, even with the help of large language models (LLMs), remains challenging for subjects like math. It is crucial to not only identify plausible distractors but also understand the error behind them. In this paper, we introduce DiVERT (Distractor Generation with Variational Errors Represented as Text), a novel variational approach that learns an interpretable representation of errors behind distractors in math MCQs. Through experiments on a real-world math MCQ dataset with 1,434 questions used by hundreds of thousands of students, we show that DiVERT, despite using a base open-source LLM with 7B parameters, outperforms state-of-the-art approaches using GPT-4o on downstream distractor generation. We also conduct a human evaluation with math educators and find that DiVERT leads to error labels that are of comparable quality to human-authored ones. | [
"Fern",
"ez, Nigel",
"Scarlatos, Alex",
"er",
"Feng, Wanyong",
"Woodhead, Simon",
"Lan, Andrew"
] | DiVERT: Distractor Generation with Variational Errors Represented as Text for Math Multiple-choice Questions | emnlp-main.512 | Poster | 2406.19356 | [
"https://github.com/umass-ml4ed/divert"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.513.bib | https://aclanthology.org/2024.emnlp-main.513/ | @inproceedings{wan-etal-2024-factuality,
title = "The Factuality Tax of Diversity-Intervened Text-to-Image Generation: Benchmark and Fact-Augmented Intervention",
author = "Wan, Yixin and
Wu, Di and
Wang, Haoran and
Chang, Kai-Wei",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.513",
pages = "9082--9100",
abstract = "Prompt-based {``}diversity interventions{''} are commonly adopted to improve the diversity of Text-to-Image (T2I) models depicting individuals with various racial or gender traits. However, will this strategy result in nonfactual demographic distribution, especially when generating real historical figures? In this work, we propose **DemOgraphic FActualIty Representation (DoFaiR)**, a benchmark to systematically quantify the trade-off between using diversity interventions and preserving demographic factuality in T2I models. DoFaiR consists of 756 meticulously fact-checked test instances to reveal the factuality tax of various diversity prompts through an automated evidence-supported evaluation pipeline. Experiments on DoFaiR unveil that diversity-oriented instructions increase the number of different gender and racial groups in DALLE-3{'}s generations at the cost of historically inaccurate demographic distributions. To resolve this issue, we propose **Fact-Augmented Intervention** (FAI), which instructs a Large Language Model (LLM) to reflect on verbalized or retrieved factual information about gender and racial compositions of generation subjects in history, and incorporate it into the generation context of T2I models. By orienting model generations using the reflected historical truths, FAI significantly improves the demographic factuality under diversity interventions while preserving diversity.",
}
| Prompt-based {``}diversity interventions{''} are commonly adopted to improve the diversity of Text-to-Image (T2I) models depicting individuals with various racial or gender traits. However, will this strategy result in nonfactual demographic distribution, especially when generating real historical figures? In this work, we propose **DemOgraphic FActualIty Representation (DoFaiR)**, a benchmark to systematically quantify the trade-off between using diversity interventions and preserving demographic factuality in T2I models. DoFaiR consists of 756 meticulously fact-checked test instances to reveal the factuality tax of various diversity prompts through an automated evidence-supported evaluation pipeline. Experiments on DoFaiR unveil that diversity-oriented instructions increase the number of different gender and racial groups in DALLE-3{'}s generations at the cost of historically inaccurate demographic distributions. To resolve this issue, we propose **Fact-Augmented Intervention** (FAI), which instructs a Large Language Model (LLM) to reflect on verbalized or retrieved factual information about gender and racial compositions of generation subjects in history, and incorporate it into the generation context of T2I models. By orienting model generations using the reflected historical truths, FAI significantly improves the demographic factuality under diversity interventions while preserving diversity. | [
"Wan, Yixin",
"Wu, Di",
"Wang, Haoran",
"Chang, Kai-Wei"
] | The Factuality Tax of Diversity-Intervened Text-to-Image Generation: Benchmark and Fact-Augmented Intervention | emnlp-main.513 | Poster | 2407.00377 | [
""
] | https://huggingface.co/papers/2407.00377 | 1 | 0 | 0 | 4 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.514.bib | https://aclanthology.org/2024.emnlp-main.514/ | @inproceedings{li-etal-2024-cleangen,
title = "{C}lean{G}en: Mitigating Backdoor Attacks for Generation Tasks in Large Language Models",
author = "Li, Yuetai and
Xu, Zhangchen and
Jiang, Fengqing and
Niu, Luyao and
Sahabandu, Dinuka and
Ramasubramanian, Bhaskar and
Poovendran, Radha",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.514",
pages = "9101--9118",
abstract = "The remarkable performance of large language models (LLMs) in generation tasks has enabled practitioners to leverage publicly available models to power custom applications, such as chatbots and virtual assistants. However, the data used to train or fine-tune these LLMs is often undisclosed, allowing an attacker to compromise the data and inject backdoors into the models. In this paper, we develop a novel inference time defense, named CleanGen, to mitigate backdoor attacks for generation tasks in LLMs. CleanGen is a lightweight and effective decoding strategy that is compatible with the state-of-the-art (SOTA) LLMs. Our insight behind CleanGen is that compared to other LLMs, backdoored LLMs assign significantly higher probabilities to tokens representing the attacker-desired contents. These discrepancies in token probabilities enable CleanGen to identify suspicious tokens favored by the attacker and replace them with tokens generated by another LLM that is not compromised by the same attacker, thereby avoiding generation of attacker-desired content. We evaluate CleanGen against five SOTA backdoor attacks. Our results show that CleanGen achieves lower attack success rates (ASR) compared to five SOTA baseline defenses for all five backdoor attacks. Moreover, LLMs deploying CleanGen maintain helpfulness in their responses when serving benign user queries with minimal added computational overhead.",
}
| The remarkable performance of large language models (LLMs) in generation tasks has enabled practitioners to leverage publicly available models to power custom applications, such as chatbots and virtual assistants. However, the data used to train or fine-tune these LLMs is often undisclosed, allowing an attacker to compromise the data and inject backdoors into the models. In this paper, we develop a novel inference time defense, named CleanGen, to mitigate backdoor attacks for generation tasks in LLMs. CleanGen is a lightweight and effective decoding strategy that is compatible with the state-of-the-art (SOTA) LLMs. Our insight behind CleanGen is that compared to other LLMs, backdoored LLMs assign significantly higher probabilities to tokens representing the attacker-desired contents. These discrepancies in token probabilities enable CleanGen to identify suspicious tokens favored by the attacker and replace them with tokens generated by another LLM that is not compromised by the same attacker, thereby avoiding generation of attacker-desired content. We evaluate CleanGen against five SOTA backdoor attacks. Our results show that CleanGen achieves lower attack success rates (ASR) compared to five SOTA baseline defenses for all five backdoor attacks. Moreover, LLMs deploying CleanGen maintain helpfulness in their responses when serving benign user queries with minimal added computational overhead. | [
"Li, Yuetai",
"Xu, Zhangchen",
"Jiang, Fengqing",
"Niu, Luyao",
"Sahab",
"u, Dinuka",
"Ramasubramanian, Bhaskar",
"Poovendran, Radha"
] | CleanGen: Mitigating Backdoor Attacks for Generation Tasks in Large Language Models | emnlp-main.514 | Poster | 2406.12257 | [
"https://github.com/uw-nsl/cleangen"
] | https://huggingface.co/papers/2406.12257 | 1 | 0 | 0 | 7 | [
"TaiGary/CB-ST",
"TaiGary/vpi_code_injection",
"TaiGary/vpi_sentiment_steering",
"TaiGary/AutoPoison"
] | [
"TaiGary/base_model_fine_tune_data_ultrachat_2k"
] | [] | [
"TaiGary/CB-ST",
"TaiGary/vpi_code_injection",
"TaiGary/vpi_sentiment_steering",
"TaiGary/AutoPoison"
] | [
"TaiGary/base_model_fine_tune_data_ultrachat_2k"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.515.bib | https://aclanthology.org/2024.emnlp-main.515/ | @inproceedings{cao-etal-2024-enhancing,
title = "Enhancing Reinforcement Learning with Dense Rewards from Language Model Critic",
author = "Cao, Meng and
Shu, Lei and
Yu, Lei and
Zhu, Yun and
Wichers, Nevan and
Liu, Yinxiao and
Meng, Lei",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.515",
pages = "9119--9138",
abstract = "Reinforcement learning (RL) can align language models with non-differentiable reward signals, such as human preferences. However, a major challenge arises from the sparsity of these reward signals - typically, there is only a single reward for an entire output. This sparsity of rewards can lead to inefficient and unstable learning. To address this challenge, our paper introduces an novel framework that utilizes the critique capability of Large Language Models (LLMs) to produce intermediate-step rewards during RL training. Our method involves coupling a policy model with a critic language model, which is responsible for providing comprehensive feedback of each part of the output. This feedback is then translated into token or span-level rewards that can be used to guide the RL training process. We investigate this approach under two different settings: one where the policy model is smaller and is paired with a more powerful critic model, and another where a single language model fulfills both roles. We assess our approach on three text generation tasks: sentiment control, language model detoxification, and summarization. Experimental results show that incorporating artificial intrinsic rewards significantly improve both sample efficiency and the overall performance of the policy model, supported by both automatic and human evaluation.",
}
| Reinforcement learning (RL) can align language models with non-differentiable reward signals, such as human preferences. However, a major challenge arises from the sparsity of these reward signals - typically, there is only a single reward for an entire output. This sparsity of rewards can lead to inefficient and unstable learning. To address this challenge, our paper introduces an novel framework that utilizes the critique capability of Large Language Models (LLMs) to produce intermediate-step rewards during RL training. Our method involves coupling a policy model with a critic language model, which is responsible for providing comprehensive feedback of each part of the output. This feedback is then translated into token or span-level rewards that can be used to guide the RL training process. We investigate this approach under two different settings: one where the policy model is smaller and is paired with a more powerful critic model, and another where a single language model fulfills both roles. We assess our approach on three text generation tasks: sentiment control, language model detoxification, and summarization. Experimental results show that incorporating artificial intrinsic rewards significantly improve both sample efficiency and the overall performance of the policy model, supported by both automatic and human evaluation. | [
"Cao, Meng",
"Shu, Lei",
"Yu, Lei",
"Zhu, Yun",
"Wichers, Nevan",
"Liu, Yinxiao",
"Meng, Lei"
] | Enhancing Reinforcement Learning with Dense Rewards from Language Model Critic | emnlp-main.515 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.516.bib | https://aclanthology.org/2024.emnlp-main.516/ | @inproceedings{bouzoubaa-etal-2024-words,
title = "Words Matter: Reducing Stigma in Online Conversations about Substance Use with Large Language Models",
author = "Bouzoubaa, Layla and
Aghakhani, Elham and
Rezapour, Rezvaneh",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.516",
pages = "9139--9156",
abstract = "Stigma is a barrier to treatment for individuals struggling with substance use disorders (SUD), which leads to significantly lower treatment engagement rates. With only 7{\%} of those affected receiving any form of help, societal stigma not only discourages individuals with SUD from seeking help but isolates them, hindering their recovery journey and perpetuating a cycle of shame and self-doubt. This study investigates how stigma manifests on social media, particularly Reddit, where anonymity can exacerbate discriminatory behaviors. We analyzed over 1.2 million posts, identifying 3,207 that exhibited stigmatizing language related to people who use substances (PWUS). Of these, 1,649 posts were classified as containing directed stigma towards PWUS, which became the focus of our de-stigmatization efforts. Using Informed and Stylized LLMs, we developed a model to transform these instances into more empathetic language.Our paper contributes to the field by proposing a computational framework for analyzing stigma and de-stigmatizing online content, and delving into the linguistic features that propagate stigma towards PWUS. Our work not only enhances understanding of stigma{'}s manifestations online but also provides practical tools for fostering a more supportive environment for those affected by SUD.",
}
| Stigma is a barrier to treatment for individuals struggling with substance use disorders (SUD), which leads to significantly lower treatment engagement rates. With only 7{\%} of those affected receiving any form of help, societal stigma not only discourages individuals with SUD from seeking help but isolates them, hindering their recovery journey and perpetuating a cycle of shame and self-doubt. This study investigates how stigma manifests on social media, particularly Reddit, where anonymity can exacerbate discriminatory behaviors. We analyzed over 1.2 million posts, identifying 3,207 that exhibited stigmatizing language related to people who use substances (PWUS). Of these, 1,649 posts were classified as containing directed stigma towards PWUS, which became the focus of our de-stigmatization efforts. Using Informed and Stylized LLMs, we developed a model to transform these instances into more empathetic language.Our paper contributes to the field by proposing a computational framework for analyzing stigma and de-stigmatizing online content, and delving into the linguistic features that propagate stigma towards PWUS. Our work not only enhances understanding of stigma{'}s manifestations online but also provides practical tools for fostering a more supportive environment for those affected by SUD. | [
"Bouzoubaa, Layla",
"Aghakhani, Elham",
"Rezapour, Rezvaneh"
] | Words Matter: Reducing Stigma in Online Conversations about Substance Use with Large Language Models | emnlp-main.516 | Poster | 2408.07873 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.517.bib | https://aclanthology.org/2024.emnlp-main.517/ | @inproceedings{chen-etal-2024-efficient,
title = "Efficient Sequential Decision Making with Large Language Models",
author = "Chen, Dingyang and
Zhang, Qi and
Zhu, Yinglun",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.517",
pages = "9157--9170",
abstract = "This paper focuses on extending the success of large language models (LLMs) to sequential decision making. Existing efforts either (i) re-train or finetune LLMs for decision making, or (ii) design prompts for pretrained LLMs. The former approach suffers from the computational burden of gradient updates, and the latter approach does not show promising results. In this paper, we propose a new approach that leverages online model selection algorithms to efficiently incorporate LLMs agents into sequential decision making. Statistically, our approach significantly outperforms both traditional decision making algorithms and vanilla LLM agents. Computationally, our approach avoids the need for expensive gradient updates of LLMs, and throughout the decision making process, it requires only a small number of LLM calls. We conduct extensive experiments to verify the effectiveness of our proposed approach. As an example, on a large-scale Amazon dataset, our approach achieves more than a 6x performance gain over baselines while calling LLMs in only 1.5{\%} of the time steps.",
}
| This paper focuses on extending the success of large language models (LLMs) to sequential decision making. Existing efforts either (i) re-train or finetune LLMs for decision making, or (ii) design prompts for pretrained LLMs. The former approach suffers from the computational burden of gradient updates, and the latter approach does not show promising results. In this paper, we propose a new approach that leverages online model selection algorithms to efficiently incorporate LLMs agents into sequential decision making. Statistically, our approach significantly outperforms both traditional decision making algorithms and vanilla LLM agents. Computationally, our approach avoids the need for expensive gradient updates of LLMs, and throughout the decision making process, it requires only a small number of LLM calls. We conduct extensive experiments to verify the effectiveness of our proposed approach. As an example, on a large-scale Amazon dataset, our approach achieves more than a 6x performance gain over baselines while calling LLMs in only 1.5{\%} of the time steps. | [
"Chen, Dingyang",
"Zhang, Qi",
"Zhu, Yinglun"
] | Efficient Sequential Decision Making with Large Language Models | emnlp-main.517 | Oral | 2406.12125 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.518.bib | https://aclanthology.org/2024.emnlp-main.518/ | @inproceedings{jiang-etal-2024-signclip,
title = "{S}ign{CLIP}: Connecting Text and Sign Language by Contrastive Learning",
author = {Jiang, Zifan and
Sant, Gerard and
Moryossef, Amit and
M{\"u}ller, Mathias and
Sennrich, Rico and
Ebling, Sarah},
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.518",
pages = "9171--9193",
abstract = "We present SignCLIP, which re-purposes CLIP (Contrastive Language-Image Pretraining) to project spoken language text and sign language videos, two classes of natural languages of distinct modalities, into the same space. SignCLIP is an efficient method of learning useful visual representations for sign language processing from large-scale, multilingual video-text pairs, without directly optimizing for a specific task or sign language which is often of limited size.We pretrain SignCLIP on Spreadthesign, a prominent sign language dictionary consisting of {\textasciitilde}500 thousand video clips in up to 44 sign languages, and evaluate it with various downstream datasets. SignCLIP discerns in-domain signing with notable text-to-video/video-to-text retrieval accuracy. It also performs competitively for out-of-domain downstream tasks such as isolated sign language recognition upon essential few-shot prompting or fine-tuning.We analyze the latent space formed by the spoken language text and sign language poses, which provides additional linguistic insights. Our code and models are openly available.",
}
| We present SignCLIP, which re-purposes CLIP (Contrastive Language-Image Pretraining) to project spoken language text and sign language videos, two classes of natural languages of distinct modalities, into the same space. SignCLIP is an efficient method of learning useful visual representations for sign language processing from large-scale, multilingual video-text pairs, without directly optimizing for a specific task or sign language which is often of limited size.We pretrain SignCLIP on Spreadthesign, a prominent sign language dictionary consisting of {\textasciitilde}500 thousand video clips in up to 44 sign languages, and evaluate it with various downstream datasets. SignCLIP discerns in-domain signing with notable text-to-video/video-to-text retrieval accuracy. It also performs competitively for out-of-domain downstream tasks such as isolated sign language recognition upon essential few-shot prompting or fine-tuning.We analyze the latent space formed by the spoken language text and sign language poses, which provides additional linguistic insights. Our code and models are openly available. | [
"Jiang, Zifan",
"Sant, Gerard",
"Moryossef, Amit",
"M{\\\"u}ller, Mathias",
"Sennrich, Rico",
"Ebling, Sarah"
] | SignCLIP: Connecting Text and Sign Language by Contrastive Learning | emnlp-main.518 | Poster | 2407.01264 | [
"https://github.com/J22Melody/fairseq"
] | https://huggingface.co/papers/2407.01264 | 0 | 0 | 0 | 6 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.519.bib | https://aclanthology.org/2024.emnlp-main.519/ | @inproceedings{guo-etal-2024-appls,
title = "{APPLS}: Evaluating Evaluation Metrics for Plain Language Summarization",
author = "Guo, Yue and
August, Tal and
Leroy, Gondy and
Cohen, Trevor and
Wang, Lucy Lu",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.519",
pages = "9194--9211",
abstract = "While there has been significant development of models for Plain Language Summarization (PLS), evaluation remains a challenge. PLS lacks a dedicated assessment metric, and the suitability of text generation evaluation metrics is unclear due to the unique transformations involved (e.g., adding background explanations, removing jargon). To address these questions, our study introduces a granular meta-evaluation testbed, APPLS, designed to evaluate metrics for PLS. We identify four PLS criteria from previous work{---}informativeness, simplification, coherence, and faithfulness{---}and define a set of perturbations corresponding to these criteria that sensitive metrics should be able to detect. We apply these perturbations to extractive hypotheses for two PLS datasets to form our testbed. Using APPLS, we assess performance of 14 metrics, including automated scores, lexical features, and LLM prompt-based evaluations. Our analysis reveals that while some current metrics show sensitivity to specific criteria, no single method captures all four criteria simultaneously. We therefore recommend a suite of automated metrics be used to capture PLS quality along all relevant criteria. This work contributes the first meta-evaluation testbed for PLS and a comprehensive evaluation of existing metrics.",
}
| While there has been significant development of models for Plain Language Summarization (PLS), evaluation remains a challenge. PLS lacks a dedicated assessment metric, and the suitability of text generation evaluation metrics is unclear due to the unique transformations involved (e.g., adding background explanations, removing jargon). To address these questions, our study introduces a granular meta-evaluation testbed, APPLS, designed to evaluate metrics for PLS. We identify four PLS criteria from previous work{---}informativeness, simplification, coherence, and faithfulness{---}and define a set of perturbations corresponding to these criteria that sensitive metrics should be able to detect. We apply these perturbations to extractive hypotheses for two PLS datasets to form our testbed. Using APPLS, we assess performance of 14 metrics, including automated scores, lexical features, and LLM prompt-based evaluations. Our analysis reveals that while some current metrics show sensitivity to specific criteria, no single method captures all four criteria simultaneously. We therefore recommend a suite of automated metrics be used to capture PLS quality along all relevant criteria. This work contributes the first meta-evaluation testbed for PLS and a comprehensive evaluation of existing metrics. | [
"Guo, Yue",
"August, Tal",
"Leroy, Gondy",
"Cohen, Trevor",
"Wang, Lucy Lu"
] | APPLS: Evaluating Evaluation Metrics for Plain Language Summarization | emnlp-main.519 | Poster | 2305.14341 | [
"https://github.com/linguisticanomalies/appls"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.520.bib | https://aclanthology.org/2024.emnlp-main.520/ | @inproceedings{weir-etal-2024-ontologically,
title = "Ontologically Faithful Generation of Non-Player Character Dialogues",
author = "Weir, Nathaniel and
Thomas, Ryan and
d{'}Amore, Randolph and
Hill, Kellie and
Van Durme, Benjamin and
Jhamtani, Harsh",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.520",
pages = "9212--9242",
abstract = "We introduce a language generation dataset grounded in a popular video game. KNUDGE (**KN**owledge Constrained **U**ser-NPC **D**ialogue **GE**neration) requires models to produce trees of dialogue between video game characters that accurately reflect quest and entity specifications stated in natural language. KNUDGE is constructed from side quest dialogues drawn directly from game data of Obsidian Entertainment{'}s {\_}The Outer Worlds{\_}, leading to real-world complexities in generation: (1) utterances must remain faithful to the game lore, including character personas and backstories; (2) a dialogue must accurately reveal new quest details to the human player; and (3) dialogues are large trees as opposed to linear chains of utterances. We report results for a set of neural generation models using supervised and in-context learning techniques; we find competent performance but room for future work addressing the challenges of creating realistic, game-quality dialogues.",
}
| We introduce a language generation dataset grounded in a popular video game. KNUDGE (**KN**owledge Constrained **U**ser-NPC **D**ialogue **GE**neration) requires models to produce trees of dialogue between video game characters that accurately reflect quest and entity specifications stated in natural language. KNUDGE is constructed from side quest dialogues drawn directly from game data of Obsidian Entertainment{'}s {\_}The Outer Worlds{\_}, leading to real-world complexities in generation: (1) utterances must remain faithful to the game lore, including character personas and backstories; (2) a dialogue must accurately reveal new quest details to the human player; and (3) dialogues are large trees as opposed to linear chains of utterances. We report results for a set of neural generation models using supervised and in-context learning techniques; we find competent performance but room for future work addressing the challenges of creating realistic, game-quality dialogues. | [
"Weir, Nathaniel",
"Thomas, Ryan",
"d{'}Amore, R",
"olph",
"Hill, Kellie",
"Van Durme, Benjamin",
"Jhamtani, Harsh"
] | Ontologically Faithful Generation of Non-Player Character Dialogues | emnlp-main.520 | Poster | 2212.10618 | [
""
] | https://huggingface.co/papers/2212.10618 | 2 | 0 | 0 | 6 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.521.bib | https://aclanthology.org/2024.emnlp-main.521/ | @inproceedings{shimabucoro-etal-2024-llm,
title = "{LLM} See, {LLM} Do: Leveraging Active Inheritance to Target Non-Differentiable Objectives",
author = "Shimabucoro, Lu{\'\i}sa and
Ruder, Sebastian and
Kreutzer, Julia and
Fadaee, Marzieh and
Hooker, Sara",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.521",
pages = "9243--9267",
abstract = "The widespread adoption of synthetic data raises new questions about how models generating the data can influence other large language models (LLMs). To start, our work exhaustively characterizes the impact of passive inheritance of model properties by systematically studying how the source of synthetic data shapes models{'} internal biases, calibration and preferences, and their generations{'} textual attributes, providing one of the most comprehensive studies to-date. We find that models are surprisingly sensitive towards certain attributes even when the synthetic data prompts appear {``}neutral{''} which invites the question: can we explicitly steer the distilled data towards desired properties? We demonstrate how such active inheritance can steer the generation profiles of models towards desirable non-differentiable attributes in both directions, e.g. increasing lexical diversity or reducing toxicity. Overall, our study broadens the understanding of the implicit biases inherited by LLMs and explores how we can leverage them to positive effect.",
}
| The widespread adoption of synthetic data raises new questions about how models generating the data can influence other large language models (LLMs). To start, our work exhaustively characterizes the impact of passive inheritance of model properties by systematically studying how the source of synthetic data shapes models{'} internal biases, calibration and preferences, and their generations{'} textual attributes, providing one of the most comprehensive studies to-date. We find that models are surprisingly sensitive towards certain attributes even when the synthetic data prompts appear {``}neutral{''} which invites the question: can we explicitly steer the distilled data towards desired properties? We demonstrate how such active inheritance can steer the generation profiles of models towards desirable non-differentiable attributes in both directions, e.g. increasing lexical diversity or reducing toxicity. Overall, our study broadens the understanding of the implicit biases inherited by LLMs and explores how we can leverage them to positive effect. | [
"Shimabucoro, Lu{\\'\\i}sa",
"Ruder, Sebastian",
"Kreutzer, Julia",
"Fadaee, Marzieh",
"Hooker, Sara"
] | LLM See, LLM Do: Leveraging Active Inheritance to Target Non-Differentiable Objectives | emnlp-main.521 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.522.bib | https://aclanthology.org/2024.emnlp-main.522/ | @inproceedings{taktasheva-etal-2024-rublimp,
title = "{R}u{BL}i{MP}: {R}ussian Benchmark of Linguistic Minimal Pairs",
author = "Taktasheva, Ekaterina and
Bazhukov, Maxim and
Koncha, Kirill and
Fenogenova, Alena and
Artemova, Ekaterina and
Mikhailov, Vladislav",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.522",
pages = "9268--9299",
abstract = "Minimal pairs are a well-established approach to evaluating the grammatical knowledge of language models. However, existing resources for minimal pairs address a limited number of languages and lack diversity of language-specific grammatical phenomena. This paper introduces the Russian Benchmark of Linguistic Minimal Pairs (RuBLiMP), which includes 45k pairs of sentences that differ in grammaticality and isolate a morphological, syntactic, or semantic phenomenon. In contrast to existing benchmarks of linguistic minimal pairs, RuBLiMP is created by applying linguistic perturbations to automatically annotated sentences from open text corpora and decontaminating test data. We describe the data collection protocol and present the results of evaluating 25 language models in various scenarios. We find that the widely used LMs for Russian are sensitive to morphological and agreement-oriented contrasts, but fall behind humans on phenomena requiring the understanding of structural relations, negation, transitivity, and tense. RuBLiMP, the codebase, and other materials are publicly available.",
}
| Minimal pairs are a well-established approach to evaluating the grammatical knowledge of language models. However, existing resources for minimal pairs address a limited number of languages and lack diversity of language-specific grammatical phenomena. This paper introduces the Russian Benchmark of Linguistic Minimal Pairs (RuBLiMP), which includes 45k pairs of sentences that differ in grammaticality and isolate a morphological, syntactic, or semantic phenomenon. In contrast to existing benchmarks of linguistic minimal pairs, RuBLiMP is created by applying linguistic perturbations to automatically annotated sentences from open text corpora and decontaminating test data. We describe the data collection protocol and present the results of evaluating 25 language models in various scenarios. We find that the widely used LMs for Russian are sensitive to morphological and agreement-oriented contrasts, but fall behind humans on phenomena requiring the understanding of structural relations, negation, transitivity, and tense. RuBLiMP, the codebase, and other materials are publicly available. | [
"Taktasheva, Ekaterina",
"Bazhukov, Maxim",
"Koncha, Kirill",
"Fenogenova, Alena",
"Artemova, Ekaterina",
"Mikhailov, Vladislav"
] | RuBLiMP: Russian Benchmark of Linguistic Minimal Pairs | emnlp-main.522 | Poster | 2406.19232 | [
"https://github.com/russiannlp/rublimp"
] | https://huggingface.co/papers/2406.19232 | 3 | 0 | 0 | 6 | [] | [
"RussianNLP/rublimp"
] | [] | [] | [
"RussianNLP/rublimp"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.523.bib | https://aclanthology.org/2024.emnlp-main.523/ | @inproceedings{deng-etal-2024-text,
title = "Text-Tuple-Table: Towards Information Integration in Text-to-Table Generation via Global Tuple Extraction",
author = "Deng, Zheye and
Chan, Chunkit and
Wang, Weiqi and
Sun, Yuxi and
Fan, Wei and
Zheng, Tianshi and
Yim, Yauwai and
Song, Yangqiu",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.523",
pages = "9300--9322",
abstract = "The task of condensing large chunks of textual information into concise and structured tables has gained attention recently due to the emergence of Large Language Models (LLMs) and their potential benefit for downstream tasks, such as text summarization and text mining. Previous approaches often generate tables that directly replicate information from the text, limiting their applicability in broader contexts, as text-to-table generation in real-life scenarios necessitates information extraction, reasoning, and integration. However, there is a lack of both datasets and methodologies towards this task. In this paper, we introduce LiveSum, a new benchmark dataset created for generating summary tables of competitions based on real-time commentary texts. We evaluate the performances of state-of-the-art LLMs on this task in both fine-tuning and zero-shot settings, and additionally propose a novel pipeline called $T^3$(Text-Tuple-Table) to improve their performances. Extensive experimental results demonstrate that LLMs still struggle with this task even after fine-tuning, while our approach can offer substantial performance gains without explicit training. Further analyses demonstrate that our method exhibits strong generalization abilities, surpassing previous approaches on several other text-to-table datasets. Our codeand data can be found at https://github.com/HKUST-KnowComp/LiveSum.",
}
| The task of condensing large chunks of textual information into concise and structured tables has gained attention recently due to the emergence of Large Language Models (LLMs) and their potential benefit for downstream tasks, such as text summarization and text mining. Previous approaches often generate tables that directly replicate information from the text, limiting their applicability in broader contexts, as text-to-table generation in real-life scenarios necessitates information extraction, reasoning, and integration. However, there is a lack of both datasets and methodologies towards this task. In this paper, we introduce LiveSum, a new benchmark dataset created for generating summary tables of competitions based on real-time commentary texts. We evaluate the performances of state-of-the-art LLMs on this task in both fine-tuning and zero-shot settings, and additionally propose a novel pipeline called $T^3$(Text-Tuple-Table) to improve their performances. Extensive experimental results demonstrate that LLMs still struggle with this task even after fine-tuning, while our approach can offer substantial performance gains without explicit training. Further analyses demonstrate that our method exhibits strong generalization abilities, surpassing previous approaches on several other text-to-table datasets. Our codeand data can be found at https://github.com/HKUST-KnowComp/LiveSum. | [
"Deng, Zheye",
"Chan, Chunkit",
"Wang, Weiqi",
"Sun, Yuxi",
"Fan, Wei",
"Zheng, Tianshi",
"Yim, Yauwai",
"Song, Yangqiu"
] | Text-Tuple-Table: Towards Information Integration in Text-to-Table Generation via Global Tuple Extraction | emnlp-main.523 | Oral | 2404.14215 | [
"https://github.com/hkust-knowcomp/livesum-ttt"
] | https://huggingface.co/papers/2404.14215 | 0 | 0 | 0 | 8 | [
"sunatte/txt2sql",
"MachoMaheen/devdock4bit"
] | [] | [
"Justinrune/LLaMA-Factory",
"smarttang/blingsec"
] | [
"sunatte/txt2sql",
"MachoMaheen/devdock4bit"
] | [] | [
"Justinrune/LLaMA-Factory",
"smarttang/blingsec"
] | 1 |
https://aclanthology.org/2024.emnlp-main.524.bib | https://aclanthology.org/2024.emnlp-main.524/ | @inproceedings{mccurdy-etal-2024-toward,
title = "Toward Compositional Behavior in Neural Models: A Survey of Current Views",
author = "McCurdy, Kate and
Soulos, Paul and
Smolensky, Paul and
Fernandez, Roland and
Gao, Jianfeng",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.524",
pages = "9323--9339",
abstract = "Compositionality is a core property of natural language, and compositional behavior (CB) is a crucial goal for modern NLP systems. The research literature, however, includes conflicting perspectives on how CB should be defined, evaluated, and achieved. We propose a conceptual framework to address these questions and survey researchers active in this area.We find consensus on several key points. Researchers broadly accept our proposed definition of CB, agree that it is not solved by current models, and doubt that scale alone will achieve the target behavior. In other areas, we find the field is split on how to move forward, identifying diverse opportunities for future research.",
}
| Compositionality is a core property of natural language, and compositional behavior (CB) is a crucial goal for modern NLP systems. The research literature, however, includes conflicting perspectives on how CB should be defined, evaluated, and achieved. We propose a conceptual framework to address these questions and survey researchers active in this area.We find consensus on several key points. Researchers broadly accept our proposed definition of CB, agree that it is not solved by current models, and doubt that scale alone will achieve the target behavior. In other areas, we find the field is split on how to move forward, identifying diverse opportunities for future research. | [
"McCurdy, Kate",
"Soulos, Paul",
"Smolensky, Paul",
"Fern",
"ez, Rol",
"",
"Gao, Jianfeng"
] | Toward Compositional Behavior in Neural Models: A Survey of Current Views | emnlp-main.524 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.525.bib | https://aclanthology.org/2024.emnlp-main.525/ | @inproceedings{opsahl-ong-etal-2024-optimizing,
title = "Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs",
author = "Opsahl-Ong, Krista and
Ryan, Michael J and
Purtell, Josh and
Broman, David and
Potts, Christopher and
Zaharia, Matei and
Khattab, Omar",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.525",
pages = "9340--9366",
abstract = "Language Model Programs, i.e. sophisticated pipelines of modular language model (LM) calls, are increasingly advancing NLP tasks, but they require crafting prompts that are jointly effective for all modules. We study prompt optimization for LM programs, i.e. how to update these prompts to maximize a downstream metric without access to module-level labels or gradients. To make this tractable, we factorize our problem into optimizing the free-form instructions and few-shot demonstrations of every module and introduce several strategies to craft task-grounded instructions and navigate credit assignment across modules. Our strategies include (i) program- and data-aware techniques for proposing effective instructions, (ii) a stochastic mini-batch evaluation function for learning a surrogate model of our objective, and (iii) a meta-optimization procedure in which we refine how LMs construct proposals over time. Using these insights we develop MIPRO, a novel algorithm for optimizing LM programs. MIPRO outperforms baseline optimizers on five of seven diverse multi-stage LM programs using a best-in-class open-source model (Llama-3-8B), by as high as 13{\%} accuracy. We have released our new optimizers and benchmark in DSPy at [http://dspy.ai](http://dspy.ai).",
}
| Language Model Programs, i.e. sophisticated pipelines of modular language model (LM) calls, are increasingly advancing NLP tasks, but they require crafting prompts that are jointly effective for all modules. We study prompt optimization for LM programs, i.e. how to update these prompts to maximize a downstream metric without access to module-level labels or gradients. To make this tractable, we factorize our problem into optimizing the free-form instructions and few-shot demonstrations of every module and introduce several strategies to craft task-grounded instructions and navigate credit assignment across modules. Our strategies include (i) program- and data-aware techniques for proposing effective instructions, (ii) a stochastic mini-batch evaluation function for learning a surrogate model of our objective, and (iii) a meta-optimization procedure in which we refine how LMs construct proposals over time. Using these insights we develop MIPRO, a novel algorithm for optimizing LM programs. MIPRO outperforms baseline optimizers on five of seven diverse multi-stage LM programs using a best-in-class open-source model (Llama-3-8B), by as high as 13{\%} accuracy. We have released our new optimizers and benchmark in DSPy at [http://dspy.ai](http://dspy.ai). | [
"Opsahl-Ong, Krista",
"Ryan, Michael J",
"Purtell, Josh",
"Broman, David",
"Potts, Christopher",
"Zaharia, Matei",
"Khattab, Omar"
] | Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs | emnlp-main.525 | Oral | 2406.11695 | [
"https://github.com/stanfordnlp/dspy"
] | https://huggingface.co/papers/2406.11695 | 0 | 1 | 0 | 7 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.526.bib | https://aclanthology.org/2024.emnlp-main.526/ | @inproceedings{kiegeland-etal-2024-reverse,
title = "Reverse-Engineering the Reader",
author = "Kiegeland, Samuel and
Wilcox, Ethan and
Amini, Afra and
Reich, David Robert and
Cotterell, Ryan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.526",
pages = "9367--9389",
abstract = "Numerous previous studies have sought to determine to what extent language models, pretrained on natural language text, can serve as useful models of human cognition.In this paper, we are interested in the opposite question: whether we can directly optimize a language model to be a useful cognitive model by aligning it to human psychometric data.To achieve this, we introduce a novel alignment technique in which we fine-tune a language model to implicitly optimize the parameters of a linear regressor that directly predicts humans{'} reading times of in-context linguistic units, e.g., phonemes, morphemes, or words, using surprisal estimates derived from the language model. Using words as a test case, we evaluate our technique across multiple model sizes and datasets and find that it improves language models{'} psychometric predictive power.However, we find an inverse relationship between psychometric power and a model{'}s performance on downstream NLP tasks as well as its perplexity on held-out test data.While this latter trend has been observed before (Oh et al., 2022; Shain et al., 2024), we are the first to induce it by manipulating a model{'}s alignment to psychometric data.",
}
| Numerous previous studies have sought to determine to what extent language models, pretrained on natural language text, can serve as useful models of human cognition.In this paper, we are interested in the opposite question: whether we can directly optimize a language model to be a useful cognitive model by aligning it to human psychometric data.To achieve this, we introduce a novel alignment technique in which we fine-tune a language model to implicitly optimize the parameters of a linear regressor that directly predicts humans{'} reading times of in-context linguistic units, e.g., phonemes, morphemes, or words, using surprisal estimates derived from the language model. Using words as a test case, we evaluate our technique across multiple model sizes and datasets and find that it improves language models{'} psychometric predictive power.However, we find an inverse relationship between psychometric power and a model{'}s performance on downstream NLP tasks as well as its perplexity on held-out test data.While this latter trend has been observed before (Oh et al., 2022; Shain et al., 2024), we are the first to induce it by manipulating a model{'}s alignment to psychometric data. | [
"Kiegel",
", Samuel",
"Wilcox, Ethan",
"Amini, Afra",
"Reich, David Robert",
"Cotterell, Ryan"
] | Reverse-Engineering the Reader | emnlp-main.526 | Poster | 2410.13086 | [
"https://github.com/samuki/reverse-engineering-the-reader"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.527.bib | https://aclanthology.org/2024.emnlp-main.527/ | @inproceedings{wu-etal-2024-synchronous,
title = "Synchronous Faithfulness Monitoring for Trustworthy Retrieval-Augmented Generation",
author = "Wu, Di and
Gu, Jia-Chen and
Yin, Fan and
Peng, Nanyun and
Chang, Kai-Wei",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.527",
pages = "9390--9406",
abstract = "Retrieval-augmented language models (RALMs) have shown strong performance and wide applicability in knowledge-intensive tasks. However, there are significant trustworthiness concerns as RALMs are prone to generating unfaithful outputs, including baseless information or contradictions with the retrieved context. This paper proposes SynCheck, a lightweight monitor that leverages fine-grained decoding dynamics including sequence likelihood, uncertainty quantification, context influence, and semantic alignment to synchronously detect unfaithful sentences. By integrating efficiently measurable and complementary signals, SynCheck enables accurate and immediate feedback and intervention. Experiments show that SynCheck significantly outperforms existing faithfulness detection baselines, achieving over 0.85 AUROC across a suite of six long-form retrieval-augmented generation tasks. Leveraging SynCheck, we further introduce FOD, a faithfulness-oriented decoding algorithm guided by beam search for long-form retrieval-augmented generation. Empirical results demonstrate that FOD outperforms traditional strategies such as abstention, reranking, or contrastive decoding significantly in terms of faithfulness, achieving over 10{\%} improvement across six datasets.",
}
| Retrieval-augmented language models (RALMs) have shown strong performance and wide applicability in knowledge-intensive tasks. However, there are significant trustworthiness concerns as RALMs are prone to generating unfaithful outputs, including baseless information or contradictions with the retrieved context. This paper proposes SynCheck, a lightweight monitor that leverages fine-grained decoding dynamics including sequence likelihood, uncertainty quantification, context influence, and semantic alignment to synchronously detect unfaithful sentences. By integrating efficiently measurable and complementary signals, SynCheck enables accurate and immediate feedback and intervention. Experiments show that SynCheck significantly outperforms existing faithfulness detection baselines, achieving over 0.85 AUROC across a suite of six long-form retrieval-augmented generation tasks. Leveraging SynCheck, we further introduce FOD, a faithfulness-oriented decoding algorithm guided by beam search for long-form retrieval-augmented generation. Empirical results demonstrate that FOD outperforms traditional strategies such as abstention, reranking, or contrastive decoding significantly in terms of faithfulness, achieving over 10{\%} improvement across six datasets. | [
"Wu, Di",
"Gu, Jia-Chen",
"Yin, Fan",
"Peng, Nanyun",
"Chang, Kai-Wei"
] | Synchronous Faithfulness Monitoring for Trustworthy Retrieval-Augmented Generation | emnlp-main.527 | Poster | 2406.13692 | [
"https://github.com/xiaowu0162/sync-ralm-faithfulness"
] | https://huggingface.co/papers/2406.13692 | 2 | 0 | 0 | 5 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.528.bib | https://aclanthology.org/2024.emnlp-main.528/ | @inproceedings{cheng-etal-2024-structure,
title = "Structure Guided Prompt: Instructing Large Language Model in Multi-Step Reasoning by Exploring Graph Structure of the Text",
author = "Cheng, Kewei and
Ahmed, Nesreen K. and
Willke, Theodore L. and
Sun, Yizhou",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.528",
pages = "9407--9430",
abstract = "Although Large Language Models (LLMs) excel at addressing straightforward reasoning tasks, they frequently struggle with difficulties when confronted by more complex multi-step reasoning due to a range of factors. Firstly, natural language often encompasses complex relationships among entities, making it challenging to maintain a clear reasoning chain over longer spans. Secondly, the abundance of linguistic diversity means that the same entities and relationships can be expressed using different terminologies and structures, complicating the task of identifying and establishing connections between multiple pieces of information. Graphs provide an effective solution to represent data rich in relational information and capture long-term dependencies among entities. To harness the potential of graphs, our paper introduces Structure Guided Prompt, an innovative three-stage task-agnostic prompting framework designed to improve the multi-step reasoning capabilities of LLMs in a zero-shot setting. This framework explicitly converts unstructured text into a graph via LLMs and instructs them to navigate this graph using task-specific strategies to formulate responses. By effectively organizing information and guiding navigation, it enables LLMs to provide more accurate and context-aware responses. Our experiments show that this framework significantly enhances the reasoning capabilities of LLMs, enabling them to excel in a broader spectrum of natural language scenarios.",
}
| Although Large Language Models (LLMs) excel at addressing straightforward reasoning tasks, they frequently struggle with difficulties when confronted by more complex multi-step reasoning due to a range of factors. Firstly, natural language often encompasses complex relationships among entities, making it challenging to maintain a clear reasoning chain over longer spans. Secondly, the abundance of linguistic diversity means that the same entities and relationships can be expressed using different terminologies and structures, complicating the task of identifying and establishing connections between multiple pieces of information. Graphs provide an effective solution to represent data rich in relational information and capture long-term dependencies among entities. To harness the potential of graphs, our paper introduces Structure Guided Prompt, an innovative three-stage task-agnostic prompting framework designed to improve the multi-step reasoning capabilities of LLMs in a zero-shot setting. This framework explicitly converts unstructured text into a graph via LLMs and instructs them to navigate this graph using task-specific strategies to formulate responses. By effectively organizing information and guiding navigation, it enables LLMs to provide more accurate and context-aware responses. Our experiments show that this framework significantly enhances the reasoning capabilities of LLMs, enabling them to excel in a broader spectrum of natural language scenarios. | [
"Cheng, Kewei",
"Ahmed, Nesreen K.",
"Willke, Theodore L.",
"Sun, Yizhou"
] | Structure Guided Prompt: Instructing Large Language Model in Multi-Step Reasoning by Exploring Graph Structure of the Text | emnlp-main.528 | Poster | 2402.13415 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.529.bib | https://aclanthology.org/2024.emnlp-main.529/ | @inproceedings{schulte-etal-2024-less,
title = "Less is More: Parameter-Efficient Selection of Intermediate Tasks for Transfer Learning",
author = "Schulte, David and
Hamborg, Felix and
Akbik, Alan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.529",
pages = "9431--9442",
abstract = "Intermediate task transfer learning can greatly improve model performance. If, for example, one has little training data for emotion detection, first fine-tuning a language model on a sentiment classification dataset may improve performance strongly. But which task to choose for transfer learning? Prior methods producing useful task rankings are infeasible for large source pools, as they require forward passes through all source language models. We overcome this by introducing Embedding Space Maps (ESMs), light-weight neural networks that approximate the effect of fine-tuning a language model. We conduct the largest study on NLP task transferability and task selection with 12k source-target pairs. We find that applying ESMs on a prior method reduces execution time and disk space usage by factors of 10 and 278, respectively, while retaining high selection performance (avg. regret@5 score of 2.95).",
}
| Intermediate task transfer learning can greatly improve model performance. If, for example, one has little training data for emotion detection, first fine-tuning a language model on a sentiment classification dataset may improve performance strongly. But which task to choose for transfer learning? Prior methods producing useful task rankings are infeasible for large source pools, as they require forward passes through all source language models. We overcome this by introducing Embedding Space Maps (ESMs), light-weight neural networks that approximate the effect of fine-tuning a language model. We conduct the largest study on NLP task transferability and task selection with 12k source-target pairs. We find that applying ESMs on a prior method reduces execution time and disk space usage by factors of 10 and 278, respectively, while retaining high selection performance (avg. regret@5 score of 2.95). | [
"Schulte, David",
"Hamborg, Felix",
"Akbik, Alan"
] | Less is More: Parameter-Efficient Selection of Intermediate Tasks for Transfer Learning | emnlp-main.529 | Poster | 2410.15148 | [
"https://github.com/davidschulte/hf-dataset-selector"
] | https://huggingface.co/papers/2410.15148 | 1 | 0 | 0 | 3 | [
"davidschulte/ESM_tab_fact_tab_fact",
"davidschulte/ESM_adv_glue_adv_mnli_mismatched",
"davidschulte/ESM_DBQ__Burberry.Product.prices.Singapore_default",
"davidschulte/ESM_zeio__auto-pale_annotated",
"davidschulte/ESM_DBQ__Chloe.Product.prices.France_default",
"davidschulte/ESM_ought__raft_tai_safety_research",
"davidschulte/ESM_ought__raft_terms_of_service",
"davidschulte/ESM_nguha__legalbench_cuad_exclusivity",
"davidschulte/ESM_glue_cola",
"davidschulte/ESM_DBQ__Mr.Porter.Product.prices.Taiwan_default",
"davidschulte/ESM_md_gender_bias_wizard",
"davidschulte/ESM_SkAndMl__CPTDS-3_default",
"davidschulte/ESM_indic_glue_wnli.hi",
"davidschulte/ESM_nguha__legalbench_privacy_policy_qa",
"davidschulte/ESM_jrahn__yolochess_deepblue_default",
"davidschulte/ESM_cardiffnlp__tweet_sentiment_multilingual_arabic",
"davidschulte/ESM_nala-cub__americas_nli_all_languages",
"davidschulte/ESM_kuroneko5943__amz20_Keyboard",
"davidschulte/ESM_Patt__ReCoRD_TH_drop_default",
"davidschulte/ESM_senti_lex_ga",
"davidschulte/ESM_prachathai67k_prachathai67k",
"davidschulte/ESM_webis__Touche23-ValueEval_main",
"davidschulte/ESM_tasksource__mmlu_astronomy",
"davidschulte/ESM_nguha__legalbench_supply_chain_disclosure_disclosed_certification",
"davidschulte/ESM_ejschwartz__oo-method-test-split_bylibrarydedup",
"davidschulte/ESM_kor_nlu_nli",
"davidschulte/ESM_nguha__legalbench_supply_chain_disclosure_disclosed_verification",
"davidschulte/ESM_offcombr_offcombr-3",
"davidschulte/ESM_HausaNLP__NaijaSenti-Twitter_yor",
"davidschulte/ESM_nguha__legalbench_citation_prediction_open",
"davidschulte/ESM_metaeval__ethics_deontology",
"davidschulte/ESM_DDSC__angry-tweets_default",
"davidschulte/ESM_fancyzhx__dbpedia_14_dbpedia_14",
"davidschulte/ESM_tasksource__mmlu_abstract_algebra",
"davidschulte/ESM_BerMaker__test_plain_text",
"davidschulte/ESM_ufukhaman__uspto_balanced_200k_ipc_classification_default",
"davidschulte/ESM_tasksource__mmlu_electrical_engineering",
"davidschulte/ESM_vadis__sv-ident_default",
"davidschulte/ESM_mwong__climatetext-claim-related-evaluation_default",
"davidschulte/ESM_senti_lex_de",
"davidschulte/ESM_guardian_authorship_cross_topic_4",
"davidschulte/ESM_se2p__code-readability-merged_default",
"davidschulte/ESM_AmazonScience__massive_zh-CN",
"davidschulte/ESM_senti_lex_fy",
"davidschulte/ESM_kuroneko5943__amz20_Baby",
"davidschulte/ESM_AhmadMustafa__Urdu-Instruct-News-Category-Classification_default",
"davidschulte/ESM_sbx__superlim-2_swenli",
"davidschulte/ESM_CATIE-AQ__paws-x_fr_prompt_paraphrase_detection_default",
"davidschulte/ESM_ar_res_reviews_default",
"davidschulte/ESM_symanto__autextification2023_detection_en",
"davidschulte/ESM_KBLab__overlim_mnli_da",
"davidschulte/ESM_glue_qqp",
"davidschulte/ESM_nguha__legalbench_cuad_ip_ownership_assignment",
"davidschulte/ESM_tyqiangz__multilingual-sentiments_hindi",
"davidschulte/ESM_RussianNLP__russian_super_glue_parus",
"davidschulte/ESM_covid_tweets_japanese_default",
"davidschulte/ESM_crows_pairs_crows_pairs",
"davidschulte/ESM_super_glue_boolq",
"davidschulte/ESM_hpprc__janli_original",
"davidschulte/ESM_nguha__legalbench_contract_nli_permissible_acquirement_of_similar_information",
"davidschulte/ESM_sem_eval_2014_task_1_default",
"davidschulte/ESM_Genius1237__TyDiP_hu",
"davidschulte/ESM_blog_authorship_corpus_blog_authorship_corpus",
"davidschulte/ESM_ought__raft_overruling",
"davidschulte/ESM_nguha__legalbench_cuad_governing_law",
"davidschulte/ESM_polemo2_out",
"davidschulte/ESM_nguha__legalbench_contract_nli_inclusion_of_verbally_conveyed_information",
"davidschulte/ESM_md_gender_bias_convai2_inferred",
"davidschulte/ESM_webis__Touche23-ValueEval_zhihu-meta",
"davidschulte/ESM_joelniklaus__lextreme_swiss_law_area_prediction_considerations",
"davidschulte/ESM_sbx__superlim-2_swesim_similarity",
"davidschulte/ESM_bgglue__bgglue_cinexio",
"davidschulte/ESM_pragmeval_switchboard",
"davidschulte/ESM_senti_lex_fa",
"davidschulte/ESM_AmazonScience__massive_kn-IN",
"davidschulte/ESM_AmazonScience__massive_ro-RO",
"davidschulte/ESM_maveriq__bigbenchhard_logical_deduction_seven_objects",
"davidschulte/ESM_slava-medvedev__zelensky-speeches_default",
"davidschulte/ESM_joelniklaus__lextreme_turkish_constitutional_court_decisions_judgment",
"davidschulte/ESM_tasksource__crowdflower_corporate-messaging",
"davidschulte/ESM_facebook__asset_ratings",
"davidschulte/ESM_indic_glue_wnli.en",
"davidschulte/ESM_clinc_oos_small",
"davidschulte/ESM_TheBritishLibrary__blbooksgenre_title_genre_classifiction",
"davidschulte/ESM_guardian_authorship_cross_genre_4",
"davidschulte/ESM_ttxy__emotion_default",
"davidschulte/ESM_nguha__legalbench_cuad_post-termination_services",
"davidschulte/ESM_nli_tr_snli_tr",
"davidschulte/ESM_almanach__hc3_french_ood_hc3_fr_sentence",
"davidschulte/ESM_AmazonScience__massive_ca-ES",
"davidschulte/ESM_maveriq__bigbenchhard_salient_translation_error_detection",
"davidschulte/ESM_tweet_eval_offensive",
"davidschulte/ESM_yangwang825__marc-ja_default",
"davidschulte/ESM_tasksource__mmlu_high_school_european_history",
"davidschulte/ESM_matthewfranglen__aste-v2_2015-restaurant-sem-eval",
"davidschulte/ESM_imppres_presupposition_both_presupposition",
"davidschulte/ESM_imppres_presupposition_cleft_existence",
"davidschulte/ESM_ScandEval__norec-mini_default",
"davidschulte/ESM_CATIE-AQ__ling_fr_prompt_textual_entailment_default",
"davidschulte/ESM_silicone_mrda"
] | [] | [] | [
"davidschulte/ESM_tab_fact_tab_fact",
"davidschulte/ESM_adv_glue_adv_mnli_mismatched",
"davidschulte/ESM_DBQ__Burberry.Product.prices.Singapore_default",
"davidschulte/ESM_zeio__auto-pale_annotated",
"davidschulte/ESM_DBQ__Chloe.Product.prices.France_default",
"davidschulte/ESM_ought__raft_tai_safety_research",
"davidschulte/ESM_ought__raft_terms_of_service",
"davidschulte/ESM_nguha__legalbench_cuad_exclusivity",
"davidschulte/ESM_glue_cola",
"davidschulte/ESM_DBQ__Mr.Porter.Product.prices.Taiwan_default",
"davidschulte/ESM_md_gender_bias_wizard",
"davidschulte/ESM_SkAndMl__CPTDS-3_default",
"davidschulte/ESM_indic_glue_wnli.hi",
"davidschulte/ESM_nguha__legalbench_privacy_policy_qa",
"davidschulte/ESM_jrahn__yolochess_deepblue_default",
"davidschulte/ESM_cardiffnlp__tweet_sentiment_multilingual_arabic",
"davidschulte/ESM_nala-cub__americas_nli_all_languages",
"davidschulte/ESM_kuroneko5943__amz20_Keyboard",
"davidschulte/ESM_Patt__ReCoRD_TH_drop_default",
"davidschulte/ESM_senti_lex_ga",
"davidschulte/ESM_prachathai67k_prachathai67k",
"davidschulte/ESM_webis__Touche23-ValueEval_main",
"davidschulte/ESM_tasksource__mmlu_astronomy",
"davidschulte/ESM_nguha__legalbench_supply_chain_disclosure_disclosed_certification",
"davidschulte/ESM_ejschwartz__oo-method-test-split_bylibrarydedup",
"davidschulte/ESM_kor_nlu_nli",
"davidschulte/ESM_nguha__legalbench_supply_chain_disclosure_disclosed_verification",
"davidschulte/ESM_offcombr_offcombr-3",
"davidschulte/ESM_HausaNLP__NaijaSenti-Twitter_yor",
"davidschulte/ESM_nguha__legalbench_citation_prediction_open",
"davidschulte/ESM_metaeval__ethics_deontology",
"davidschulte/ESM_DDSC__angry-tweets_default",
"davidschulte/ESM_fancyzhx__dbpedia_14_dbpedia_14",
"davidschulte/ESM_tasksource__mmlu_abstract_algebra",
"davidschulte/ESM_BerMaker__test_plain_text",
"davidschulte/ESM_ufukhaman__uspto_balanced_200k_ipc_classification_default",
"davidschulte/ESM_tasksource__mmlu_electrical_engineering",
"davidschulte/ESM_vadis__sv-ident_default",
"davidschulte/ESM_mwong__climatetext-claim-related-evaluation_default",
"davidschulte/ESM_senti_lex_de",
"davidschulte/ESM_guardian_authorship_cross_topic_4",
"davidschulte/ESM_se2p__code-readability-merged_default",
"davidschulte/ESM_AmazonScience__massive_zh-CN",
"davidschulte/ESM_senti_lex_fy",
"davidschulte/ESM_kuroneko5943__amz20_Baby",
"davidschulte/ESM_AhmadMustafa__Urdu-Instruct-News-Category-Classification_default",
"davidschulte/ESM_sbx__superlim-2_swenli",
"davidschulte/ESM_CATIE-AQ__paws-x_fr_prompt_paraphrase_detection_default",
"davidschulte/ESM_ar_res_reviews_default",
"davidschulte/ESM_symanto__autextification2023_detection_en",
"davidschulte/ESM_KBLab__overlim_mnli_da",
"davidschulte/ESM_glue_qqp",
"davidschulte/ESM_nguha__legalbench_cuad_ip_ownership_assignment",
"davidschulte/ESM_tyqiangz__multilingual-sentiments_hindi",
"davidschulte/ESM_RussianNLP__russian_super_glue_parus",
"davidschulte/ESM_covid_tweets_japanese_default",
"davidschulte/ESM_crows_pairs_crows_pairs",
"davidschulte/ESM_super_glue_boolq",
"davidschulte/ESM_hpprc__janli_original",
"davidschulte/ESM_nguha__legalbench_contract_nli_permissible_acquirement_of_similar_information",
"davidschulte/ESM_sem_eval_2014_task_1_default",
"davidschulte/ESM_Genius1237__TyDiP_hu",
"davidschulte/ESM_blog_authorship_corpus_blog_authorship_corpus",
"davidschulte/ESM_ought__raft_overruling",
"davidschulte/ESM_nguha__legalbench_cuad_governing_law",
"davidschulte/ESM_polemo2_out",
"davidschulte/ESM_nguha__legalbench_contract_nli_inclusion_of_verbally_conveyed_information",
"davidschulte/ESM_md_gender_bias_convai2_inferred",
"davidschulte/ESM_webis__Touche23-ValueEval_zhihu-meta",
"davidschulte/ESM_joelniklaus__lextreme_swiss_law_area_prediction_considerations",
"davidschulte/ESM_sbx__superlim-2_swesim_similarity",
"davidschulte/ESM_bgglue__bgglue_cinexio",
"davidschulte/ESM_pragmeval_switchboard",
"davidschulte/ESM_senti_lex_fa",
"davidschulte/ESM_AmazonScience__massive_kn-IN",
"davidschulte/ESM_AmazonScience__massive_ro-RO",
"davidschulte/ESM_maveriq__bigbenchhard_logical_deduction_seven_objects",
"davidschulte/ESM_slava-medvedev__zelensky-speeches_default",
"davidschulte/ESM_joelniklaus__lextreme_turkish_constitutional_court_decisions_judgment",
"davidschulte/ESM_tasksource__crowdflower_corporate-messaging",
"davidschulte/ESM_facebook__asset_ratings",
"davidschulte/ESM_indic_glue_wnli.en",
"davidschulte/ESM_clinc_oos_small",
"davidschulte/ESM_TheBritishLibrary__blbooksgenre_title_genre_classifiction",
"davidschulte/ESM_guardian_authorship_cross_genre_4",
"davidschulte/ESM_ttxy__emotion_default",
"davidschulte/ESM_nguha__legalbench_cuad_post-termination_services",
"davidschulte/ESM_nli_tr_snli_tr",
"davidschulte/ESM_almanach__hc3_french_ood_hc3_fr_sentence",
"davidschulte/ESM_AmazonScience__massive_ca-ES",
"davidschulte/ESM_maveriq__bigbenchhard_salient_translation_error_detection",
"davidschulte/ESM_tweet_eval_offensive",
"davidschulte/ESM_yangwang825__marc-ja_default",
"davidschulte/ESM_tasksource__mmlu_high_school_european_history",
"davidschulte/ESM_matthewfranglen__aste-v2_2015-restaurant-sem-eval",
"davidschulte/ESM_imppres_presupposition_both_presupposition",
"davidschulte/ESM_imppres_presupposition_cleft_existence",
"davidschulte/ESM_ScandEval__norec-mini_default",
"davidschulte/ESM_CATIE-AQ__ling_fr_prompt_textual_entailment_default",
"davidschulte/ESM_silicone_mrda"
] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.530.bib | https://aclanthology.org/2024.emnlp-main.530/ | @inproceedings{lee-vu-2024-effects,
title = "The effects of distance on {NPI} illusive effects in {BERT}",
author = "Lee, So Young and
Vu, Mai Ha",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.530",
pages = "9443--9457",
abstract = "Previous studies have examined the syntactic capabilities of large pre-trained language models, such as BERT, by using stimuli from psycholinguistic studies. Studying well-known processing errors, such as NPI illusive effects can reveal whether a model prioritizes linear or hierarchical information when processing language. Recent experiments have found that BERT is mildly susceptible to Negative Polarity Item (NPI) illusion effects (Shin et al., 2023; Vu and Lee, 2022). We expand on these results by examining the effect of distance on the illusive effect, using and modifying stimuli from Parker and Phillips (2016). We also further tease apart whether the model is more affected by hierarchical distance or linear distance. We find that BERT is highly sensitive to syntactic hierarchical information: added hierarchical layers affected its processing capabilities compared to added linear distance.",
}
| Previous studies have examined the syntactic capabilities of large pre-trained language models, such as BERT, by using stimuli from psycholinguistic studies. Studying well-known processing errors, such as NPI illusive effects can reveal whether a model prioritizes linear or hierarchical information when processing language. Recent experiments have found that BERT is mildly susceptible to Negative Polarity Item (NPI) illusion effects (Shin et al., 2023; Vu and Lee, 2022). We expand on these results by examining the effect of distance on the illusive effect, using and modifying stimuli from Parker and Phillips (2016). We also further tease apart whether the model is more affected by hierarchical distance or linear distance. We find that BERT is highly sensitive to syntactic hierarchical information: added hierarchical layers affected its processing capabilities compared to added linear distance. | [
"Lee, So Young",
"Vu, Mai Ha"
] | The effects of distance on NPI illusive effects in BERT | emnlp-main.530 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.531.bib | https://aclanthology.org/2024.emnlp-main.531/ | @inproceedings{weir-etal-2024-enhancing,
title = "Enhancing Systematic Decompositional Natural Language Inference Using Informal Logic",
author = "Weir, Nathaniel and
Sanders, Kate and
Weller, Orion and
Sharma, Shreya and
Jiang, Dongwei and
Jiang, Zhengping and
Dalvi Mishra, Bhavana and
Tafjord, Oyvind and
Jansen, Peter and
Clark, Peter and
Van Durme, Benjamin",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.531",
pages = "9458--9482",
abstract = "Recent language models enable new opportunities for structured reasoning with text, such as the construction of intuitive, proof-like textual entailment trees without relying on brittle formal logic. However, progress in this direction has been hampered by a long-standing lack of a clear protocol for determining what {\_}valid decompositional entailment{\_} is. This absence causes noisy datasets and limited performance gains by modern neuro-symbolic entailment engines. To address these problems, we formulate a consistent and theoretically grounded approach to annotating decompositional entailment and evaluate its impact on LLM-based textual inference. We find that our new dataset, RDTE (Recognizing Decompositional Textual Entailment), has a substantially higher internal consistency than prior decompositional entailment datasets, suggesting that RDTE is a significant step forward in the long-standing problem of forming a clear protocol for discerning entailment. We also find that training an RDTE-oriented entailment classifier via knowledge distillation and employing it in an entailment tree reasoning engine significantly improves both accuracy and proof quality, illustrating the practical benefit of this advance for textual inference.",
}
| Recent language models enable new opportunities for structured reasoning with text, such as the construction of intuitive, proof-like textual entailment trees without relying on brittle formal logic. However, progress in this direction has been hampered by a long-standing lack of a clear protocol for determining what {\_}valid decompositional entailment{\_} is. This absence causes noisy datasets and limited performance gains by modern neuro-symbolic entailment engines. To address these problems, we formulate a consistent and theoretically grounded approach to annotating decompositional entailment and evaluate its impact on LLM-based textual inference. We find that our new dataset, RDTE (Recognizing Decompositional Textual Entailment), has a substantially higher internal consistency than prior decompositional entailment datasets, suggesting that RDTE is a significant step forward in the long-standing problem of forming a clear protocol for discerning entailment. We also find that training an RDTE-oriented entailment classifier via knowledge distillation and employing it in an entailment tree reasoning engine significantly improves both accuracy and proof quality, illustrating the practical benefit of this advance for textual inference. | [
"Weir, Nathaniel",
"S",
"ers, Kate",
"Weller, Orion",
"Sharma, Shreya",
"Jiang, Dongwei",
"Jiang, Zhengping",
"Dalvi Mishra, Bhavana",
"Tafjord, Oyvind",
"Jansen, Peter",
"Clark, Peter",
"Van Durme, Benjamin"
] | Enhancing Systematic Decompositional Natural Language Inference Using Informal Logic | emnlp-main.531 | Poster | 2402.14798 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.532.bib | https://aclanthology.org/2024.emnlp-main.532/ | @inproceedings{acquaye-etal-2024-susu,
title = "{S}usu Box or Piggy Bank: Assessing Cultural Commonsense Knowledge between {G}hana and the {US}",
author = "Acquaye, Christabel and
An, Haozhe and
Rudinger, Rachel",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.532",
pages = "9483--9502",
abstract = "Recent work has highlighted the culturally-contingent nature of commonsense knowledge. We introduce AMAMMERε, a test set of 525 multiple-choice questions designed to evaluate the commonsense knowledge of English LLMs, relative to the cultural contexts of Ghana and the United States. To create AMAMMERε, we select a set of multiple-choice questions (MCQs) from existing commonsense datasets and rewrite them in a multi-stage process involving surveys of Ghanaian and U.S. participants. In three rounds of surveys, participants from both pools are solicited to (1) write correct and incorrect answer choices, (2) rate individual answer choices on a 5-point Likert scale, and (3) select the best answer choice from the newly-constructed MCQ items, in a final validation step. By engaging participants at multiple stages, our procedure ensures that participant perspectives are incorporated both in the creation and validation of test items, resulting in high levels of agreement within each pool. We evaluate several off-the-shelf English LLMs on AMAMMERε. Uniformly, models prefer answers choices that align with the preferences of U.S. annotators over Ghanaian annotators. Additionally, when test items specify a cultural context (Ghana or the U.S.), models exhibit some ability to adapt, but performance is consistently better in U.S. contexts than Ghanaian. As large resources are devoted to the advancement of English LLMs, our findings underscore the need for culturally adaptable models and evaluations to meet the needs of diverse English-speaking populations around the world.",
}
| Recent work has highlighted the culturally-contingent nature of commonsense knowledge. We introduce AMAMMERε, a test set of 525 multiple-choice questions designed to evaluate the commonsense knowledge of English LLMs, relative to the cultural contexts of Ghana and the United States. To create AMAMMERε, we select a set of multiple-choice questions (MCQs) from existing commonsense datasets and rewrite them in a multi-stage process involving surveys of Ghanaian and U.S. participants. In three rounds of surveys, participants from both pools are solicited to (1) write correct and incorrect answer choices, (2) rate individual answer choices on a 5-point Likert scale, and (3) select the best answer choice from the newly-constructed MCQ items, in a final validation step. By engaging participants at multiple stages, our procedure ensures that participant perspectives are incorporated both in the creation and validation of test items, resulting in high levels of agreement within each pool. We evaluate several off-the-shelf English LLMs on AMAMMERε. Uniformly, models prefer answers choices that align with the preferences of U.S. annotators over Ghanaian annotators. Additionally, when test items specify a cultural context (Ghana or the U.S.), models exhibit some ability to adapt, but performance is consistently better in U.S. contexts than Ghanaian. As large resources are devoted to the advancement of English LLMs, our findings underscore the need for culturally adaptable models and evaluations to meet the needs of diverse English-speaking populations around the world. | [
"Acquaye, Christabel",
"An, Haozhe",
"Rudinger, Rachel"
] | Susu Box or Piggy Bank: Assessing Cultural Commonsense Knowledge between Ghana and the US | emnlp-main.532 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.533.bib | https://aclanthology.org/2024.emnlp-main.533/ | @inproceedings{fan-etal-2024-read,
title = "Read Anywhere Pointed: Layout-aware {GUI} Screen Reading with Tree-of-Lens Grounding",
author = "Fan, Yue and
Ding, Lei and
Kuo, Ching-Chen and
Jiang, Shan and
Zhao, Yang and
Guan, Xinze and
Yang, Jie and
Zhang, Yi and
Wang, Xin Eric",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.533",
pages = "9503--9522",
abstract = "Graphical User Interfaces (GUIs) are central to our interaction with digital devices and growing efforts have been made to build models for various GUI understanding tasks. However, these efforts largely overlook an important GUI-referring task: screen reading based on user-indicated points, which we name the Screen Point-and-Read (ScreenPR) task. Currently, this task is predominantly handled by rigid accessible screen reading tools, in great need of new models driven by advancements in Multimodal Large Language Models (MLLMs). In this paper, we propose a Tree-of-Lens (ToL) agent, utilizing a novel ToL grounding mechanism, to address the ScreenPR task. Based on the input point coordinate and the corresponding GUI screenshot, our ToL agent constructs a Hierarchical Layout Tree. Based on the tree, our ToL agent not only comprehends the content of the indicated area but also articulates the layout and spatial relationships between elements. Such layout information is crucial for accurately interpreting information on the screen, distinguishing our ToL agent from other screen reading tools. We also thoroughly evaluate the ToL agent against other baselines on a newly proposed ScreenPR benchmark, which includes GUIs from mobile, web, and operating systems. Last but not least, we test the ToL agent on mobile GUI navigation tasks, demonstrating its utility in identifying incorrect actions along the path of agent execution trajectories. Code and data: https://screen-point-and-read.github.io.",
}
| Graphical User Interfaces (GUIs) are central to our interaction with digital devices and growing efforts have been made to build models for various GUI understanding tasks. However, these efforts largely overlook an important GUI-referring task: screen reading based on user-indicated points, which we name the Screen Point-and-Read (ScreenPR) task. Currently, this task is predominantly handled by rigid accessible screen reading tools, in great need of new models driven by advancements in Multimodal Large Language Models (MLLMs). In this paper, we propose a Tree-of-Lens (ToL) agent, utilizing a novel ToL grounding mechanism, to address the ScreenPR task. Based on the input point coordinate and the corresponding GUI screenshot, our ToL agent constructs a Hierarchical Layout Tree. Based on the tree, our ToL agent not only comprehends the content of the indicated area but also articulates the layout and spatial relationships between elements. Such layout information is crucial for accurately interpreting information on the screen, distinguishing our ToL agent from other screen reading tools. We also thoroughly evaluate the ToL agent against other baselines on a newly proposed ScreenPR benchmark, which includes GUIs from mobile, web, and operating systems. Last but not least, we test the ToL agent on mobile GUI navigation tasks, demonstrating its utility in identifying incorrect actions along the path of agent execution trajectories. Code and data: https://screen-point-and-read.github.io. | [
"Fan, Yue",
"Ding, Lei",
"Kuo, Ching-Chen",
"Jiang, Shan",
"Zhao, Yang",
"Guan, Xinze",
"Yang, Jie",
"Zhang, Yi",
"Wang, Xin Eric"
] | Read Anywhere Pointed: Layout-aware GUI Screen Reading with Tree-of-Lens Grounding | emnlp-main.533 | Poster | 2406.19263 | [
"https://github.com/eric-ai-lab/Screen-Point-and-Read"
] | https://huggingface.co/papers/2406.19263 | 2 | 9 | 2 | 9 | [] | [
"yfan1997/ScreenPR",
"orlando23/failed_agent_trajectory",
"orlando23/screendata",
"orlando23/mobile_pc_web_osworld"
] | [] | [] | [
"yfan1997/ScreenPR",
"orlando23/failed_agent_trajectory",
"orlando23/screendata",
"orlando23/mobile_pc_web_osworld"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.534.bib | https://aclanthology.org/2024.emnlp-main.534/ | @inproceedings{pfrommer-etal-2024-ranking,
title = "Ranking Manipulation for Conversational Search Engines",
author = "Pfrommer, Samuel and
Bai, Yatong and
Gautam, Tanmay and
Sojoudi, Somayeh",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.534",
pages = "9523--9552",
abstract = "Major search engine providers are rapidly incorporating Large Language Model (LLM)-generated content in response to user queries. These *conversational search engines* operate by loading retrieved website text into the LLM context for summarization and interpretation. Recent research demonstrates that LLMs are highly vulnerable to jailbreaking and prompt injection attacks, which disrupt the safety and quality goals of LLMs using adversarial strings. This work investigates the impact of prompt injections on the ranking order of sources referenced by conversational search engines. To this end, we introduce a focused dataset of real-world consumer product websites and formalize conversational search ranking as an adversarial problem. Experimentally, we analyze conversational search rankings in the absence of adversarial injections and show that different LLMs vary significantly in prioritizing product name, document content, and context position. We then present a tree-of-attacks-based jailbreaking technique which reliably promotes low-ranked products. Importantly, these attacks transfer effectively to state-of-the-art conversational search engines such as *perplexity.ai*. Given the strong financial incentive for website owners to boost their search ranking, we argue that our problem formulation is of critical importance for future robustness work.",
}
| Major search engine providers are rapidly incorporating Large Language Model (LLM)-generated content in response to user queries. These *conversational search engines* operate by loading retrieved website text into the LLM context for summarization and interpretation. Recent research demonstrates that LLMs are highly vulnerable to jailbreaking and prompt injection attacks, which disrupt the safety and quality goals of LLMs using adversarial strings. This work investigates the impact of prompt injections on the ranking order of sources referenced by conversational search engines. To this end, we introduce a focused dataset of real-world consumer product websites and formalize conversational search ranking as an adversarial problem. Experimentally, we analyze conversational search rankings in the absence of adversarial injections and show that different LLMs vary significantly in prioritizing product name, document content, and context position. We then present a tree-of-attacks-based jailbreaking technique which reliably promotes low-ranked products. Importantly, these attacks transfer effectively to state-of-the-art conversational search engines such as *perplexity.ai*. Given the strong financial incentive for website owners to boost their search ranking, we argue that our problem formulation is of critical importance for future robustness work. | [
"Pfrommer, Samuel",
"Bai, Yatong",
"Gautam, Tanmay",
"Sojoudi, Somayeh"
] | Ranking Manipulation for Conversational Search Engines | emnlp-main.534 | Oral | 2406.03589 | [
"https://github.com/spfrommer/ranking_manipulation"
] | https://huggingface.co/papers/2406.03589 | 1 | 0 | 0 | 4 | [] | [
"Bai-YT/RAGDOLL"
] | [] | [] | [
"Bai-YT/RAGDOLL"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.535.bib | https://aclanthology.org/2024.emnlp-main.535/ | @inproceedings{rahamim-etal-2024-fast,
title = "Fast Forwarding Low-Rank Training",
author = "Rahamim, Adir and
Saphra, Naomi and
Kangaslahti, Sara and
Belinkov, Yonatan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.535",
pages = "9553--9562",
abstract = "Parameter efficient finetuning methods like low-rank adaptation (LoRA) aim to reduce the computational costs of finetuning pretrained Language Models (LMs). Enabled by these low-rank settings, we propose an even more efficient optimization strategy: Fast Forward, a simple and effective approach to accelerate large segments of SGD training. In a Fast Forward stage, we repeat the most recent optimizer step until the loss stops improving on a tiny validation set. By alternating between regular optimization steps and Fast Forward stages, Fast Forward provides up to an 87{\%} reduction in FLOPs over standard SGD with Adam. We validate Fast Forward by finetuning various models on different tasks and demonstrate that it speeds up training without compromising model performance. Additionally, we analyze when and how to apply Fast Forward.",
}
| Parameter efficient finetuning methods like low-rank adaptation (LoRA) aim to reduce the computational costs of finetuning pretrained Language Models (LMs). Enabled by these low-rank settings, we propose an even more efficient optimization strategy: Fast Forward, a simple and effective approach to accelerate large segments of SGD training. In a Fast Forward stage, we repeat the most recent optimizer step until the loss stops improving on a tiny validation set. By alternating between regular optimization steps and Fast Forward stages, Fast Forward provides up to an 87{\%} reduction in FLOPs over standard SGD with Adam. We validate Fast Forward by finetuning various models on different tasks and demonstrate that it speeds up training without compromising model performance. Additionally, we analyze when and how to apply Fast Forward. | [
"Rahamim, Adir",
"Saphra, Naomi",
"Kangaslahti, Sara",
"Belinkov, Yonatan"
] | Fast Forwarding Low-Rank Training | emnlp-main.535 | Poster | 2409.04206 | [
""
] | https://huggingface.co/papers/2409.04206 | 0 | 0 | 0 | 4 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.536.bib | https://aclanthology.org/2024.emnlp-main.536/ | @inproceedings{fogliato-etal-2024-precise,
title = "Precise Model Benchmarking with Only a Few Observations",
author = "Fogliato, Riccardo and
Patil, Pratik and
Akpinar, Nil-Jana and
Monfort, Mathew",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.536",
pages = "9563--9575",
abstract = "How can we precisely estimate a large language model{'}s (LLM) accuracy on questions belonging to a specific topic within a larger question-answering dataset? The standard direct estimator, which averages the model{'}s accuracy on the questions in each subgroup, may exhibit high variance for subgroups (topics) with small sample sizes. Synthetic regression modeling, which leverages the model{'}s accuracy on questions about other topics, may yield biased estimates that are too unreliable for large subgroups. We prescribe a simple yet effective solution: an empirical Bayes (EB) estimator that balances direct and regression estimates for each subgroup separately, improving the precision of subgroup-level estimates of model performance. Our experiments on multiple datasets show that this approach consistently provides more precise estimates of the LLM performance compared to the direct and regression approaches, achieving substantial reductions in the mean squared error. Confidence intervals for EB estimates also have near-nominal coverage and are narrower compared to those for the direct estimator. Additional experiments on tabular and vision data validate the benefits of this EB approach.",
}
| How can we precisely estimate a large language model{'}s (LLM) accuracy on questions belonging to a specific topic within a larger question-answering dataset? The standard direct estimator, which averages the model{'}s accuracy on the questions in each subgroup, may exhibit high variance for subgroups (topics) with small sample sizes. Synthetic regression modeling, which leverages the model{'}s accuracy on questions about other topics, may yield biased estimates that are too unreliable for large subgroups. We prescribe a simple yet effective solution: an empirical Bayes (EB) estimator that balances direct and regression estimates for each subgroup separately, improving the precision of subgroup-level estimates of model performance. Our experiments on multiple datasets show that this approach consistently provides more precise estimates of the LLM performance compared to the direct and regression approaches, achieving substantial reductions in the mean squared error. Confidence intervals for EB estimates also have near-nominal coverage and are narrower compared to those for the direct estimator. Additional experiments on tabular and vision data validate the benefits of this EB approach. | [
"Fogliato, Riccardo",
"Patil, Pratik",
"Akpinar, Nil-Jana",
"Monfort, Mathew"
] | Precise Model Benchmarking with Only a Few Observations | emnlp-main.536 | Poster | 2410.05222 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.537.bib | https://aclanthology.org/2024.emnlp-main.537/ | @inproceedings{berlot-attwell-etal-2024-attribute,
title = "Attribute Diversity Determines the Systematicity Gap in {VQA}",
author = "Berlot-Attwell, Ian and
Agrawal, Kumar Krishna and
Carrell, Annabelle Michael and
Sharma, Yash and
Saphra, Naomi",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.537",
pages = "9576--9611",
abstract = "Although modern neural networks often generalize to new combinations of familiar concepts, the conditions that enable such compositionality have long been an open question. In this work, we study the systematicity gap in visual question answering: the performance difference between reasoning on previously seen and unseen combinations of object attributes. To test, we introduce a novel diagnostic dataset, CLEVR-HOPE. We find that the systematicity gap is not reduced by increasing the quantity of training data, but is reduced by increasing the diversity of training data. In particular, our experiments suggest that the more distinct attribute type combinations are seen during training, the more systematic we can expect the resulting model to be.",
}
| Although modern neural networks often generalize to new combinations of familiar concepts, the conditions that enable such compositionality have long been an open question. In this work, we study the systematicity gap in visual question answering: the performance difference between reasoning on previously seen and unseen combinations of object attributes. To test, we introduce a novel diagnostic dataset, CLEVR-HOPE. We find that the systematicity gap is not reduced by increasing the quantity of training data, but is reduced by increasing the diversity of training data. In particular, our experiments suggest that the more distinct attribute type combinations are seen during training, the more systematic we can expect the resulting model to be. | [
"Berlot-Attwell, Ian",
"Agrawal, Kumar Krishna",
"Carrell, Annabelle Michael",
"Sharma, Yash",
"Saphra, Naomi"
] | Attribute Diversity Determines the Systematicity Gap in VQA | emnlp-main.537 | Poster | 2311.08695 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.538.bib | https://aclanthology.org/2024.emnlp-main.538/ | @inproceedings{newman-etal-2024-arxivdigestables,
title = "{A}rxiv{DIGEST}ables: Synthesizing Scientific Literature into Tables using Language Models",
author = "Newman, Benjamin and
Lee, Yoonjoo and
Naik, Aakanksha and
Siangliulue, Pao and
Fok, Raymond and
Kim, Juho and
Weld, Daniel S and
Chang, Joseph Chee and
Lo, Kyle",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.538",
pages = "9612--9631",
abstract = "When conducting literature reviews, scientists often create literature review tables{---}tables whose rows are publications and whose columns constitute a schema, a set of aspects used to compare and contrast the papers. Can we automatically generate these tables using language models (LMs)? In this work, we introduce a framework that leverages LMs to perform this task by decomposing it into separate schema and value generation steps. To enable experimentation, we address two main challenges: First, we overcome a lack of high-quality datasets to benchmark table generation by curating and releasing arxivDIGESTables, a new dataset of 2,228 literature review tables extracted from ArXiv papers that synthesize a total of 7,542 research papers. Second, to support scalable evaluation of model generations against human-authored reference tables, we develop DecontextEval, an automatic evaluation method that aligns elements of tables with the same underlying aspects despite differing surface forms. Given these tools, we evaluate LMs{'} abilities to reconstruct reference tables, finding this task benefits from additional context to ground the generation (e.g. table captions, in-text references). Finally, through a human evaluation study we find that even when LMs fail to fully reconstruct a reference table, their generated novel aspects can still be useful.",
}
| When conducting literature reviews, scientists often create literature review tables{---}tables whose rows are publications and whose columns constitute a schema, a set of aspects used to compare and contrast the papers. Can we automatically generate these tables using language models (LMs)? In this work, we introduce a framework that leverages LMs to perform this task by decomposing it into separate schema and value generation steps. To enable experimentation, we address two main challenges: First, we overcome a lack of high-quality datasets to benchmark table generation by curating and releasing arxivDIGESTables, a new dataset of 2,228 literature review tables extracted from ArXiv papers that synthesize a total of 7,542 research papers. Second, to support scalable evaluation of model generations against human-authored reference tables, we develop DecontextEval, an automatic evaluation method that aligns elements of tables with the same underlying aspects despite differing surface forms. Given these tools, we evaluate LMs{'} abilities to reconstruct reference tables, finding this task benefits from additional context to ground the generation (e.g. table captions, in-text references). Finally, through a human evaluation study we find that even when LMs fail to fully reconstruct a reference table, their generated novel aspects can still be useful. | [
"Newman, Benjamin",
"Lee, Yoonjoo",
"Naik, Aakanksha",
"Siangliulue, Pao",
"Fok, Raymond",
"Kim, Juho",
"Weld, Daniel S",
"Chang, Joseph Chee",
"Lo, Kyle"
] | ArxivDIGESTables: Synthesizing Scientific Literature into Tables using Language Models | emnlp-main.538 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.539.bib | https://aclanthology.org/2024.emnlp-main.539/ | @inproceedings{shah-etal-2024-development,
title = "Development of Cognitive Intelligence in Pre-trained Language Models",
author = "Shah, Raj Sanjay and
Bhardwaj, Khushi and
Varma, Sashank",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.539",
pages = "9632--9657",
abstract = "Recent studies show evidence for emergent cognitive abilities in Large Pre-trained Language Models (PLMs). The increasing cognitive alignment of these models has made them candidates for cognitive science theories. Prior research into the emergent cognitive abilities of PLMs has been path-independent to model training, i.e. has only looked at the final model weights and not the intermediate steps. However, building plausible models of human cognition using PLMs also requires aligning their performance during training to the developmental trajectories of children{'}s thinking. Guided by psychometric tests of human intelligence, we choose four task categories to investigate the alignment of ten popular families of PLMs and evaluate each of their available intermediate and final training steps: Numerical ability, Linguistic abilities, Conceptual understanding, and Fluid reasoning. We find a striking regularity: regardless of model size, the developmental trajectories of PLMs consistently exhibit a window of maximal alignment to human cognitive development. Before that window, training appears to endow models with the requisite structure to be poised to rapidly learn from experience. After that window, training appears to serve the engineering goal of reducing loss but not the scientific goal of increasing alignment with human cognition.",
}
| Recent studies show evidence for emergent cognitive abilities in Large Pre-trained Language Models (PLMs). The increasing cognitive alignment of these models has made them candidates for cognitive science theories. Prior research into the emergent cognitive abilities of PLMs has been path-independent to model training, i.e. has only looked at the final model weights and not the intermediate steps. However, building plausible models of human cognition using PLMs also requires aligning their performance during training to the developmental trajectories of children{'}s thinking. Guided by psychometric tests of human intelligence, we choose four task categories to investigate the alignment of ten popular families of PLMs and evaluate each of their available intermediate and final training steps: Numerical ability, Linguistic abilities, Conceptual understanding, and Fluid reasoning. We find a striking regularity: regardless of model size, the developmental trajectories of PLMs consistently exhibit a window of maximal alignment to human cognitive development. Before that window, training appears to endow models with the requisite structure to be poised to rapidly learn from experience. After that window, training appears to serve the engineering goal of reducing loss but not the scientific goal of increasing alignment with human cognition. | [
"Shah, Raj Sanjay",
"Bhardwaj, Khushi",
"Varma, Sashank"
] | Development of Cognitive Intelligence in Pre-trained Language Models | emnlp-main.539 | Poster | 2407.01047 | [
""
] | https://huggingface.co/papers/2407.01047 | 1 | 0 | 0 | 3 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.540.bib | https://aclanthology.org/2024.emnlp-main.540/ | @inproceedings{zhang-etal-2024-modeling,
title = "Modeling Layout Reading Order as Ordering Relations for Visually-rich Document Understanding",
author = "Zhang, Chong and
Tu, Yi and
Zhao, Yixi and
Yuan, Chenshu and
Chen, Huan and
Zhang, Yue and
Chai, Mingxu and
Guo, Ya and
Zhu, Huijia and
Zhang, Qi and
Gui, Tao",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.540",
pages = "9658--9678",
abstract = "Modeling and leveraging layout reading order in visually-rich documents (VrDs) is critical in document intelligence as it captures the rich structure semantics within documents.Previous works typically formulated layout reading order as a permutation of layout elements, i.e. a sequence containing all the layout elements.However, we argue that this formulation does not adequately convey the complete reading order information in the layout, which may potentially lead to performance decline in downstream tasks.To address this issue, we propose to model the layout reading order as ordering relations over the set of layout elements, which have sufficient expressive capability for the complete reading order information. To enable empirical evaluation on methods towards the improved form of reading order prediction (ROP), we establish a comprehensive benchmark dataset including the reading order annotation as relations over layout elements, together with a relation-extraction-based method that outperforms previous models. Moreover, we propose a reading-order-relation-enhancing pipeline to improve model performance on any arbitrary VrD task by introducing additional reading order relation inputs.We conduct comprehensive experiments to demonstrate that the pipeline generally benefits downstream VrD tasks:(1) with utilizing the reading order relation information, the enhanced downstream models achieve SOTA results on both two task settings of the targeted dataset; (2) with utilizing the pseudo reading order information generated by the proposed ROP model, the performance of the enhanced models has improved across all three models and eight cross-domain VrD-IE/QA task settings without targeted optimization.",
}
| Modeling and leveraging layout reading order in visually-rich documents (VrDs) is critical in document intelligence as it captures the rich structure semantics within documents.Previous works typically formulated layout reading order as a permutation of layout elements, i.e. a sequence containing all the layout elements.However, we argue that this formulation does not adequately convey the complete reading order information in the layout, which may potentially lead to performance decline in downstream tasks.To address this issue, we propose to model the layout reading order as ordering relations over the set of layout elements, which have sufficient expressive capability for the complete reading order information. To enable empirical evaluation on methods towards the improved form of reading order prediction (ROP), we establish a comprehensive benchmark dataset including the reading order annotation as relations over layout elements, together with a relation-extraction-based method that outperforms previous models. Moreover, we propose a reading-order-relation-enhancing pipeline to improve model performance on any arbitrary VrD task by introducing additional reading order relation inputs.We conduct comprehensive experiments to demonstrate that the pipeline generally benefits downstream VrD tasks:(1) with utilizing the reading order relation information, the enhanced downstream models achieve SOTA results on both two task settings of the targeted dataset; (2) with utilizing the pseudo reading order information generated by the proposed ROP model, the performance of the enhanced models has improved across all three models and eight cross-domain VrD-IE/QA task settings without targeted optimization. | [
"Zhang, Chong",
"Tu, Yi",
"Zhao, Yixi",
"Yuan, Chenshu",
"Chen, Huan",
"Zhang, Yue",
"Chai, Mingxu",
"Guo, Ya",
"Zhu, Huijia",
"Zhang, Qi",
"Gui, Tao"
] | Modeling Layout Reading Order as Ordering Relations for Visually-rich Document Understanding | emnlp-main.540 | Poster | 2409.19672 | [
"https://github.com/chongzhangFDU/ROOR"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.541.bib | https://aclanthology.org/2024.emnlp-main.541/ | @inproceedings{blouir-etal-2024-birdie,
title = "Birdie: Advancing State Space Language Modeling with Dynamic Mixtures of Training Objectives",
author = "Blouir, Sam and
Smith, Jimmy T.h. and
Anastasopoulos, Antonios and
Shehu, Amarda",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.541",
pages = "9679--9705",
abstract = "Efficient state space models (SSMs), including linear recurrent neural networks and linear attention variants, have emerged as potential alternative language models to Transformers. While efficient, SSMs struggle with tasks requiring in-context retrieval, such as text copying and associative recall, limiting their usefulness in practical settings. Prior work on how to meet this challenge has focused on the internal model architecture and not investigated the role of the training procedure. This paper proposes a new training procedure that improve the performance of SSMs on retrieval-intensive tasks. This novel pre-training procedure combines a bidirectional processing of the input with dynamic mixtures of pre-training objectives to improve the utilization of the SSM{'}s fixed-size state. Our experimental evaluations show that this procedure significantly improves performance on retrieval-intensive tasks that challenge current SSMs, such as phone book lookup, long paragraph question-answering, and infilling tasks. Our findings offer insights into a new direction to advance the training of SSMs to close the performance gap with Transformers.",
}
| Efficient state space models (SSMs), including linear recurrent neural networks and linear attention variants, have emerged as potential alternative language models to Transformers. While efficient, SSMs struggle with tasks requiring in-context retrieval, such as text copying and associative recall, limiting their usefulness in practical settings. Prior work on how to meet this challenge has focused on the internal model architecture and not investigated the role of the training procedure. This paper proposes a new training procedure that improve the performance of SSMs on retrieval-intensive tasks. This novel pre-training procedure combines a bidirectional processing of the input with dynamic mixtures of pre-training objectives to improve the utilization of the SSM{'}s fixed-size state. Our experimental evaluations show that this procedure significantly improves performance on retrieval-intensive tasks that challenge current SSMs, such as phone book lookup, long paragraph question-answering, and infilling tasks. Our findings offer insights into a new direction to advance the training of SSMs to close the performance gap with Transformers. | [
"Blouir, Sam",
"Smith, Jimmy T.h.",
"Anastasopoulos, Antonios",
"Shehu, Amarda"
] | Birdie: Advancing State Space Language Modeling with Dynamic Mixtures of Training Objectives | emnlp-main.541 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.542.bib | https://aclanthology.org/2024.emnlp-main.542/ | @inproceedings{chen-etal-2024-good-data,
title = "Is It Good Data for Multilingual Instruction Tuning or Just Bad Multilingual Evaluation for Large Language Models?",
author = "Chen, Pinzhen and
Yu, Simon and
Guo, Zhicheng and
Haddow, Barry",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.542",
pages = "9706--9726",
abstract = "Multilingual large language models are designed, claimed, and expected to cater to speakers of varied languages. We hypothesise that the current practices of fine-tuning and evaluating these models may not perfectly align with this objective owing to a heavy reliance on translation, which cannot cover language-specific knowledge but can introduce translation defects. It remains unknown whether the nature of the instruction data has an impact on the model output; conversely, it is questionable whether translated test sets can capture such nuances. Due to the often coupled practices of using translated data in both stages, such imperfections could have been overlooked. This work investigates these issues using controlled native or translated data during the instruction tuning and evaluation stages. We show that native or generation benchmarks reveal a notable difference between native and translated instruction data especially when model performance is high, whereas other types of test sets cannot. The comparison between round-trip and single-pass translations reflects the importance of knowledge from language-native resources. Finally, we demonstrate that regularization is beneficial to bridging this gap on structured but not generative tasks.",
}
| Multilingual large language models are designed, claimed, and expected to cater to speakers of varied languages. We hypothesise that the current practices of fine-tuning and evaluating these models may not perfectly align with this objective owing to a heavy reliance on translation, which cannot cover language-specific knowledge but can introduce translation defects. It remains unknown whether the nature of the instruction data has an impact on the model output; conversely, it is questionable whether translated test sets can capture such nuances. Due to the often coupled practices of using translated data in both stages, such imperfections could have been overlooked. This work investigates these issues using controlled native or translated data during the instruction tuning and evaluation stages. We show that native or generation benchmarks reveal a notable difference between native and translated instruction data especially when model performance is high, whereas other types of test sets cannot. The comparison between round-trip and single-pass translations reflects the importance of knowledge from language-native resources. Finally, we demonstrate that regularization is beneficial to bridging this gap on structured but not generative tasks. | [
"Chen, Pinzhen",
"Yu, Simon",
"Guo, Zhicheng",
"Haddow, Barry"
] | Is It Good Data for Multilingual Instruction Tuning or Just Bad Multilingual Evaluation for Large Language Models? | emnlp-main.542 | Poster | 2406.12822 | [
""
] | https://huggingface.co/papers/2406.12822 | 2 | 0 | 0 | 4 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.543.bib | https://aclanthology.org/2024.emnlp-main.543/ | @inproceedings{feucht-etal-2024-token,
title = "Token Erasure as a Footprint of Implicit Vocabulary Items in {LLM}s",
author = "Feucht, Sheridan and
Atkinson, David and
Wallace, Byron C and
Bau, David",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.543",
pages = "9727--9739",
abstract = "LLMs process text as sequences of tokens that roughly correspond to words, where less common words are represented by multiple tokens. However, individual tokens are often semantically unrelated to the meanings of the words/concepts they comprise. For example, Llama-2-7b{'}s tokenizer splits the word {``}patrolling{''} into two tokens, {``}pat{''} and {``}rolling{''}, neither of which correspond to semantically meaningful units like {``}patrol{''} or ''-ing.{''} Similarly, the overall meanings of named entities like {``}Neil Young{''} and multi-word expressions like {``}break a leg{''} cannot be directly inferred from their constituent tokens. Mechanistically, how do LLMs convert such arbitrary groups of tokens into useful higher-level representations? In this work, we find that last token representations of named entities and multi-token words exhibit a pronounced {``}erasure{''} effect, where information about previous and current tokens is rapidly forgotten in early layers. Using this observation, we propose a method to {``}read out{''} the implicit vocabulary of an autoregressive LLM by examining differences in token representations across layers, and present results of this method for Llama-2-7b and Llama-3-8B. To our knowledge, this is the first attempt to probe the implicit vocabulary of an LLM.",
}
| LLMs process text as sequences of tokens that roughly correspond to words, where less common words are represented by multiple tokens. However, individual tokens are often semantically unrelated to the meanings of the words/concepts they comprise. For example, Llama-2-7b{'}s tokenizer splits the word {``}patrolling{''} into two tokens, {``}pat{''} and {``}rolling{''}, neither of which correspond to semantically meaningful units like {``}patrol{''} or ''-ing.{''} Similarly, the overall meanings of named entities like {``}Neil Young{''} and multi-word expressions like {``}break a leg{''} cannot be directly inferred from their constituent tokens. Mechanistically, how do LLMs convert such arbitrary groups of tokens into useful higher-level representations? In this work, we find that last token representations of named entities and multi-token words exhibit a pronounced {``}erasure{''} effect, where information about previous and current tokens is rapidly forgotten in early layers. Using this observation, we propose a method to {``}read out{''} the implicit vocabulary of an autoregressive LLM by examining differences in token representations across layers, and present results of this method for Llama-2-7b and Llama-3-8B. To our knowledge, this is the first attempt to probe the implicit vocabulary of an LLM. | [
"Feucht, Sheridan",
"Atkinson, David",
"Wallace, Byron C",
"Bau, David"
] | Token Erasure as a Footprint of Implicit Vocabulary Items in LLMs | emnlp-main.543 | Poster | 2406.20086 | [
""
] | https://huggingface.co/papers/2406.20086 | 4 | 4 | 2 | 4 | [
"sfeucht/footprints"
] | [] | [] | [
"sfeucht/footprints"
] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.544.bib | https://aclanthology.org/2024.emnlp-main.544/ | @inproceedings{shang-etal-2024-traveler,
title = "{T}rave{LER}: A Modular Multi-{LMM} Agent Framework for Video Question-Answering",
author = "Shang, Chuyi and
You, Amos and
Subramanian, Sanjay and
Darrell, Trevor and
Herzig, Roei",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.544",
pages = "9740--9766",
abstract = "Recently, image-based Large Multimodal Models (LMMs) have made significant progress in video question-answering (VideoQA) using a frame-wise approach by leveraging large-scale pretraining in a zero-shot manner. Nevertheless, these models need to be capable of finding relevant information, extracting it, and answering the question simultaneously. Currently, existing methods perform all of these steps in a single pass without being able to adapt if insufficient or incorrect information is collected. To overcome this, we introduce a modular multi-LMM agent framework based on several agents with different roles, instructed by a Planner agent that updates its instructions using shared feedback from the other agents. Specifically, we propose TraveLER, a method that can create a plan to ''**Trave**rse{''} through the video, ask questions about individual frames to ''**L**ocate{''} and store key information, and then ''**E**valuate{''} if there is enough information to answer the question. Finally, if there is not enough information, our method is able to ''**R**eplan{''} based on its collected knowledge. Through extensive experiments, we find that the proposed TraveLER approach improves performance on several VideoQA benchmarks without the need to fine-tune on specific datasets. Our code is available at https://github.com/traveler-framework/TraveLER.",
}
| Recently, image-based Large Multimodal Models (LMMs) have made significant progress in video question-answering (VideoQA) using a frame-wise approach by leveraging large-scale pretraining in a zero-shot manner. Nevertheless, these models need to be capable of finding relevant information, extracting it, and answering the question simultaneously. Currently, existing methods perform all of these steps in a single pass without being able to adapt if insufficient or incorrect information is collected. To overcome this, we introduce a modular multi-LMM agent framework based on several agents with different roles, instructed by a Planner agent that updates its instructions using shared feedback from the other agents. Specifically, we propose TraveLER, a method that can create a plan to ''**Trave**rse{''} through the video, ask questions about individual frames to ''**L**ocate{''} and store key information, and then ''**E**valuate{''} if there is enough information to answer the question. Finally, if there is not enough information, our method is able to ''**R**eplan{''} based on its collected knowledge. Through extensive experiments, we find that the proposed TraveLER approach improves performance on several VideoQA benchmarks without the need to fine-tune on specific datasets. Our code is available at https://github.com/traveler-framework/TraveLER. | [
"Shang, Chuyi",
"You, Amos",
"Subramanian, Sanjay",
"Darrell, Trevor",
"Herzig, Roei"
] | TraveLER: A Modular Multi-LMM Agent Framework for Video Question-Answering | emnlp-main.544 | Poster | 2404.01476 | [
"https://github.com/traveler-framework/traveler"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.545.bib | https://aclanthology.org/2024.emnlp-main.545/ | @inproceedings{mohapatra-etal-2024-evaluating,
title = "Evaluating the Effectiveness of Large Language Models in Establishing Conversational Grounding",
author = "Mohapatra, Biswesh and
Kapadnis, Manav Nitin and
Romary, Laurent and
Cassell, Justine",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.545",
pages = "9767--9781",
abstract = "Conversational grounding, vital for building dependable dialog systems, involves ensuring a mutual understanding of shared information. Despite its importance, there has been limited research on this aspect of conversation in recent years, especially after the advent of Large Language Models (LLMs). Previous studies have highlighted the shortcomings of pre-trained language models in conversational grounding. However, most testing for conversational grounding capabilities involves human evaluations that are costly and time-consuming. This has led to a lack of testing across multiple models of varying sizes, a critical need given the rapid rate of new model releases. This gap in research becomes more significant considering recent advances in language models, which have led to new emergent capabilities. In this paper, we aim to evaluate the performance of LLMs in various aspects of conversational grounding and analyze why some models perform better than others. We demonstrate a direct correlation between the size of the pre-training data and conversational grounding abilities, meaning that they have independently acquired a specific form of pragmatic capabilities from larger pre-training datasets. Finally, we propose ways to enhance the capabilities of the models that lag in this aspect.",
}
| Conversational grounding, vital for building dependable dialog systems, involves ensuring a mutual understanding of shared information. Despite its importance, there has been limited research on this aspect of conversation in recent years, especially after the advent of Large Language Models (LLMs). Previous studies have highlighted the shortcomings of pre-trained language models in conversational grounding. However, most testing for conversational grounding capabilities involves human evaluations that are costly and time-consuming. This has led to a lack of testing across multiple models of varying sizes, a critical need given the rapid rate of new model releases. This gap in research becomes more significant considering recent advances in language models, which have led to new emergent capabilities. In this paper, we aim to evaluate the performance of LLMs in various aspects of conversational grounding and analyze why some models perform better than others. We demonstrate a direct correlation between the size of the pre-training data and conversational grounding abilities, meaning that they have independently acquired a specific form of pragmatic capabilities from larger pre-training datasets. Finally, we propose ways to enhance the capabilities of the models that lag in this aspect. | [
"Mohapatra, Biswesh",
"Kapadnis, Manav Nitin",
"Romary, Laurent",
"Cassell, Justine"
] | Evaluating the Effectiveness of Large Language Models in Establishing Conversational Grounding | emnlp-main.545 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.546.bib | https://aclanthology.org/2024.emnlp-main.546/ | @inproceedings{wang-etal-2024-unlocking,
title = "Unlocking Memorization in Large Language Models with Dynamic Soft Prompting",
author = "Wang, Zhepeng and
Bao, Runxue and
Wu, Yawen and
Taylor, Jackson and
Xiao, Cao and
Zheng, Feng and
Jiang, Weiwen and
Gao, Shangqian and
Zhang, Yanfu",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.546",
pages = "9782--9796",
abstract = "Pretrained large language models (LLMs) have excelled in a variety of natural language processing (NLP) tasks, including summarization, question answering, and translation. However, LLMs pose significant security risks due to their tendency to memorize training data, leading to potential privacy breaches and copyright infringement. Therefore, accurate measurement of the memorization is essential to evaluate and mitigate these potential risks. However, previous attempts to characterize memorization are constrained by either using prefixes only or by prepending a constant soft prompt to the prefixes, which cannot react to changes in input. To address this challenge, we propose a novel method for estimating LLM memorization using dynamic, prefix-dependent soft prompts. Our approach involves training a transformer-based generator to produce soft prompts that adapt to changes in input, thereby enabling more accurate extraction of memorized data. Our method not only addresses the limitations of previous methods but also demonstrates superior performance in diverse experimental settings compared to state-of-the-art techniques. In particular, our method can achieve the maximum relative improvement of 135.3{\%} and 39.8{\%} over the vanilla baseline on average in terms of *discoverable memorization rate* for the text generation task and code generation task, respectively. Our code is available at https://github.com/wangger/llm-memorization-dsp.",
}
| Pretrained large language models (LLMs) have excelled in a variety of natural language processing (NLP) tasks, including summarization, question answering, and translation. However, LLMs pose significant security risks due to their tendency to memorize training data, leading to potential privacy breaches and copyright infringement. Therefore, accurate measurement of the memorization is essential to evaluate and mitigate these potential risks. However, previous attempts to characterize memorization are constrained by either using prefixes only or by prepending a constant soft prompt to the prefixes, which cannot react to changes in input. To address this challenge, we propose a novel method for estimating LLM memorization using dynamic, prefix-dependent soft prompts. Our approach involves training a transformer-based generator to produce soft prompts that adapt to changes in input, thereby enabling more accurate extraction of memorized data. Our method not only addresses the limitations of previous methods but also demonstrates superior performance in diverse experimental settings compared to state-of-the-art techniques. In particular, our method can achieve the maximum relative improvement of 135.3{\%} and 39.8{\%} over the vanilla baseline on average in terms of *discoverable memorization rate* for the text generation task and code generation task, respectively. Our code is available at https://github.com/wangger/llm-memorization-dsp. | [
"Wang, Zhepeng",
"Bao, Runxue",
"Wu, Yawen",
"Taylor, Jackson",
"Xiao, Cao",
"Zheng, Feng",
"Jiang, Weiwen",
"Gao, Shangqian",
"Zhang, Yanfu"
] | Unlocking Memorization in Large Language Models with Dynamic Soft Prompting | emnlp-main.546 | Poster | 2409.13853 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.547.bib | https://aclanthology.org/2024.emnlp-main.547/ | @inproceedings{esfandiarpoor-etal-2024-clip,
title = "If {CLIP} Could Talk: Understanding Vision-Language Model Representations Through Their Preferred Concept Descriptions",
author = "Esfandiarpoor, Reza and
Menghini, Cristina and
Bach, Stephen",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.547",
pages = "9797--9819",
abstract = "Recent works often assume that Vision-Language Model (VLM) representations are based on visual attributes like shape. However, it is unclear to what extent VLMs prioritize this information to represent concepts. We propose Extract and Explore (EX2), a novel approach to characterize textual features that are important for VLMs. EX2 uses reinforcement learning to align a large language model with VLM preferences and generates descriptions that incorporate features that are important for the VLM. Then, we inspect the descriptions to identify features that contribute to VLM representations. Using EX2, we find that spurious descriptions have a major role in VLM representations despite providing no helpful information, e.g., Click to enlarge photo of CONCEPT. More importantly, among informative descriptions, VLMs rely significantly on non-visual attributes like habitat (e.g., North America) to represent visual concepts. Also, our analysis reveals that different VLMs prioritize different attributes in their representations. Overall, we show that VLMs do not simply match images to scene descriptions and that non-visual or even spurious descriptions significantly influence their representations.",
}
| Recent works often assume that Vision-Language Model (VLM) representations are based on visual attributes like shape. However, it is unclear to what extent VLMs prioritize this information to represent concepts. We propose Extract and Explore (EX2), a novel approach to characterize textual features that are important for VLMs. EX2 uses reinforcement learning to align a large language model with VLM preferences and generates descriptions that incorporate features that are important for the VLM. Then, we inspect the descriptions to identify features that contribute to VLM representations. Using EX2, we find that spurious descriptions have a major role in VLM representations despite providing no helpful information, e.g., Click to enlarge photo of CONCEPT. More importantly, among informative descriptions, VLMs rely significantly on non-visual attributes like habitat (e.g., North America) to represent visual concepts. Also, our analysis reveals that different VLMs prioritize different attributes in their representations. Overall, we show that VLMs do not simply match images to scene descriptions and that non-visual or even spurious descriptions significantly influence their representations. | [
"Esf",
"iarpoor, Reza",
"Menghini, Cristina",
"Bach, Stephen"
] | If CLIP Could Talk: Understanding Vision-Language Model Representations Through Their Preferred Concept Descriptions | emnlp-main.547 | Poster | 2403.16442 | [
"https://github.com/batsresearch/ex2"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.548.bib | https://aclanthology.org/2024.emnlp-main.548/ | @inproceedings{zhang-soh-2024-extract,
title = "Extract, Define, Canonicalize: An {LLM}-based Framework for Knowledge Graph Construction",
author = "Zhang, Bowen and
Soh, Harold",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.548",
pages = "9820--9836",
abstract = "In this work, we are interested in automated methods for knowledge graph creation (KGC) from input text. Progress on large language models (LLMs) has prompted a series of recent works applying them to KGC, e.g., via zero/few-shot prompting. Despite successes on small domain-specific datasets, these models face difficulties scaling up to text common in many real-world applications. A principal issue is that, in prior methods, the KG schema has to be included in the LLM prompt to generate valid triplets; larger and more complex schemas easily exceed the LLMs{'} context window length. Furthermore, there are scenarios where a fixed pre-defined schema is not available and we would like the method to construct a high-quality KG with a succinct self-generated schema. To address these problems, we propose a three-phase framework named Extract-Define-Canonicalize (EDC): open information extraction followed by schema definition and post-hoc canonicalization. EDC is flexible in that it can be applied to settings where a pre-defined target schema is available and when it is not; in the latter case, it constructs a schema automatically and applies self-canonicalization. To further improve performance, we introduce a trained component that retrieves schema elements relevant to the input text; this improves the LLMs{'} extraction performance in a retrieval-augmented generation-like manner. We demonstrate on three KGC benchmarks that EDC is able to extract high-quality triplets without any parameter tuning and with significantly larger schemas compared to prior works. Code for EDC is available at https://github.com/clear-nus/edc.",
}
| In this work, we are interested in automated methods for knowledge graph creation (KGC) from input text. Progress on large language models (LLMs) has prompted a series of recent works applying them to KGC, e.g., via zero/few-shot prompting. Despite successes on small domain-specific datasets, these models face difficulties scaling up to text common in many real-world applications. A principal issue is that, in prior methods, the KG schema has to be included in the LLM prompt to generate valid triplets; larger and more complex schemas easily exceed the LLMs{'} context window length. Furthermore, there are scenarios where a fixed pre-defined schema is not available and we would like the method to construct a high-quality KG with a succinct self-generated schema. To address these problems, we propose a three-phase framework named Extract-Define-Canonicalize (EDC): open information extraction followed by schema definition and post-hoc canonicalization. EDC is flexible in that it can be applied to settings where a pre-defined target schema is available and when it is not; in the latter case, it constructs a schema automatically and applies self-canonicalization. To further improve performance, we introduce a trained component that retrieves schema elements relevant to the input text; this improves the LLMs{'} extraction performance in a retrieval-augmented generation-like manner. We demonstrate on three KGC benchmarks that EDC is able to extract high-quality triplets without any parameter tuning and with significantly larger schemas compared to prior works. Code for EDC is available at https://github.com/clear-nus/edc. | [
"Zhang, Bowen",
"Soh, Harold"
] | Extract, Define, Canonicalize: An LLM-based Framework for Knowledge Graph Construction | emnlp-main.548 | Poster | 2404.03868 | [
"https://github.com/clear-nus/edc"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.549.bib | https://aclanthology.org/2024.emnlp-main.549/ | @inproceedings{liu-etal-2024-mquine,
title = "{MQ}uin{E}: a Cure for {``}{Z}-paradox{''} in Knowledge Graph Embedding",
author = "Liu, Yang and
Fang, Huang and
Cai, Yunfeng and
Sun, Mingming",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.549",
pages = "9837--9850",
abstract = "Knowledge graph embedding (KGE) models achieved state-of-the-art results on many knowledge graph tasks including link prediction and information retrieval. Despite the superior performance of KGE models in practice, we discover a deficiency in the expressiveness of some popular existing KGE models called \textit{Z-paradox}. Motivated by the existence of Z-paradox, we propose a new KGE model called \textit{MQuinE} that does not suffer from Z-paradox while preserves strong expressiveness to model various relation patterns including symmetric/asymmetric, inverse, 1-N/N-1/N-N, and composition relations with theoretical justification. Experiments on real-world knowledge bases indicate that Z-paradox indeed degrades the performance of existing KGE models, and can cause more than 20{\%} accuracy drop on some challenging test samples. Our experiments further demonstrate that MQuinE can mitigate the negative impact of Z-paradox and outperform existing KGE models by a visible margin on link prediction tasks.",
}
| Knowledge graph embedding (KGE) models achieved state-of-the-art results on many knowledge graph tasks including link prediction and information retrieval. Despite the superior performance of KGE models in practice, we discover a deficiency in the expressiveness of some popular existing KGE models called \textit{Z-paradox}. Motivated by the existence of Z-paradox, we propose a new KGE model called \textit{MQuinE} that does not suffer from Z-paradox while preserves strong expressiveness to model various relation patterns including symmetric/asymmetric, inverse, 1-N/N-1/N-N, and composition relations with theoretical justification. Experiments on real-world knowledge bases indicate that Z-paradox indeed degrades the performance of existing KGE models, and can cause more than 20{\%} accuracy drop on some challenging test samples. Our experiments further demonstrate that MQuinE can mitigate the negative impact of Z-paradox and outperform existing KGE models by a visible margin on link prediction tasks. | [
"Liu, Yang",
"Fang, Huang",
"Cai, Yunfeng",
"Sun, Mingming"
] | MQuinE: a Cure for “Z-paradox” in Knowledge Graph Embedding | emnlp-main.549 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.550.bib | https://aclanthology.org/2024.emnlp-main.550/ | @inproceedings{svete-etal-2024-transformers,
title = "Can Transformers Learn $n$-gram Language Models?",
author = "Svete, Anej and
Borenstein, Nadav and
Zhou, Mike and
Augenstein, Isabelle and
Cotterell, Ryan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.550",
pages = "9851--9867",
abstract = "Much theoretical work has described the ability of transformers to represent formal languages. However, linking theoretical results to empirical performance is not straightforward due to the complex interplay between the architecture, the learning algorithm, and training data. To test whether theoretical lower bounds imply \textit{learnability} of formal languages, we turn to recent work relating transformers to $n$-gram language models (LMs). We study transformers{'} ability to learn random $n$-gram LMs of two kinds: ones with arbitrary next-symbol probabilities and ones where those are defined with shared parameters. We find that classic estimation techniques for $n$-gram LMs such as add-$\lambda$ smoothing outperform transformers on the former, while transformers perform better on the latter, outperforming methods specifically designed to learn $n$-gram LMs.",
}
| Much theoretical work has described the ability of transformers to represent formal languages. However, linking theoretical results to empirical performance is not straightforward due to the complex interplay between the architecture, the learning algorithm, and training data. To test whether theoretical lower bounds imply \textit{learnability} of formal languages, we turn to recent work relating transformers to $n$-gram language models (LMs). We study transformers{'} ability to learn random $n$-gram LMs of two kinds: ones with arbitrary next-symbol probabilities and ones where those are defined with shared parameters. We find that classic estimation techniques for $n$-gram LMs such as add-$\lambda$ smoothing outperform transformers on the former, while transformers perform better on the latter, outperforming methods specifically designed to learn $n$-gram LMs. | [
"Svete, Anej",
"Borenstein, Nadav",
"Zhou, Mike",
"Augenstein, Isabelle",
"Cotterell, Ryan"
] | Can Transformers Learn n-gram Language Models? | emnlp-main.550 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.551.bib | https://aclanthology.org/2024.emnlp-main.551/ | @inproceedings{kwon-etal-2024-stableprompt,
title = "{S}table{P}rompt : Automatic Prompt Tuning using Reinforcement Learning for Large Language Model",
author = "Kwon, Minchan and
Kim, Gaeun and
Kim, Jongsuk and
Lee, Haeil and
Kim, Junmo",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.551",
pages = "9868--9884",
abstract = "Finding appropriate prompts for the specific task has become an important issue as the usage of Large Language Models (LLM) have expanded. However, the variety of input-output formats complicate finding the prompts. Reinforcement Learning (RL) is a promising for prompt tuning due to its ability to incrementally produce better results through interaction with the environment. But its inherent training instability and environmental dependency make it difficult to use in practice. In this paper, we propose StablePrompt, a prompt tuning method based on RL. We formulate prompt tuning as RL problem between agent and target LLM, and introduce Adaptive Proximal Policy Optimization (APPO), an modified version of PPO for prompt tuning. APPO introduces an anchor model and updates it adaptively based on the training trajectory. Using this anchor model for the KL divergence term in PPO keeps the search space flexible and ensures training stability. We evaluate StablePrompt on various tasks, including text classification, question answering, and text generation. StablePrompt achieves State-of-The-Art performance across diverse tasks. We demonstrates that StablePrompt performs well across various types and sizes of LLMs. Furthermore, we present TTE-StablePrompt, an extension for generating input-dependent prompts. It outperforms StablePrompt in tasks that are hard to solve with a single prompt. This shows that StablePrompt is an extensible and stable RL framework for LLM.",
}
| Finding appropriate prompts for the specific task has become an important issue as the usage of Large Language Models (LLM) have expanded. However, the variety of input-output formats complicate finding the prompts. Reinforcement Learning (RL) is a promising for prompt tuning due to its ability to incrementally produce better results through interaction with the environment. But its inherent training instability and environmental dependency make it difficult to use in practice. In this paper, we propose StablePrompt, a prompt tuning method based on RL. We formulate prompt tuning as RL problem between agent and target LLM, and introduce Adaptive Proximal Policy Optimization (APPO), an modified version of PPO for prompt tuning. APPO introduces an anchor model and updates it adaptively based on the training trajectory. Using this anchor model for the KL divergence term in PPO keeps the search space flexible and ensures training stability. We evaluate StablePrompt on various tasks, including text classification, question answering, and text generation. StablePrompt achieves State-of-The-Art performance across diverse tasks. We demonstrates that StablePrompt performs well across various types and sizes of LLMs. Furthermore, we present TTE-StablePrompt, an extension for generating input-dependent prompts. It outperforms StablePrompt in tasks that are hard to solve with a single prompt. This shows that StablePrompt is an extensible and stable RL framework for LLM. | [
"Kwon, Minchan",
"Kim, Gaeun",
"Kim, Jongsuk",
"Lee, Haeil",
"Kim, Junmo"
] | StablePrompt : Automatic Prompt Tuning using Reinforcement Learning for Large Language Model | emnlp-main.551 | Poster | [
"https://github.com/kmc0207/Stableprompt"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.552.bib | https://aclanthology.org/2024.emnlp-main.552/ | @inproceedings{laban-etal-2024-summary,
title = "Summary of a Haystack: A Challenge to Long-Context {LLM}s and {RAG} Systems",
author = "Laban, Philippe and
Fabbri, Alexander and
Xiong, Caiming and
Wu, Chien-Sheng",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.552",
pages = "9885--9903",
abstract = "LLMs and RAG systems are now capable of handling millions of input tokens or more. However, evaluating the output quality of such systems on long-context tasks remains challenging, as tasks like Needle-in-a-Haystack lack complexity. In this work, we argue that summarization can play a central role in such evaluation. We design a procedure to synthesize Haystacks of documents, ensuring that specific insights repeat across documents. The {``}Summary of a Haystack{''} (SummHay) task then requires a system to process the Haystack and generate, given a query, a summary that identifies the relevant insights and precisely cites the source documents. Since we have precise knowledge of what insights should appear in a haystack summary and what documents should be cited, we implement a highly reproducible automatic evaluation that can score summaries on two aspects {--} Coverage and Citation. We generate Haystacks in two domains (conversation, news), and perform a large-scale evaluation of 10 LLMs and corresponding 50 RAG systems. Our findings indicate that SummHay is an open challenge for current systems, as even systems provided with an Oracle signal of document relevance lag our estimate of human performance (56{\%}) by 10+ points on a Joint Score. Without a retriever, long-context LLMs like GPT-4o and Claude 3 Opus score below 20{\%} on SummHay. We show SummHay can also be used to study enterprise RAG systems and position bias in long-context models. We hope future systems can equal and surpass human performance on SummHay.",
}
| LLMs and RAG systems are now capable of handling millions of input tokens or more. However, evaluating the output quality of such systems on long-context tasks remains challenging, as tasks like Needle-in-a-Haystack lack complexity. In this work, we argue that summarization can play a central role in such evaluation. We design a procedure to synthesize Haystacks of documents, ensuring that specific insights repeat across documents. The {``}Summary of a Haystack{''} (SummHay) task then requires a system to process the Haystack and generate, given a query, a summary that identifies the relevant insights and precisely cites the source documents. Since we have precise knowledge of what insights should appear in a haystack summary and what documents should be cited, we implement a highly reproducible automatic evaluation that can score summaries on two aspects {--} Coverage and Citation. We generate Haystacks in two domains (conversation, news), and perform a large-scale evaluation of 10 LLMs and corresponding 50 RAG systems. Our findings indicate that SummHay is an open challenge for current systems, as even systems provided with an Oracle signal of document relevance lag our estimate of human performance (56{\%}) by 10+ points on a Joint Score. Without a retriever, long-context LLMs like GPT-4o and Claude 3 Opus score below 20{\%} on SummHay. We show SummHay can also be used to study enterprise RAG systems and position bias in long-context models. We hope future systems can equal and surpass human performance on SummHay. | [
"Laban, Philippe",
"Fabbri, Alex",
"er",
"Xiong, Caiming",
"Wu, Chien-Sheng"
] | Summary of a Haystack: A Challenge to Long-Context LLMs and RAG Systems | emnlp-main.552 | Poster | 2407.01370 | [
"https://github.com/salesforce/summary-of-a-haystack"
] | https://huggingface.co/papers/2407.01370 | 4 | 85 | 6 | 4 | [] | [
"Salesforce/summary-of-a-haystack"
] | [] | [] | [
"Salesforce/summary-of-a-haystack"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.553.bib | https://aclanthology.org/2024.emnlp-main.553/ | @inproceedings{wang-etal-2024-multi-pass,
title = "Multi-pass Decoding for Grammatical Error Correction",
author = "Wang, Xiaoying and
Mu, Lingling and
Zhang, Jingyi and
Xu, Hongfei",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.553",
pages = "9904--9916",
abstract = "Sequence-to-sequence (seq2seq) models achieve comparable or better grammatical error correction performance compared to sequence-to-edit (seq2edit) models. Seq2edit models normally iteratively refine the correction result, while seq2seq models decode only once without aware of subsequent tokens. Iteratively refining the correction results of seq2seq models via Multi-Pass Decoding (MPD) may lead to better performance. However, MPD increases the inference costs. Deleting or replacing corrections in previous rounds may lose useful information in the source input. We present an early-stop mechanism to alleviate the efficiency issue. To address the source information loss issue, we propose to merge the source input with the previous round correction result into one sequence. Experiments on the CoNLL-14 test set and BEA-19 test set show that our approach can lead to consistent and significant improvements over strong BART and T5 baselines (+1.80, +1.35, and +2.02 F0.5 for BART 12-2, large and T5 large respectively on CoNLL-14 and +2.99, +1.82, and +2.79 correspondingly on BEA-19), obtaining F0.5 scores of 68.41 and 75.36 on CoNLL-14 and BEA-19 respectively.",
}
| Sequence-to-sequence (seq2seq) models achieve comparable or better grammatical error correction performance compared to sequence-to-edit (seq2edit) models. Seq2edit models normally iteratively refine the correction result, while seq2seq models decode only once without aware of subsequent tokens. Iteratively refining the correction results of seq2seq models via Multi-Pass Decoding (MPD) may lead to better performance. However, MPD increases the inference costs. Deleting or replacing corrections in previous rounds may lose useful information in the source input. We present an early-stop mechanism to alleviate the efficiency issue. To address the source information loss issue, we propose to merge the source input with the previous round correction result into one sequence. Experiments on the CoNLL-14 test set and BEA-19 test set show that our approach can lead to consistent and significant improvements over strong BART and T5 baselines (+1.80, +1.35, and +2.02 F0.5 for BART 12-2, large and T5 large respectively on CoNLL-14 and +2.99, +1.82, and +2.79 correspondingly on BEA-19), obtaining F0.5 scores of 68.41 and 75.36 on CoNLL-14 and BEA-19 respectively. | [
"Wang, Xiaoying",
"Mu, Lingling",
"Zhang, Jingyi",
"Xu, Hongfei"
] | Multi-pass Decoding for Grammatical Error Correction | emnlp-main.553 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.554.bib | https://aclanthology.org/2024.emnlp-main.554/ | @inproceedings{jiang-etal-2024-unknown,
title = "Into the Unknown Unknowns: Engaged Human Learning through Participation in Language Model Agent Conversations",
author = "Jiang, Yucheng and
Shao, Yijia and
Ma, Dekun and
Semnani, Sina and
Lam, Monica",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.554",
pages = "9917--9955",
abstract = "While language model (LM)-powered chatbots and generative search engines excel at answering concrete queries, discovering information in the terrain of unknown unknowns remains challenging for users. To emulate the common educational scenario where children/students learn by listening to and participating in conversations of their parents/teachers, we create Collaborative STORM (Co-STORM). Unlike QA systems that require users to ask all the questions, Co-STORM lets users observe and occasionally steer the discourse among several LM agents. The agents ask questions on the user{'}s behalf, allowing the user to discover unknown unknowns serendipitously. To facilitate user interaction, Co-STORM assists users in tracking the discourse by organizing the uncovered information into a dynamic mind map, ultimately generating a comprehensive report as takeaways. For automatic evaluation, we construct the WildSeek dataset by collecting real information-seeking records with user goals. Co-STORM outperforms baseline methods on both discourse trace and report quality. In a further human evaluation, 70{\%} of participants prefer Co-STORM over a search engine, and 78{\%} favor it over a RAG chatbot.",
}
| While language model (LM)-powered chatbots and generative search engines excel at answering concrete queries, discovering information in the terrain of unknown unknowns remains challenging for users. To emulate the common educational scenario where children/students learn by listening to and participating in conversations of their parents/teachers, we create Collaborative STORM (Co-STORM). Unlike QA systems that require users to ask all the questions, Co-STORM lets users observe and occasionally steer the discourse among several LM agents. The agents ask questions on the user{'}s behalf, allowing the user to discover unknown unknowns serendipitously. To facilitate user interaction, Co-STORM assists users in tracking the discourse by organizing the uncovered information into a dynamic mind map, ultimately generating a comprehensive report as takeaways. For automatic evaluation, we construct the WildSeek dataset by collecting real information-seeking records with user goals. Co-STORM outperforms baseline methods on both discourse trace and report quality. In a further human evaluation, 70{\%} of participants prefer Co-STORM over a search engine, and 78{\%} favor it over a RAG chatbot. | [
"Jiang, Yucheng",
"Shao, Yijia",
"Ma, Dekun",
"Semnani, Sina",
"Lam, Monica"
] | Into the Unknown Unknowns: Engaged Human Learning through Participation in Language Model Agent Conversations | emnlp-main.554 | Poster | 2408.15232 | [
""
] | https://huggingface.co/papers/2408.15232 | 0 | 0 | 0 | 5 | [] | [
"YuchengJiang/WildSeek"
] | [] | [] | [
"YuchengJiang/WildSeek"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.555.bib | https://aclanthology.org/2024.emnlp-main.555/ | @inproceedings{tang-etal-2024-scoi,
title = "{SCOI}: Syntax-augmented Coverage-based In-context Example Selection for Machine Translation",
author = "Tang, Chenming and
Wang, Zhixiang and
Wu, Yunfang",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.555",
pages = "9956--9971",
abstract = "In-context learning (ICL) greatly improves the performance of large language models (LLMs) on various down-stream tasks, where the improvement highly depends on the quality of demonstrations. In this work, we introduce syntactic knowledge to select better in-context examples for machine translation (MT). We propose a new strategy, namely \textbf{S}yntax-augmented \textbf{CO}verage-based \textbf{I}n-context example selection (SCOI), leveraging the deep syntactic structure beyond conventional word matching. Specifically, we measure the set-level syntactic coverage by computing the coverage of polynomial terms with the help of a simplified tree-to-polynomial algorithm, and lexical coverage using word overlap. Furthermore, we devise an alternate selection approach to combine both coverage measures, taking advantage of syntactic and lexical information. We conduct experiments with two multi-lingual LLMs on six translation directions. Empirical results show that our proposed SCOI obtains the highest average COMET score among all learning-free methods, indicating that combining syntactic and lexical coverage successfully helps to select better in-context examples for MT. Our code is available at https://github.com/JamyDon/SCOI.",
}
| In-context learning (ICL) greatly improves the performance of large language models (LLMs) on various down-stream tasks, where the improvement highly depends on the quality of demonstrations. In this work, we introduce syntactic knowledge to select better in-context examples for machine translation (MT). We propose a new strategy, namely \textbf{S}yntax-augmented \textbf{CO}verage-based \textbf{I}n-context example selection (SCOI), leveraging the deep syntactic structure beyond conventional word matching. Specifically, we measure the set-level syntactic coverage by computing the coverage of polynomial terms with the help of a simplified tree-to-polynomial algorithm, and lexical coverage using word overlap. Furthermore, we devise an alternate selection approach to combine both coverage measures, taking advantage of syntactic and lexical information. We conduct experiments with two multi-lingual LLMs on six translation directions. Empirical results show that our proposed SCOI obtains the highest average COMET score among all learning-free methods, indicating that combining syntactic and lexical coverage successfully helps to select better in-context examples for MT. Our code is available at https://github.com/JamyDon/SCOI. | [
"Tang, Chenming",
"Wang, Zhixiang",
"Wu, Yunfang"
] | SCOI: Syntax-augmented Coverage-based In-context Example Selection for Machine Translation | emnlp-main.555 | Poster | 2408.04872 | [
"https://github.com/jamydon/scoi"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.556.bib | https://aclanthology.org/2024.emnlp-main.556/ | @inproceedings{wang-etal-2024-efficient,
title = "Efficient Temporal Extrapolation of Multimodal Large Language Models with Temporal Grounding Bridge",
author = "Wang, Yuxuan and
Wang, Yueqian and
Wu, Pengfei and
Liang, Jianxin and
Zhao, Dongyan and
Liu, Yang and
Zheng, Zilong",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.556",
pages = "9972--9987",
abstract = "Despite progress in multimodal large language models (MLLMs), the challenge of interpreting long-form videos in response to linguistic queries persists, largely due to the inefficiency in temporal grounding and limited pre-trained context window size. In this work, we introduce Temporal Grounding Bridge (TGB), a novel framework that bootstraps MLLMs with advanced temporal grounding capabilities and broadens their contextual scope. Our framework significantly enhances the temporal capabilities of current MLLMs through three key innovations: an efficient multi-span temporal grounding algorithm applied to low-dimension temporal features projected from flow; a multimodal length extrapolation training paradigm that utilizes low-dimension temporal features to extend the training context window size; and a bootstrapping framework that bridges our model with pluggable MLLMs without requiring annotation. We validate TGB across seven video benchmarks and demonstrate substantial performance improvements compared with prior MLLMs. Notably, our model, initially trained on sequences of four frames, effectively handles sequences up to 16 longer without sacrificing performance, highlighting its scalability and effectiveness in real-world applications. Our code is publicly available.",
}
| Despite progress in multimodal large language models (MLLMs), the challenge of interpreting long-form videos in response to linguistic queries persists, largely due to the inefficiency in temporal grounding and limited pre-trained context window size. In this work, we introduce Temporal Grounding Bridge (TGB), a novel framework that bootstraps MLLMs with advanced temporal grounding capabilities and broadens their contextual scope. Our framework significantly enhances the temporal capabilities of current MLLMs through three key innovations: an efficient multi-span temporal grounding algorithm applied to low-dimension temporal features projected from flow; a multimodal length extrapolation training paradigm that utilizes low-dimension temporal features to extend the training context window size; and a bootstrapping framework that bridges our model with pluggable MLLMs without requiring annotation. We validate TGB across seven video benchmarks and demonstrate substantial performance improvements compared with prior MLLMs. Notably, our model, initially trained on sequences of four frames, effectively handles sequences up to 16 longer without sacrificing performance, highlighting its scalability and effectiveness in real-world applications. Our code is publicly available. | [
"Wang, Yuxuan",
"Wang, Yueqian",
"Wu, Pengfei",
"Liang, Jianxin",
"Zhao, Dongyan",
"Liu, Yang",
"Zheng, Zilong"
] | Efficient Temporal Extrapolation of Multimodal Large Language Models with Temporal Grounding Bridge | emnlp-main.556 | Poster | 2402.16050 | [
"https://github.com/bigai-nlco/lstp-chat"
] | https://huggingface.co/papers/2402.16050 | 1 | 1 | 0 | 6 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.557.bib | https://aclanthology.org/2024.emnlp-main.557/ | @inproceedings{subbiah-etal-2024-storysumm,
title = "{STORYSUMM}: Evaluating Faithfulness in Story Summarization",
author = "Subbiah, Melanie and
Ladhak, Faisal and
Mishra, Akankshya and
Adams, Griffin Thomas and
Chilton, Lydia and
McKeown, Kathleen",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.557",
pages = "9988--10005",
abstract = "Human evaluation has been the gold standard for checking faithfulness in abstractive summarization. However, with a challenging source domain like narrative, multiple annotators can agree a summary is faithful, while missing details that are obvious errors only once pointed out. We therefore introduce a new dataset, StorySumm, comprising LLM summaries of short stories with localized faithfulness labels and error explanations. This benchmark is for evaluation methods, testing whether a given method can detect challenging inconsistencies. Using this dataset, we first show that any one human annotation protocol is likely to miss inconsistencies, and we advocate for pursuing a range of methods when establishing ground truth for a summarization dataset. We finally test recent automatic metrics and find that none of them achieve more than 70{\%} balanced accuracy on this task, demonstrating that it is a challenging benchmark for future work in faithfulness evaluation.",
}
| Human evaluation has been the gold standard for checking faithfulness in abstractive summarization. However, with a challenging source domain like narrative, multiple annotators can agree a summary is faithful, while missing details that are obvious errors only once pointed out. We therefore introduce a new dataset, StorySumm, comprising LLM summaries of short stories with localized faithfulness labels and error explanations. This benchmark is for evaluation methods, testing whether a given method can detect challenging inconsistencies. Using this dataset, we first show that any one human annotation protocol is likely to miss inconsistencies, and we advocate for pursuing a range of methods when establishing ground truth for a summarization dataset. We finally test recent automatic metrics and find that none of them achieve more than 70{\%} balanced accuracy on this task, demonstrating that it is a challenging benchmark for future work in faithfulness evaluation. | [
"Subbiah, Melanie",
"Ladhak, Faisal",
"Mishra, Akankshya",
"Adams, Griffin Thomas",
"Chilton, Lydia",
"McKeown, Kathleen"
] | STORYSUMM: Evaluating Faithfulness in Story Summarization | emnlp-main.557 | Poster | 2407.06501 | [
"https://github.com/melaniesubbiah/storysumm"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.558.bib | https://aclanthology.org/2024.emnlp-main.558/ | @inproceedings{yu-etal-2024-mmoe,
title = "{MM}o{E}: Enhancing Multimodal Models with Mixtures of Multimodal Interaction Experts",
author = "Yu, Haofei and
Qi, Zhengyang and
Jang, Lawrence Keunho and
Salakhutdinov, Russ and
Morency, Louis-Philippe and
Liang, Paul Pu",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.558",
pages = "10006--10030",
abstract = "Advances in multimodal models have greatly improved how interactions relevant to various tasks are modeled. Today{'}s multimodal models mainly focus on the correspondence between images and text, using this for tasks like image-text matching. However, this covers only a subset of real-world interactions. Novel interactions, such as sarcasm expressed through opposing spoken words and gestures or humor expressed through utterances and tone of voice, remain challenging. In this paper, we introduce an approach to enhance multimodal models, which we call Multimodal Mixtures of Experts (MMoE). The key idea in MMoE is to train separate expert models for each type of multimodal interaction, such as redundancy present in both modalities, uniqueness in one modality, or synergy that emerges when both modalities are fused. On a sarcasm detection task (MUStARD) and a humor detection task (URFUNNY), we obtain new state-of-the-art results. MMoE is also able to be applied to various types of models to gain improvement.",
}
| Advances in multimodal models have greatly improved how interactions relevant to various tasks are modeled. Today{'}s multimodal models mainly focus on the correspondence between images and text, using this for tasks like image-text matching. However, this covers only a subset of real-world interactions. Novel interactions, such as sarcasm expressed through opposing spoken words and gestures or humor expressed through utterances and tone of voice, remain challenging. In this paper, we introduce an approach to enhance multimodal models, which we call Multimodal Mixtures of Experts (MMoE). The key idea in MMoE is to train separate expert models for each type of multimodal interaction, such as redundancy present in both modalities, uniqueness in one modality, or synergy that emerges when both modalities are fused. On a sarcasm detection task (MUStARD) and a humor detection task (URFUNNY), we obtain new state-of-the-art results. MMoE is also able to be applied to various types of models to gain improvement. | [
"Yu, Haofei",
"Qi, Zhengyang",
"Jang, Lawrence Keunho",
"Salakhutdinov, Russ",
"Morency, Louis-Philippe",
"Liang, Paul Pu"
] | MMoE: Enhancing Multimodal Models with Mixtures of Multimodal Interaction Experts | emnlp-main.558 | Poster | 2311.09580 | [
"https://github.com/lwaekfjlk/mmoe"
] | https://huggingface.co/papers/2311.09580 | 0 | 0 | 0 | 4 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.559.bib | https://aclanthology.org/2024.emnlp-main.559/ | @inproceedings{zhang-etal-2024-omagent,
title = "{O}m{A}gent: A Multi-modal Agent Framework for Complex Video Understanding with Task Divide-and-Conquer",
author = "Zhang, Lu and
Zhao, Tiancheng and
Ying, Heting and
Ma, Yibo and
Lee, Kyusong",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.559",
pages = "10031--10045",
abstract = "Recent advancements in Large Language Models (LLMs) have expanded their capabilities to multimodal contexts, including comprehensive video understanding. However, processing extensive videos such as 24-hour CCTV footage or full-length films presents significant challenges due to the vast data and processing demands. Traditional methods, like extracting key frames or converting frames to text, often result in substantial information loss. To address these shortcomings, we develop OmAgent, efficiently stores and retrieves relevant video frames for specific queries, preserving the detailed content of videos. Additionally, it features an Divide-and-Conquer Loop capable of autonomous reasoning, dynamically invoking APIs and tools to enhance query processing and accuracy. This approach ensures robust video understanding, significantly reducing information loss. Experimental results affirm OmAgent{'}s efficacy in handling various types of videos and complex tasks. Moreover, we have endowed it with greater autonomy and a robust tool-calling system, enabling it to accomplish even more intricate tasks.",
}
| Recent advancements in Large Language Models (LLMs) have expanded their capabilities to multimodal contexts, including comprehensive video understanding. However, processing extensive videos such as 24-hour CCTV footage or full-length films presents significant challenges due to the vast data and processing demands. Traditional methods, like extracting key frames or converting frames to text, often result in substantial information loss. To address these shortcomings, we develop OmAgent, efficiently stores and retrieves relevant video frames for specific queries, preserving the detailed content of videos. Additionally, it features an Divide-and-Conquer Loop capable of autonomous reasoning, dynamically invoking APIs and tools to enhance query processing and accuracy. This approach ensures robust video understanding, significantly reducing information loss. Experimental results affirm OmAgent{'}s efficacy in handling various types of videos and complex tasks. Moreover, we have endowed it with greater autonomy and a robust tool-calling system, enabling it to accomplish even more intricate tasks. | [
"Zhang, Lu",
"Zhao, Tiancheng",
"Ying, Heting",
"Ma, Yibo",
"Lee, Kyusong"
] | OmAgent: A Multi-modal Agent Framework for Complex Video Understanding with Task Divide-and-Conquer | emnlp-main.559 | Poster | 2406.16620 | [
"https://github.com/om-ai-lab/OmAgent"
] | https://huggingface.co/papers/2406.16620 | 0 | 0 | 0 | 5 | [
"omlab/omchat-v2.0-13B-single-beta_hf"
] | [] | [] | [
"omlab/omchat-v2.0-13B-single-beta_hf"
] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.560.bib | https://aclanthology.org/2024.emnlp-main.560/ | @inproceedings{ai-etal-2024-enhancing,
title = "Enhancing Pre-Trained Generative Language Models with Question Attended Span Extraction on Machine Reading Comprehension",
author = "Ai, Lin and
Hui, Zheng and
Liu, Zizhou and
Hirschberg, Julia",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.560",
pages = "10046--10063",
}
| No abstract found | [
"Ai, Lin",
"Hui, Zheng",
"Liu, Zizhou",
"Hirschberg, Julia"
] | Enhancing Pre-Trained Generative Language Models with Question Attended Span Extraction on Machine Reading Comprehension | emnlp-main.560 | Poster | 2404.17991 | [
""
] | https://huggingface.co/papers/2404.17991 | 0 | 0 | 3 | 4 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.561.bib | https://aclanthology.org/2024.emnlp-main.561/ | @inproceedings{rao-etal-2024-commonit,
title = "{C}ommon{IT}: Commonality-Aware Instruction Tuning for Large Language Models via Data Partitions",
author = "Rao, Jun and
Liu, Xuebo and
Lian, Lian and
Cheng, Shengjun and
Liao, Yunjie and
Zhang, Min",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.561",
pages = "10064--10083",
abstract = "With instruction tuning, Large Language Models (LLMs) can enhance their ability to adhere to commands. Diverging from most works focusing on data mixing, our study concentrates on enhancing the model{'}s capabilities from the perspective of data sampling during training. Drawing inspiration from the human learning process, where it is generally easier to master solutions to similar topics through focused practice on a single type of topic, we introduce a novel instruction tuning strategy termed CommonIT: Commonality-aware Instruction Tuning. Specifically, we cluster instruction datasets into distinct groups with three proposed metrics Task, Embedding and Length). We ensure each training mini-batch, or {``}partition{''}, consists solely of data from a single group, which brings about both data randomness across mini-batches and intra-batch data similarity. Rigorous testing on LLaMa models demonstrates CommonIT{'}s effectiveness in enhancing the instruction-following capabilities of LLMs through IT datasets (FLAN, CoT, and Alpaca) and models (LLaMa2-7B, Qwen2-7B, LLaMa 13B, and BLOOM 7B). CommonIT consistently boosts an average improvement of 2.1{\%} on the general domain (i.e., the average score of Knowledge, Reasoning, Multilinguality and Coding) with the Length metric, and 5.2{\%} on the special domain (i.e., GSM, Openfunctions and Code) with the Task metric, and 3.8{\%} on the specific tasks (i.e., MMLU) with the Embedding metric. Code is available at https://github.com/raojay7/CommonIT.",
}
| With instruction tuning, Large Language Models (LLMs) can enhance their ability to adhere to commands. Diverging from most works focusing on data mixing, our study concentrates on enhancing the model{'}s capabilities from the perspective of data sampling during training. Drawing inspiration from the human learning process, where it is generally easier to master solutions to similar topics through focused practice on a single type of topic, we introduce a novel instruction tuning strategy termed CommonIT: Commonality-aware Instruction Tuning. Specifically, we cluster instruction datasets into distinct groups with three proposed metrics Task, Embedding and Length). We ensure each training mini-batch, or {``}partition{''}, consists solely of data from a single group, which brings about both data randomness across mini-batches and intra-batch data similarity. Rigorous testing on LLaMa models demonstrates CommonIT{'}s effectiveness in enhancing the instruction-following capabilities of LLMs through IT datasets (FLAN, CoT, and Alpaca) and models (LLaMa2-7B, Qwen2-7B, LLaMa 13B, and BLOOM 7B). CommonIT consistently boosts an average improvement of 2.1{\%} on the general domain (i.e., the average score of Knowledge, Reasoning, Multilinguality and Coding) with the Length metric, and 5.2{\%} on the special domain (i.e., GSM, Openfunctions and Code) with the Task metric, and 3.8{\%} on the specific tasks (i.e., MMLU) with the Embedding metric. Code is available at https://github.com/raojay7/CommonIT. | [
"Rao, Jun",
"Liu, Xuebo",
"Lian, Lian",
"Cheng, Shengjun",
"Liao, Yunjie",
"Zhang, Min"
] | CommonIT: Commonality-Aware Instruction Tuning for Large Language Models via Data Partitions | emnlp-main.561 | Poster | 2410.03077 | [
"https://github.com/raojay7/commonit"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.562.bib | https://aclanthology.org/2024.emnlp-main.562/ | @inproceedings{gu-diao-2024-esc,
title = "{ESC}: Efficient Speech Coding with Cross-Scale Residual Vector Quantized Transformers",
author = "Gu, Yuzhe and
Diao, Enmao",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.562",
pages = "10084--10096",
abstract = "Neural speech codecs aim to compress input signals into minimal bits while maintaining content quality in a low-latency manner. However, existing neural codecs often trade model complexity for reconstruction performance. These codecs primarily use convolutional blocks for feature transformation, which are not inherently suited for capturing the local redundancies in speech signals. To compensate, they require either adversarial discriminators or a large number of model parameters to enhance audio quality. In response to these challenges, we introduce the Efficient Speech Codec (ESC), a lightweight, parameter-efficient speech codec based on a cross-scale residual vector quantization scheme and transformers. Our model employs mirrored hierarchical window transformer blocks and performs step-wise decoding from coarse-to-fine feature representations. To enhance bitrate efficiency, we propose a novel combination of vector quantization techniques along with a pre-training paradigm. Extensive experiments demonstrate that ESC can achieve high-fidelity speech reconstruction with significantly lower model complexity, making it a promising alternative to existing convolutional audio codecs.",
}
| Neural speech codecs aim to compress input signals into minimal bits while maintaining content quality in a low-latency manner. However, existing neural codecs often trade model complexity for reconstruction performance. These codecs primarily use convolutional blocks for feature transformation, which are not inherently suited for capturing the local redundancies in speech signals. To compensate, they require either adversarial discriminators or a large number of model parameters to enhance audio quality. In response to these challenges, we introduce the Efficient Speech Codec (ESC), a lightweight, parameter-efficient speech codec based on a cross-scale residual vector quantization scheme and transformers. Our model employs mirrored hierarchical window transformer blocks and performs step-wise decoding from coarse-to-fine feature representations. To enhance bitrate efficiency, we propose a novel combination of vector quantization techniques along with a pre-training paradigm. Extensive experiments demonstrate that ESC can achieve high-fidelity speech reconstruction with significantly lower model complexity, making it a promising alternative to existing convolutional audio codecs. | [
"Gu, Yuzhe",
"Diao, Enmao"
] | ESC: Efficient Speech Coding with Cross-Scale Residual Vector Quantized Transformers | emnlp-main.562 | Poster | 2404.19441 | [
"https://github.com/yzguu830/efficient-speech-codec"
] | https://huggingface.co/papers/2404.19441 | 0 | 0 | 0 | 2 | [] | [
"Tracygu/efficient-speech-codec"
] | [] | [] | [
"Tracygu/efficient-speech-codec"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.563.bib | https://aclanthology.org/2024.emnlp-main.563/ | @inproceedings{lee-etal-2024-breaking,
title = "Breaking {R}e{LU} Barrier: Generalized {M}o{E}fication for Dense Pretrained Models",
author = "Lee, Jaeseong and
Hwang, Seung-won and
Park, Wonpyo and
Ji, Mingi",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.563",
pages = "10097--10107",
abstract = "As the scale of language models (LMs) continues to grow, there is a heightened interest in reducing the inference cost associated with these models. Mixture-of-Experts (MoEs) present an efficient alternative to dense models, while the existing methods to convert pretrained dense models to MoEs is limited to ReLU-based models with natural sparsity. This paper introduces G-MoEfication, applicable to arbitrary dense models, where ReLU-based activation sparsity assumptions no longer hold. For generalizations, we encounter the dilemma of needing to zero-out deactivated experts, while also avoiding excessive zeroing-out to retain dense activation information. We publicly release our code and report results conducted with mBERT, SantaCoder-1.1B, Phi-2-2.7B, and Falcon-7B demonstrating the efficacy of our approach in general scenarios: from multitask to multilingual, from fine-tuning to zero-shot evaluation.",
}
| As the scale of language models (LMs) continues to grow, there is a heightened interest in reducing the inference cost associated with these models. Mixture-of-Experts (MoEs) present an efficient alternative to dense models, while the existing methods to convert pretrained dense models to MoEs is limited to ReLU-based models with natural sparsity. This paper introduces G-MoEfication, applicable to arbitrary dense models, where ReLU-based activation sparsity assumptions no longer hold. For generalizations, we encounter the dilemma of needing to zero-out deactivated experts, while also avoiding excessive zeroing-out to retain dense activation information. We publicly release our code and report results conducted with mBERT, SantaCoder-1.1B, Phi-2-2.7B, and Falcon-7B demonstrating the efficacy of our approach in general scenarios: from multitask to multilingual, from fine-tuning to zero-shot evaluation. | [
"Lee, Jaeseong",
"Hwang, Seung-won",
"Park, Wonpyo",
"Ji, Mingi"
] | Breaking ReLU Barrier: Generalized MoEfication for Dense Pretrained Models | emnlp-main.563 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.564.bib | https://aclanthology.org/2024.emnlp-main.564/ | @inproceedings{xu-etal-2024-detecting,
title = "Detecting Subtle Differences between Human and Model Languages Using Spectrum of Relative Likelihood",
author = "Xu, Yang and
Wang, Yu and
An, Hao and
Liu, Zhichen and
Li, Yongyuan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.564",
pages = "10108--10121",
abstract = "Human and model-generated texts can be distinguished by examining the magnitude of likelihood in language. However, it is becoming increasingly difficult as language model{'}s capabilities of generating human-like texts keep evolving. This study provides a new perspective by using the relative likelihood values instead of absolute ones, and extracting useful features from the spectrum-view of likelihood for the human-model text detection task. We propose a detection procedure with two classification methods, supervised and heuristic-based, respectively, which results in competitive performances with previous zero-shot detection methods and a new state-of-the-art on short-text detection. Our method can also reveal subtle differences between human and model languages, which find theoretical roots in psycholinguistics studies.",
}
| Human and model-generated texts can be distinguished by examining the magnitude of likelihood in language. However, it is becoming increasingly difficult as language model{'}s capabilities of generating human-like texts keep evolving. This study provides a new perspective by using the relative likelihood values instead of absolute ones, and extracting useful features from the spectrum-view of likelihood for the human-model text detection task. We propose a detection procedure with two classification methods, supervised and heuristic-based, respectively, which results in competitive performances with previous zero-shot detection methods and a new state-of-the-art on short-text detection. Our method can also reveal subtle differences between human and model languages, which find theoretical roots in psycholinguistics studies. | [
"Xu, Yang",
"Wang, Yu",
"An, Hao",
"Liu, Zhichen",
"Li, Yongyuan"
] | Detecting Subtle Differences between Human and Model Languages Using Spectrum of Relative Likelihood | emnlp-main.564 | Poster | 2406.19874 | [
"https://github.com/clcs-sustech/fouriergpt"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.565.bib | https://aclanthology.org/2024.emnlp-main.565/ | @inproceedings{li-etal-2024-optimizing-language,
title = "Optimizing Language Models with Fair and Stable Reward Composition in Reinforcement Learning",
author = "Li, Jiahui and
Zhang, Hanlin and
Zhang, Fengda and
Chang, Tai-Wei and
Kuang, Kun and
Chen, Long and
Zhou, Jun",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.565",
pages = "10122--10140",
abstract = "Reinforcement learning from human feedback (RLHF) and AI-generated feedback (RLAIF) have become prominent techniques that significantly enhance the functionality of pre-trained language models (LMs). These methods harness feedback, sourced either from humans or AI, as direct rewards or to shape reward models that steer LM optimization. Nonetheless, the effective integration of rewards from diverse sources presents a significant challenge due to their disparate characteristics. To address this, recent research has developed algorithms incorporating strategies such as weighting, ranking, and constraining to handle this complexity. Despite these innovations, a bias toward disproportionately high rewards can still skew the reinforcement learning process and negatively impact LM performance. This paper explores a methodology for reward composition that enables simultaneous improvements in LMs across multiple dimensions. Inspired by fairness theory, we introduce a training algorithm that aims to reduce disparity and enhance stability among various rewards. Our method treats the aggregate reward as a dynamic weighted sum of individual rewards, with alternating updates to the weights and model parameters. For efficient and straightforward implementation, we employ an estimation technique rooted in the mirror descent method for weight updates, eliminating the need for gradient computations. The empirical results under various types of rewards across a wide range of scenarios demonstrate the effectiveness of our method.",
}
| Reinforcement learning from human feedback (RLHF) and AI-generated feedback (RLAIF) have become prominent techniques that significantly enhance the functionality of pre-trained language models (LMs). These methods harness feedback, sourced either from humans or AI, as direct rewards or to shape reward models that steer LM optimization. Nonetheless, the effective integration of rewards from diverse sources presents a significant challenge due to their disparate characteristics. To address this, recent research has developed algorithms incorporating strategies such as weighting, ranking, and constraining to handle this complexity. Despite these innovations, a bias toward disproportionately high rewards can still skew the reinforcement learning process and negatively impact LM performance. This paper explores a methodology for reward composition that enables simultaneous improvements in LMs across multiple dimensions. Inspired by fairness theory, we introduce a training algorithm that aims to reduce disparity and enhance stability among various rewards. Our method treats the aggregate reward as a dynamic weighted sum of individual rewards, with alternating updates to the weights and model parameters. For efficient and straightforward implementation, we employ an estimation technique rooted in the mirror descent method for weight updates, eliminating the need for gradient computations. The empirical results under various types of rewards across a wide range of scenarios demonstrate the effectiveness of our method. | [
"Li, Jiahui",
"Zhang, Hanlin",
"Zhang, Fengda",
"Chang, Tai-Wei",
"Kuang, Kun",
"Chen, Long",
"Zhou, Jun"
] | Optimizing Language Models with Fair and Stable Reward Composition in Reinforcement Learning | emnlp-main.565 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.566.bib | https://aclanthology.org/2024.emnlp-main.566/ | @inproceedings{feng-etal-2024-fine,
title = "Fine-grained Pluggable Gradient Ascent for Knowledge Unlearning in Language Models",
author = "Feng, XiaoHua and
Chen, Chaochao and
Li, Yuyuan and
Lin, Zibin",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.566",
pages = "10141--10155",
abstract = "Pre-trained language models acquire knowledge from vast amounts of text data, which can inadvertently contain sensitive information. To mitigate the presence of undesirable knowledge, the task of knowledge unlearning becomes crucial for language models. Previous research relies on gradient ascent methods to achieve knowledge unlearning, which is simple and effective. However, this approach calculates all the gradients of tokens in the sequence, potentially compromising the general ability of language models. To overcome this limitation, we propose an adaptive objective that calculates gradients with fine-grained control specifically targeting sensitive tokens. Our adaptive objective is pluggable, ensuring simplicity and enabling extension to the regularization-based framework that utilizes non-target data or other models to preserve general ability. Through extensive experiments targeting the removal of typical sensitive data, we demonstrate that our proposed method enhances the general ability of language models while achieving knowledge unlearning. Additionally, it demonstrates the capability to adapt to behavior alignment, eliminating all the undesirable knowledge within a specific domain.",
}
| Pre-trained language models acquire knowledge from vast amounts of text data, which can inadvertently contain sensitive information. To mitigate the presence of undesirable knowledge, the task of knowledge unlearning becomes crucial for language models. Previous research relies on gradient ascent methods to achieve knowledge unlearning, which is simple and effective. However, this approach calculates all the gradients of tokens in the sequence, potentially compromising the general ability of language models. To overcome this limitation, we propose an adaptive objective that calculates gradients with fine-grained control specifically targeting sensitive tokens. Our adaptive objective is pluggable, ensuring simplicity and enabling extension to the regularization-based framework that utilizes non-target data or other models to preserve general ability. Through extensive experiments targeting the removal of typical sensitive data, we demonstrate that our proposed method enhances the general ability of language models while achieving knowledge unlearning. Additionally, it demonstrates the capability to adapt to behavior alignment, eliminating all the undesirable knowledge within a specific domain. | [
"Feng, XiaoHua",
"Chen, Chaochao",
"Li, Yuyuan",
"Lin, Zibin"
] | Fine-grained Pluggable Gradient Ascent for Knowledge Unlearning in Language Models | emnlp-main.566 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.567.bib | https://aclanthology.org/2024.emnlp-main.567/ | @inproceedings{liu-etal-2024-arm,
title = "{ARM}: An Alignment-and-Replacement Module for {C}hinese Spelling Check Based on {LLM}s",
author = "Liu, Changchun and
Zhang, Kai and
Jiang, Junzhe and
Liu, Zirui and
Tao, Hanqing and
Gao, Min and
Chen, Enhong",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.567",
pages = "10156--10168",
abstract = "Chinese Spelling Check (CSC) aims to identify and correct spelling errors in Chinese texts, where enhanced semantic understanding of a sentence can significantly improve correction accuracy. Recently, Large Language Models (LLMs) have demonstrated exceptional mastery of world knowledge and semantic understanding, rendering them more robust against spelling errors. However, the application of LLMs in CSC is a double-edged sword, as they tend to unnecessarily alter sentence length and modify rare but correctly used phrases. In this paper, by leveraging the capabilities of LLMs while mitigating their limitations, we propose a novel plug-and-play Alignment-and-Replacement Module ARM that enhances the performance of existing CSC models and without the need for retraining or fine-tuning. Experiment results and analysis on three benchmark datasets demonstrate the effectiveness and competitiveness of the proposed module.",
}
| Chinese Spelling Check (CSC) aims to identify and correct spelling errors in Chinese texts, where enhanced semantic understanding of a sentence can significantly improve correction accuracy. Recently, Large Language Models (LLMs) have demonstrated exceptional mastery of world knowledge and semantic understanding, rendering them more robust against spelling errors. However, the application of LLMs in CSC is a double-edged sword, as they tend to unnecessarily alter sentence length and modify rare but correctly used phrases. In this paper, by leveraging the capabilities of LLMs while mitigating their limitations, we propose a novel plug-and-play Alignment-and-Replacement Module ARM that enhances the performance of existing CSC models and without the need for retraining or fine-tuning. Experiment results and analysis on three benchmark datasets demonstrate the effectiveness and competitiveness of the proposed module. | [
"Liu, Changchun",
"Zhang, Kai",
"Jiang, Junzhe",
"Liu, Zirui",
"Tao, Hanqing",
"Gao, Min",
"Chen, Enhong"
] | ARM: An Alignment-and-Replacement Module for Chinese Spelling Check Based on LLMs | emnlp-main.567 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.568.bib | https://aclanthology.org/2024.emnlp-main.568/ | @inproceedings{jiang-etal-2024-context,
title = "On the In-context Generation of Language Models",
author = "Jiang, Zhongtao and
Zhang, Yuanzhe and
Luo, Kun and
Yuan, Xiaowei and
Zhao, Jun and
Liu, Kang",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.568",
pages = "10169--10187",
abstract = "Large language models (LLMs) are found to have the ability of in-context generation (ICG): when they are fed with an in-context prompt concatenating a few somehow similar examples, they can implicitly recognize the pattern of them and then complete the prompt in the same pattern. ICG is curious, since language models are usually not explicitly trained in the same way as the in-context prompt, and the distribution of examples in the prompt differs from that of sequences in the pretrained corpora. This paper provides a systematic study of the ICG ability of language models, covering discussions about its source and influential factors, in the view of both theory and empirical experiments. Concretely, we first propose a plausible latent variable model to model the distribution of the pretrained corpora, and then formalize ICG as a problem of next topic prediction. With this framework, we can prove that the repetition nature of a few topics ensures the ICG ability on them theoretically. Then, we use this controllable pretrained distribution to generate several medium-scale synthetic datasets (token scale: 2.1B-3.9B) and experiment with different settings of Transformer architectures (parameter scale: 4M-234M). Our experimental results further offer insights into how the data and model architectures influence ICG.",
}
| Large language models (LLMs) are found to have the ability of in-context generation (ICG): when they are fed with an in-context prompt concatenating a few somehow similar examples, they can implicitly recognize the pattern of them and then complete the prompt in the same pattern. ICG is curious, since language models are usually not explicitly trained in the same way as the in-context prompt, and the distribution of examples in the prompt differs from that of sequences in the pretrained corpora. This paper provides a systematic study of the ICG ability of language models, covering discussions about its source and influential factors, in the view of both theory and empirical experiments. Concretely, we first propose a plausible latent variable model to model the distribution of the pretrained corpora, and then formalize ICG as a problem of next topic prediction. With this framework, we can prove that the repetition nature of a few topics ensures the ICG ability on them theoretically. Then, we use this controllable pretrained distribution to generate several medium-scale synthetic datasets (token scale: 2.1B-3.9B) and experiment with different settings of Transformer architectures (parameter scale: 4M-234M). Our experimental results further offer insights into how the data and model architectures influence ICG. | [
"Jiang, Zhongtao",
"Zhang, Yuanzhe",
"Luo, Kun",
"Yuan, Xiaowei",
"Zhao, Jun",
"Liu, Kang"
] | On the In-context Generation of Language Models | emnlp-main.568 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.569.bib | https://aclanthology.org/2024.emnlp-main.569/ | @inproceedings{stacey-etal-2024-atomic,
title = "Atomic Inference for {NLI} with Generated Facts as Atoms",
author = "Stacey, Joe and
Minervini, Pasquale and
Dubossarsky, Haim and
Camburu, Oana-Maria and
Rei, Marek",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.569",
pages = "10188--10204",
abstract = "With recent advances, neural models can achieve human-level performance on various natural language tasks. However, there are no guarantees that any explanations from these models are faithful, i.e. that they reflect the inner workings of the model. Atomic inference overcomes this issue, providing interpretable and faithful model decisions. This approach involves making predictions for different components (or atoms) of an instance, before using interpretable and deterministic rules to derive the overall prediction based on the individual atom-level predictions. We investigate the effectiveness of using LLM-generated facts as atoms, decomposing Natural Language Inference premises into lists of facts. While directly using generated facts in atomic inference systems can result in worse performance, with 1) a multi-stage fact generation process, and 2) a training regime that incorporates the facts, our fact-based method outperforms other approaches.",
}
| With recent advances, neural models can achieve human-level performance on various natural language tasks. However, there are no guarantees that any explanations from these models are faithful, i.e. that they reflect the inner workings of the model. Atomic inference overcomes this issue, providing interpretable and faithful model decisions. This approach involves making predictions for different components (or atoms) of an instance, before using interpretable and deterministic rules to derive the overall prediction based on the individual atom-level predictions. We investigate the effectiveness of using LLM-generated facts as atoms, decomposing Natural Language Inference premises into lists of facts. While directly using generated facts in atomic inference systems can result in worse performance, with 1) a multi-stage fact generation process, and 2) a training regime that incorporates the facts, our fact-based method outperforms other approaches. | [
"Stacey, Joe",
"Minervini, Pasquale",
"Dubossarsky, Haim",
"Camburu, Oana-Maria",
"Rei, Marek"
] | Atomic Inference for NLI with Generated Facts as Atoms | emnlp-main.569 | Poster | 2305.13214 | [
"https://github.com/joestacey/atomic_inference_anli"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.570.bib | https://aclanthology.org/2024.emnlp-main.570/ | @inproceedings{chen-etal-2024-towards-robust,
title = "Towards Robust Speech Representation Learning for Thousands of Languages",
author = "Chen, William and
Zhang, Wangyou and
Peng, Yifan and
Li, Xinjian and
Tian, Jinchuan and
Shi, Jiatong and
Chang, Xuankai and
Maiti, Soumi and
Livescu, Karen and
Watanabe, Shinji",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.570",
pages = "10205--10224",
abstract = "Self-supervised learning (SSL) has helped extend speech technologies to more languages by reducing the need for labeled data. However, models are still far from supporting the world{'}s 7000+ languages. We propose XEUS, a Cross-lingual Encoder for Universal Speech, trained on over 1 million hours of data across 4057 languages, extending the language coverage of SSL models 4-fold. We combine 1 million hours of speech from existing publicly accessible corpora with a newly created corpus of 7400+ hours from 4057 languages, which will be publicly released. To handle the diverse conditions of multilingual speech data, we augment the typical SSL masked prediction approach with a novel dereverberation objective, increasing robustness. We evaluate XEUS on several benchmarks, and show that it consistently outperforms or achieves comparable results to state-of-the-art (SOTA) SSL models across a variety of tasks. XEUS sets a new SOTA on the ML-SUPERB benchmark: it outperforms MMS 1B and w2v-BERT 2.0 v2 by 0.8{\%} and 4.4{\%} respectively, despite having less parameters or pre-training data. Checkpoints, code, and data are found in https://www.wavlab.org/activities/2024/xeus/.",
}
| Self-supervised learning (SSL) has helped extend speech technologies to more languages by reducing the need for labeled data. However, models are still far from supporting the world{'}s 7000+ languages. We propose XEUS, a Cross-lingual Encoder for Universal Speech, trained on over 1 million hours of data across 4057 languages, extending the language coverage of SSL models 4-fold. We combine 1 million hours of speech from existing publicly accessible corpora with a newly created corpus of 7400+ hours from 4057 languages, which will be publicly released. To handle the diverse conditions of multilingual speech data, we augment the typical SSL masked prediction approach with a novel dereverberation objective, increasing robustness. We evaluate XEUS on several benchmarks, and show that it consistently outperforms or achieves comparable results to state-of-the-art (SOTA) SSL models across a variety of tasks. XEUS sets a new SOTA on the ML-SUPERB benchmark: it outperforms MMS 1B and w2v-BERT 2.0 v2 by 0.8{\%} and 4.4{\%} respectively, despite having less parameters or pre-training data. Checkpoints, code, and data are found in https://www.wavlab.org/activities/2024/xeus/. | [
"Chen, William",
"Zhang, Wangyou",
"Peng, Yifan",
"Li, Xinjian",
"Tian, Jinchuan",
"Shi, Jiatong",
"Chang, Xuankai",
"Maiti, Soumi",
"Livescu, Karen",
"Watanabe, Shinji"
] | Towards Robust Speech Representation Learning for Thousands of Languages | emnlp-main.570 | Poster | 2407.00837 | [
""
] | https://huggingface.co/papers/2407.00837 | 9 | 10 | 1 | 10 | [
"espnet/xeus"
] | [
"espnet/mms_ulab_v2",
"espnet/wikitongues",
"espnet/jesus_dramas"
] | [] | [
"espnet/xeus"
] | [
"espnet/mms_ulab_v2",
"espnet/wikitongues",
"espnet/jesus_dramas"
] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.571.bib | https://aclanthology.org/2024.emnlp-main.571/ | @inproceedings{ren-etal-2024-learn,
title = "{I} Learn Better If You Speak My Language: Understanding the Superior Performance of Fine-Tuning Large Language Models with {LLM}-Generated Responses",
author = "Ren, Xuan and
Wu, Biao and
Liu, Lingqiao",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.571",
pages = "10225--10245",
abstract = "This paper explores an intriguing observation: fine-tuning a large language model (LLM) with responses generated by a LLM often yields better results than using responses generated by humans, particularly in reasoning tasks. We conduct an in-depth investigation to understand why this occurs. Contrary to the common belief that these instances is due to the more detailed nature of LLM-generated content, our study identifies another contributing factor: an LLM is inherently more {``}familiar{''} with LLM generated responses. This familiarity is evidenced by lower perplexity before fine-tuning. We design a series of experiments to understand the impact of the {``}familiarity{''} and our conclusion reveals that this {``}familiarity{''} significantly impacts learning performance. Training with LLM-generated responses not only enhances performance but also helps maintain the model{'}s capabilities in other reasoning tasks after fine-tuning on a specific task.",
}
| This paper explores an intriguing observation: fine-tuning a large language model (LLM) with responses generated by a LLM often yields better results than using responses generated by humans, particularly in reasoning tasks. We conduct an in-depth investigation to understand why this occurs. Contrary to the common belief that these instances is due to the more detailed nature of LLM-generated content, our study identifies another contributing factor: an LLM is inherently more {``}familiar{''} with LLM generated responses. This familiarity is evidenced by lower perplexity before fine-tuning. We design a series of experiments to understand the impact of the {``}familiarity{''} and our conclusion reveals that this {``}familiarity{''} significantly impacts learning performance. Training with LLM-generated responses not only enhances performance but also helps maintain the model{'}s capabilities in other reasoning tasks after fine-tuning on a specific task. | [
"Ren, Xuan",
"Wu, Biao",
"Liu, Lingqiao"
] | I Learn Better If You Speak My Language: Understanding the Superior Performance of Fine-Tuning Large Language Models with LLM-Generated Responses | emnlp-main.571 | Poster | 2402.11192 | [
"https://github.com/xuanren4470/i-learn-better-if-you-speak-my-language"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.572.bib | https://aclanthology.org/2024.emnlp-main.572/ | @inproceedings{li-etal-2024-prealign,
title = "{P}re{A}lign: Boosting Cross-Lingual Transfer by Early Establishment of Multilingual Alignment",
author = "Li, Jiahuan and
Huang, Shujian and
Ching, Aarron and
Dai, Xinyu and
Chen, Jiajun",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.572",
pages = "10246--10257",
abstract = "Large language models demonstrate reasonable multilingual abilities, despite predominantly English-centric pretraining. However, the spontaneous multilingual alignment in these models is shown to be weak, leading to unsatisfactory cross-lingual transfer and knowledge sharing. Previous works attempt to address this issue by explicitly injecting multilingual alignment information during or after pretraining. Thus for the early stage in pretraining, the alignment is weak for sharing information or knowledge across languages. In this paper, we propose PreAlign, a framework that establishes multilingual alignment prior to language model pretraining. PreAlign injects multilingual alignment by initializing the model to generate similar representations of aligned words and preserves this alignment using a code-switching strategy during pretraining. Extensive experiments in a synthetic English to English-Clone setting demonstrate that PreAlign significantly outperforms standard multilingual joint training in language modeling, zero-shot cross-lingual transfer, and cross-lingual knowledge application. Further experiments in real-world scenarios further validate PreAlign{'}s effectiveness across various model sizes.",
}
| Large language models demonstrate reasonable multilingual abilities, despite predominantly English-centric pretraining. However, the spontaneous multilingual alignment in these models is shown to be weak, leading to unsatisfactory cross-lingual transfer and knowledge sharing. Previous works attempt to address this issue by explicitly injecting multilingual alignment information during or after pretraining. Thus for the early stage in pretraining, the alignment is weak for sharing information or knowledge across languages. In this paper, we propose PreAlign, a framework that establishes multilingual alignment prior to language model pretraining. PreAlign injects multilingual alignment by initializing the model to generate similar representations of aligned words and preserves this alignment using a code-switching strategy during pretraining. Extensive experiments in a synthetic English to English-Clone setting demonstrate that PreAlign significantly outperforms standard multilingual joint training in language modeling, zero-shot cross-lingual transfer, and cross-lingual knowledge application. Further experiments in real-world scenarios further validate PreAlign{'}s effectiveness across various model sizes. | [
"Li, Jiahuan",
"Huang, Shujian",
"Ching, Aarron",
"Dai, Xinyu",
"Chen, Jiajun"
] | PreAlign: Boosting Cross-Lingual Transfer by Early Establishment of Multilingual Alignment | emnlp-main.572 | Poster | 2407.16222 | [
"https://github.com/saltychtao/prealign"
] | https://huggingface.co/papers/2407.16222 | 0 | 0 | 0 | 5 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.573.bib | https://aclanthology.org/2024.emnlp-main.573/ | @inproceedings{khanuja-etal-2024-image,
title = "An image speaks a thousand words, but can everyone listen? On image transcreation for cultural relevance",
author = "Khanuja, Simran and
Ramamoorthy, Sathyanarayanan and
Song, Yueqi and
Neubig, Graham",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.573",
pages = "10258--10279",
abstract = "Given the rise of multimedia content, human translators increasingly focus on culturally adapting not only words but also other modalities such as images to convey the same meaning. While several applications stand to benefit from this, machine translation systems remain confined to dealing with language in speech and text. In this work, we introduce a new task of translating images to make them culturally relevant. First, we build three pipelines comprising state-of-the-art generative models to do the task. Next, we build a two-part evaluation dataset {--} (i) concept: comprising 600 images that are cross-culturally coherent, focusing on a single concept per image; and (ii) application: comprising 100 images curated from real-world applications. We conduct a multi-faceted human evaluation of translated images to assess for cultural relevance and meaning preservation. We find that as of today, image-editing models fail at this task, but can be improved by leveraging LLMs and retrievers in the loop. Best pipelines can only translate 5{\%} of images for some countries in the easier concept dataset and no translation is successful for some countries in the application dataset, highlighting the challenging nature of the task. Our project webpage is here: https://machine-transcreation.github.io/image-transcreation and our code, data and model outputs can be found here: https://github.com/simran-khanuja/image-transcreation.",
}
| Given the rise of multimedia content, human translators increasingly focus on culturally adapting not only words but also other modalities such as images to convey the same meaning. While several applications stand to benefit from this, machine translation systems remain confined to dealing with language in speech and text. In this work, we introduce a new task of translating images to make them culturally relevant. First, we build three pipelines comprising state-of-the-art generative models to do the task. Next, we build a two-part evaluation dataset {--} (i) concept: comprising 600 images that are cross-culturally coherent, focusing on a single concept per image; and (ii) application: comprising 100 images curated from real-world applications. We conduct a multi-faceted human evaluation of translated images to assess for cultural relevance and meaning preservation. We find that as of today, image-editing models fail at this task, but can be improved by leveraging LLMs and retrievers in the loop. Best pipelines can only translate 5{\%} of images for some countries in the easier concept dataset and no translation is successful for some countries in the application dataset, highlighting the challenging nature of the task. Our project webpage is here: https://machine-transcreation.github.io/image-transcreation and our code, data and model outputs can be found here: https://github.com/simran-khanuja/image-transcreation. | [
"Khanuja, Simran",
"Ramamoorthy, Sathyanarayanan",
"Song, Yueqi",
"Neubig, Graham"
] | An image speaks a thousand words, but can everyone listen? On image transcreation for cultural relevance | emnlp-main.573 | Poster | 2404.01247 | [
"https://github.com/simran-khanuja/image-transcreation"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.574.bib | https://aclanthology.org/2024.emnlp-main.574/ | @inproceedings{chang-etal-2024-parts,
title = "When Parts Are Greater Than Sums: Individual {LLM} Components Can Outperform Full Models",
author = "Chang, Ting-Yun and
Thomason, Jesse and
Jia, Robin",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.574",
pages = "10280--10299",
abstract = "This paper studies in-context learning by decomposing the output of large language models into the individual contributions of attention heads and MLPs (components). We observe curious components: good-performing ones that individually do well on a classification task, even when the model performs poorly; bad-performing ones that do much worse than chance; and label-biased components that always predict the same label. We find that component accuracies are well-correlated across different demonstration sets and perturbations of prompt templates. Based on our findings, we propose component reweighting, which learns to linearly re-scale the component activations from a few labeled examples. Given 24 labeled examples, our method improves by an average of 6.0{\%} accuracy points over 24-shot ICL across 8 tasks on Llama-2-7B. Overall, this paper both enriches our understanding of ICL and provides a practical method for improvement by examining model internals.",
}
| This paper studies in-context learning by decomposing the output of large language models into the individual contributions of attention heads and MLPs (components). We observe curious components: good-performing ones that individually do well on a classification task, even when the model performs poorly; bad-performing ones that do much worse than chance; and label-biased components that always predict the same label. We find that component accuracies are well-correlated across different demonstration sets and perturbations of prompt templates. Based on our findings, we propose component reweighting, which learns to linearly re-scale the component activations from a few labeled examples. Given 24 labeled examples, our method improves by an average of 6.0{\%} accuracy points over 24-shot ICL across 8 tasks on Llama-2-7B. Overall, this paper both enriches our understanding of ICL and provides a practical method for improvement by examining model internals. | [
"Chang, Ting-Yun",
"Thomason, Jesse",
"Jia, Robin"
] | When Parts Are Greater Than Sums: Individual LLM Components Can Outperform Full Models | emnlp-main.574 | Poster | 2406.13131 | [
"https://github.com/terarachang/LLMDecomp"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.575.bib | https://aclanthology.org/2024.emnlp-main.575/ | @inproceedings{yu-etal-2024-multimodal,
title = "Multimodal Clickbait Detection by De-confounding Biases Using Causal Representation Inference",
author = "Yu, Jianxing and
Wang, Shiqi and
Yin, Han and
Sun, Zhenlong and
Xie, Ruobing and
Zhang, Bo and
Rao, Yanghui",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.575",
pages = "10300--10317",
abstract = "This paper focuses on detecting clickbait posts on the Web. These posts often use eye-catching disinformation in mixed modalities to mislead users to click for profit. That affects the user experience and thus would be blocked by content provider. To escape detection, malicious creators use tricks to add some irrelevant non-bait content into bait posts, dressing them up as legal to fool the detector. This content often has biased relations with non-bait labels, yet traditional detectors tend to make predictions based on simple co-occurrence rather than grasping inherent factors that lead to malicious behavior. This spurious bias would easily cause misjudgments. To address this problem, we propose a new debiased method based on causal inference. We first employ a set of features in multiple modalities to characterize the posts. Considering these features are often mixed up with unknown biases, we then disentangle three kinds of latent factors from them, including the invariant factor that indicates intrinsic bait intention; the causal factor which reflects deceptive patterns in a certain scenario, and non-causal noise. By eliminating the noise that causes bias, we can use invariant and causal factors to build a robust model with good generalization ability. Experiments on three popular datasets show the effectiveness of our approach.",
}
| This paper focuses on detecting clickbait posts on the Web. These posts often use eye-catching disinformation in mixed modalities to mislead users to click for profit. That affects the user experience and thus would be blocked by content provider. To escape detection, malicious creators use tricks to add some irrelevant non-bait content into bait posts, dressing them up as legal to fool the detector. This content often has biased relations with non-bait labels, yet traditional detectors tend to make predictions based on simple co-occurrence rather than grasping inherent factors that lead to malicious behavior. This spurious bias would easily cause misjudgments. To address this problem, we propose a new debiased method based on causal inference. We first employ a set of features in multiple modalities to characterize the posts. Considering these features are often mixed up with unknown biases, we then disentangle three kinds of latent factors from them, including the invariant factor that indicates intrinsic bait intention; the causal factor which reflects deceptive patterns in a certain scenario, and non-causal noise. By eliminating the noise that causes bias, we can use invariant and causal factors to build a robust model with good generalization ability. Experiments on three popular datasets show the effectiveness of our approach. | [
"Yu, Jianxing",
"Wang, Shiqi",
"Yin, Han",
"Sun, Zhenlong",
"Xie, Ruobing",
"Zhang, Bo",
"Rao, Yanghui"
] | Multimodal Clickbait Detection by De-confounding Biases Using Causal Representation Inference | emnlp-main.575 | Poster | 2410.07673 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.576.bib | https://aclanthology.org/2024.emnlp-main.576/ | @inproceedings{yoon-etal-2024-matryoshka,
title = "Matryoshka-Adaptor: Unsupervised and Supervised Tuning for Smaller Embedding Dimensions",
author = "Yoon, Jinsung and
Sinha, Rajarishi and
Arik, Sercan O and
Pfister, Tomas",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.576",
pages = "10318--10336",
abstract = "Embeddings from Large Language Models (LLMs) have emerged as critical components in various applications, particularly for information retrieval. While high-dimensional embeddings generally demonstrate superior performance as they contain more salient information, their practical application is frequently hindered by elevated computational latency and the associated higher cost. To address these challenges, we propose Matryoshka-Adaptor, a novel tuning framework designed for the customization of LLM embeddings. Matryoshka-Adaptor facilitates substantial dimensionality reduction while maintaining comparable performance levels, thereby achieving a significant enhancement in computational efficiency and cost-effectiveness. Our framework directly modifies the embeddings from pre-trained LLMs which is designed to be seamlessly integrated with any LLM architecture, encompassing those accessible exclusively through black-box APIs. Also, it exhibits efficacy in both unsupervised and supervised learning settings. A rigorous evaluation conducted across a diverse corpus of English, multilingual, and multimodal datasets consistently reveals substantial gains with Matryoshka-Adaptor. Notably, with Google and OpenAI Embedding APIs, Matryoshka-Adaptor achieves a reduction in dimensionality ranging from two- to twelve-fold without compromising performance across multiple BEIR datasets.",
}
| Embeddings from Large Language Models (LLMs) have emerged as critical components in various applications, particularly for information retrieval. While high-dimensional embeddings generally demonstrate superior performance as they contain more salient information, their practical application is frequently hindered by elevated computational latency and the associated higher cost. To address these challenges, we propose Matryoshka-Adaptor, a novel tuning framework designed for the customization of LLM embeddings. Matryoshka-Adaptor facilitates substantial dimensionality reduction while maintaining comparable performance levels, thereby achieving a significant enhancement in computational efficiency and cost-effectiveness. Our framework directly modifies the embeddings from pre-trained LLMs which is designed to be seamlessly integrated with any LLM architecture, encompassing those accessible exclusively through black-box APIs. Also, it exhibits efficacy in both unsupervised and supervised learning settings. A rigorous evaluation conducted across a diverse corpus of English, multilingual, and multimodal datasets consistently reveals substantial gains with Matryoshka-Adaptor. Notably, with Google and OpenAI Embedding APIs, Matryoshka-Adaptor achieves a reduction in dimensionality ranging from two- to twelve-fold without compromising performance across multiple BEIR datasets. | [
"Yoon, Jinsung",
"Sinha, Rajarishi",
"Arik, Sercan O",
"Pfister, Tomas"
] | Matryoshka-Adaptor: Unsupervised and Supervised Tuning for Smaller Embedding Dimensions | emnlp-main.576 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.577.bib | https://aclanthology.org/2024.emnlp-main.577/ | @inproceedings{kou-etal-2024-knn,
title = "{KNN}-Instruct: Automatic Instruction Construction with K Nearest Neighbor Deduction",
author = "Kou, Jianshang and
Xu, Benfeng and
Zhu, Chiwei and
Mao, Zhendong",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.577",
pages = "10337--10350",
abstract = "Supervised fine-tuning (SFT) is a critical procedure for aligning large language models. Despite its efficiency, the construction of SFT data often struggles with issues of quality, diversity, and scalability. Many existing methods, inspired by the Self-Instruct framework, typically generate synthetic instructions by prompting aligned proprietary models like ChatGPT. However, such process suffers from stale distribution, resulting in instructions that are merely trivial variations of existing ones. In this paper, we introduce a novel bootstrapping approach termed KNN-Instruct, which incorporates KNN deduction to produce meaningful new instructions by effectively summarizing and learning from similar existing ones. We conduct an economical controlled experiment to preliminarily validate its effectiveness. In the further experiment, we construct a high-quality SFT dataset named KNN-Inst-12k*. Applying the dataset to Qwen-2-7B, we get a MT-Bench score of 7.64, which outperforms all 7B models on the LMSYS leaderboard, including Starling-LM-7B (7.48), OpenChat-3.5 (7.06) and Zephyr-7B-beta (6.53). Our code and data are available at https://github.com/CrossmodalGroup/KNN-Instruct/.",
}
| Supervised fine-tuning (SFT) is a critical procedure for aligning large language models. Despite its efficiency, the construction of SFT data often struggles with issues of quality, diversity, and scalability. Many existing methods, inspired by the Self-Instruct framework, typically generate synthetic instructions by prompting aligned proprietary models like ChatGPT. However, such process suffers from stale distribution, resulting in instructions that are merely trivial variations of existing ones. In this paper, we introduce a novel bootstrapping approach termed KNN-Instruct, which incorporates KNN deduction to produce meaningful new instructions by effectively summarizing and learning from similar existing ones. We conduct an economical controlled experiment to preliminarily validate its effectiveness. In the further experiment, we construct a high-quality SFT dataset named KNN-Inst-12k*. Applying the dataset to Qwen-2-7B, we get a MT-Bench score of 7.64, which outperforms all 7B models on the LMSYS leaderboard, including Starling-LM-7B (7.48), OpenChat-3.5 (7.06) and Zephyr-7B-beta (6.53). Our code and data are available at https://github.com/CrossmodalGroup/KNN-Instruct/. | [
"Kou, Jianshang",
"Xu, Benfeng",
"Zhu, Chiwei",
"Mao, Zhendong"
] | KNN-Instruct: Automatic Instruction Construction with K Nearest Neighbor Deduction | emnlp-main.577 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.578.bib | https://aclanthology.org/2024.emnlp-main.578/ | @inproceedings{lin-etal-2024-contextualized,
title = "Contextualized Sequence Likelihood: Enhanced Confidence Scores for Natural Language Generation",
author = "Lin, Zhen and
Trivedi, Shubhendu and
Sun, Jimeng",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.578",
pages = "10351--10368",
abstract = "The advent of large language models (LLMs) has dramatically advanced the state-of-the-art in numerous natural language generation tasks. For LLMs to be applied reliably, it is essential to have an accurate measure of their confidence. Currently, the most commonly used confidence score function is the likelihood of the generated sequence, which, however, conflates semantic and syntactic components. For instance, in question-answering (QA) tasks, an awkward phrasing of the correct answer might result in a lower probability prediction. Additionally, different tokens should be weighted differently depending on the context. In this work, we propose enhancing the predicted sequence probability by assigning different weights to various tokens using attention values elicited from the base LLM. By employing a validation set, we can identify the relevant attention heads, thereby significantly improving the reliability of the vanilla sequence probability confidence measure. We refer to this new score as the Contextualized Sequence Likelihood (CSL). CSL is easy to implement, fast to compute, and offers considerable potential for further improvement with task-specific prompts. Across several QA datasets and a diverse array of LLMs, CSL has demonstrated significantly higher reliability than state-of-the-art baselines in predicting generation quality, as measured by the AUROC or AUARC.",
}
| The advent of large language models (LLMs) has dramatically advanced the state-of-the-art in numerous natural language generation tasks. For LLMs to be applied reliably, it is essential to have an accurate measure of their confidence. Currently, the most commonly used confidence score function is the likelihood of the generated sequence, which, however, conflates semantic and syntactic components. For instance, in question-answering (QA) tasks, an awkward phrasing of the correct answer might result in a lower probability prediction. Additionally, different tokens should be weighted differently depending on the context. In this work, we propose enhancing the predicted sequence probability by assigning different weights to various tokens using attention values elicited from the base LLM. By employing a validation set, we can identify the relevant attention heads, thereby significantly improving the reliability of the vanilla sequence probability confidence measure. We refer to this new score as the Contextualized Sequence Likelihood (CSL). CSL is easy to implement, fast to compute, and offers considerable potential for further improvement with task-specific prompts. Across several QA datasets and a diverse array of LLMs, CSL has demonstrated significantly higher reliability than state-of-the-art baselines in predicting generation quality, as measured by the AUROC or AUARC. | [
"Lin, Zhen",
"Trivedi, Shubhendu",
"Sun, Jimeng"
] | Contextualized Sequence Likelihood: Enhanced Confidence Scores for Natural Language Generation | emnlp-main.578 | Poster | 2406.01806 | [
"https://github.com/zlin7/contextsl"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.579.bib | https://aclanthology.org/2024.emnlp-main.579/ | @inproceedings{cai-etal-2024-mixgr,
title = "{M}ix{GR}: Enhancing Retriever Generalization for Scientific Domain through Complementary Granularity",
author = "Cai, Fengyu and
Zhao, Xinran and
Chen, Tong and
Chen, Sihao and
Zhang, Hongming and
Gurevych, Iryna and
Koeppl, Heinz",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.579",
pages = "10369--10391",
}
| No abstract found | [
"Cai, Fengyu",
"Zhao, Xinran",
"Chen, Tong",
"Chen, Sihao",
"Zhang, Hongming",
"Gurevych, Iryna",
"Koeppl, Heinz"
] | MixGR: Enhancing Retriever Generalization for Scientific Domain through Complementary Granularity | emnlp-main.579 | Poster | [
"https://github.com/TRUMANCFY/MixGR"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.580.bib | https://aclanthology.org/2024.emnlp-main.580/ | @inproceedings{nguyen-etal-2024-carer,
title = "{CARER} - {C}linic{A}l Reasoning-Enhanced Representation for Temporal Health Risk Prediction",
author = "Nguyen, Tuan Dung and
Huynh, Thanh Trung and
Phan, Minh Hieu and
Nguyen, Quoc Viet Hung and
Nguyen, Phi Le",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.580",
pages = "10392--10407",
abstract = "The increasing availability of multimodal data from electronic health records (EHR) has paved the way for deep learning methods to improve diagnosis accuracy. However, deep learning models are data-driven, requiring large-scale datasets to achieve high generalizability. Inspired by how human experts leverage reasoning for medical diagnosis, we propose CARER, a novel health risk prediction framework, that enhances deep learning models with clinical rationales derived from medically proficient Large Language Models (LLMs). In addition, we provide a cross-view alignment loss which aligns the {``}local{''} view from the patient{'}s health status with the {``}global{''} view from the external LLM{'}s clinical reasoning to boost the mutual feature learning. Through extensive experiments on two predictive tasks using two popular EHR datasets, our CARER{'}s significantly exceeds the performance of state-of-the-art models by up to 11.2{\%}, especially in improving data efficiency and generalizability. Our code is available at https://github.com/tuandung2812/CARER-EMNLP-2024",
}
| The increasing availability of multimodal data from electronic health records (EHR) has paved the way for deep learning methods to improve diagnosis accuracy. However, deep learning models are data-driven, requiring large-scale datasets to achieve high generalizability. Inspired by how human experts leverage reasoning for medical diagnosis, we propose CARER, a novel health risk prediction framework, that enhances deep learning models with clinical rationales derived from medically proficient Large Language Models (LLMs). In addition, we provide a cross-view alignment loss which aligns the {``}local{''} view from the patient{'}s health status with the {``}global{''} view from the external LLM{'}s clinical reasoning to boost the mutual feature learning. Through extensive experiments on two predictive tasks using two popular EHR datasets, our CARER{'}s significantly exceeds the performance of state-of-the-art models by up to 11.2{\%}, especially in improving data efficiency and generalizability. Our code is available at https://github.com/tuandung2812/CARER-EMNLP-2024 | [
"Nguyen, Tuan Dung",
"Huynh, Thanh Trung",
"Phan, Minh Hieu",
"Nguyen, Quoc Viet Hung",
"Nguyen, Phi Le"
] | CARER - ClinicAl Reasoning-Enhanced Representation for Temporal Health Risk Prediction | emnlp-main.580 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.581.bib | https://aclanthology.org/2024.emnlp-main.581/ | @inproceedings{cheng-etal-2024-dialogues,
title = "{``}In-Dialogues We Learn{''}: Towards Personalized Dialogue Without Pre-defined Profiles through In-Dialogue Learning",
author = "Cheng, Chuanqi and
Tu, Quan and
Wu, Wei and
Shang, Shuo and
Mao, Cunli and
Yu, Zhengtao and
Yan, Rui",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.581",
pages = "10408--10422",
abstract = "Personalized dialogue systems have gained significant attention in recent years for their ability to generate responses in alignment with different personas. However, most existing approaches rely on pre-defined personal profiles, which are not only time-consuming and labor-intensive to create but also lack flexibility. We propose In-Dialogue Learning (IDL), a fine-tuning framework that enhances the ability of pre-trained large language models to leverage dialogue history to characterize persona for personalized dialogue generation tasks without pre-defined profiles. Our experiments on three datasets demonstrate that IDL brings substantial improvements, with BLEU and ROUGE scores increasing by up to 200{\%} and 247{\%}, respectively. Additionally, the results of human evaluations further validate the efficacy of our proposed method.",
}
| Personalized dialogue systems have gained significant attention in recent years for their ability to generate responses in alignment with different personas. However, most existing approaches rely on pre-defined personal profiles, which are not only time-consuming and labor-intensive to create but also lack flexibility. We propose In-Dialogue Learning (IDL), a fine-tuning framework that enhances the ability of pre-trained large language models to leverage dialogue history to characterize persona for personalized dialogue generation tasks without pre-defined profiles. Our experiments on three datasets demonstrate that IDL brings substantial improvements, with BLEU and ROUGE scores increasing by up to 200{\%} and 247{\%}, respectively. Additionally, the results of human evaluations further validate the efficacy of our proposed method. | [
"Cheng, Chuanqi",
"Tu, Quan",
"Wu, Wei",
"Shang, Shuo",
"Mao, Cunli",
"Yu, Zhengtao",
"Yan, Rui"
] | “In-Dialogues We Learn”: Towards Personalized Dialogue Without Pre-defined Profiles through In-Dialogue Learning | emnlp-main.581 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.582.bib | https://aclanthology.org/2024.emnlp-main.582/ | @inproceedings{yan-etal-2024-encourage,
title = "Encourage or Inhibit Monosemanticity? Revisit Monosemanticity from a Feature Decorrelation Perspective",
author = "Yan, Hanqi and
Xiang, Yanzheng and
Chen, Guangyi and
Wang, Yifei and
Gui, Lin and
He, Yulan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.582",
pages = "10423--10435",
abstract = "To better interpret the intrinsic mechanism of large language models (LLMs), recent studies focus on monosemanticity on its basic units. A monosemantic neuron is dedicated to a single and specific concept, which forms a one-to-one correlation between neurons and concepts. Despite extensive research in monosemanticity probing, it remains unclear whether monosemanticity is beneficial or harmful to model capacity. To explore this question, we revisit monosemanticity from the feature decorrelation perspective and advocate for its encouragement. We experimentally observe that the current conclusion by (CITATION), which suggests that decreasing monosemanticity enhances model performance, does not hold when the model changes. Instead, we demonstrate that monosemanticity consistently exhibits a positive correlation with model capacity, in the preference alignment process. Consequently, we apply feature correlation as a proxy for monosemanticity and incorporate a feature decorrelation regularizer into the dynamic preference optimization process. The experiments show that our method not only enhances representation diversity and activation sparsity but also improves preference alignment performance.",
}
| To better interpret the intrinsic mechanism of large language models (LLMs), recent studies focus on monosemanticity on its basic units. A monosemantic neuron is dedicated to a single and specific concept, which forms a one-to-one correlation between neurons and concepts. Despite extensive research in monosemanticity probing, it remains unclear whether monosemanticity is beneficial or harmful to model capacity. To explore this question, we revisit monosemanticity from the feature decorrelation perspective and advocate for its encouragement. We experimentally observe that the current conclusion by (CITATION), which suggests that decreasing monosemanticity enhances model performance, does not hold when the model changes. Instead, we demonstrate that monosemanticity consistently exhibits a positive correlation with model capacity, in the preference alignment process. Consequently, we apply feature correlation as a proxy for monosemanticity and incorporate a feature decorrelation regularizer into the dynamic preference optimization process. The experiments show that our method not only enhances representation diversity and activation sparsity but also improves preference alignment performance. | [
"Yan, Hanqi",
"Xiang, Yanzheng",
"Chen, Guangyi",
"Wang, Yifei",
"Gui, Lin",
"He, Yulan"
] | Encourage or Inhibit Monosemanticity? Revisit Monosemanticity from a Feature Decorrelation Perspective | emnlp-main.582 | Poster | 2406.17969 | [
"https://github.com/hanqi-qi/revisit_monosemanticity"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.583.bib | https://aclanthology.org/2024.emnlp-main.583/ | @inproceedings{liu-etal-2024-enhancing-language,
title = "Enhancing Language Model Factuality via Activation-Based Confidence Calibration and Guided Decoding",
author = "Liu, Xin and
Fatahi Bayat, Farima and
Wang, Lu",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.583",
pages = "10436--10448",
abstract = "Calibrating language models (LMs) aligns their generation confidence with the actual likelihood of answer correctness, which can inform users about LMs{'} reliability and mitigate hallucinated content. However, prior calibration methods, such as self-consistency-based and logit-based approaches, are either limited in inference-time efficiency or fall short of providing informative signals. Moreover, simply filtering out low-confidence responses reduces the LM{'}s helpfulness when the answers are correct. Therefore, effectively using calibration techniques to enhance an LM{'}s factuality remains an unsolved challenge. In this paper, we first propose an activation-based calibration method, ActCab, which trains a linear layer on top of the LM{'}s last-layer activations that can better capture the representations of knowledge. Built on top of ActCab, we further propose CoDec, a confidence-guided decoding strategy to elicit truthful answers with high confidence from LMs. By evaluating on five popular QA benchmarks, ActCab achieves superior calibration performance than all competitive baselines, e.g., by reducing the average expected calibration error (ECE) score by up to 39{\%}. Further experiments on CoDec show consistent improvements in several LMs{'} factuality on challenging QA datasets, such as TruthfulQA, highlighting the value of confidence signals in enhancing the factuality.",
}
| Calibrating language models (LMs) aligns their generation confidence with the actual likelihood of answer correctness, which can inform users about LMs{'} reliability and mitigate hallucinated content. However, prior calibration methods, such as self-consistency-based and logit-based approaches, are either limited in inference-time efficiency or fall short of providing informative signals. Moreover, simply filtering out low-confidence responses reduces the LM{'}s helpfulness when the answers are correct. Therefore, effectively using calibration techniques to enhance an LM{'}s factuality remains an unsolved challenge. In this paper, we first propose an activation-based calibration method, ActCab, which trains a linear layer on top of the LM{'}s last-layer activations that can better capture the representations of knowledge. Built on top of ActCab, we further propose CoDec, a confidence-guided decoding strategy to elicit truthful answers with high confidence from LMs. By evaluating on five popular QA benchmarks, ActCab achieves superior calibration performance than all competitive baselines, e.g., by reducing the average expected calibration error (ECE) score by up to 39{\%}. Further experiments on CoDec show consistent improvements in several LMs{'} factuality on challenging QA datasets, such as TruthfulQA, highlighting the value of confidence signals in enhancing the factuality. | [
"Liu, Xin",
"Fatahi Bayat, Farima",
"Wang, Lu"
] | Enhancing Language Model Factuality via Activation-Based Confidence Calibration and Guided Decoding | emnlp-main.583 | Poster | 2406.13230 | [
"https://github.com/launchnlp/actcab"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.584.bib | https://aclanthology.org/2024.emnlp-main.584/ | @inproceedings{gan-etal-2024-reasoning,
title = "Reasoning Robustness of {LLM}s to Adversarial Typographical Errors",
author = "Gan, Esther and
Zhao, Yiran and
Cheng, Liying and
Yancan, Mao and
Goyal, Anirudh and
Kawaguchi, Kenji and
Kan, Min-Yen and
Shieh, Michael",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.584",
pages = "10449--10459",
}
| No abstract found | [
"Gan, Esther",
"Zhao, Yiran",
"Cheng, Liying",
"Yancan, Mao",
"Goyal, Anirudh",
"Kawaguchi, Kenji",
"Kan, Min-Yen",
"Shieh, Michael"
] | Reasoning Robustness of LLMs to Adversarial Typographical Errors | emnlp-main.584 | Poster | 2411.05345 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.585.bib | https://aclanthology.org/2024.emnlp-main.585/ | @inproceedings{wang-etal-2024-inferaligner,
title = "{I}nfer{A}ligner: Inference-Time Alignment for Harmlessness through Cross-Model Guidance",
author = "Wang, Pengyu and
Zhang, Dong and
Li, Linyang and
Tan, Chenkun and
Wang, Xinghao and
Zhang, Mozhi and
Ren, Ke and
Jiang, Botian and
Qiu, Xipeng",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.585",
pages = "10460--10479",
abstract = "As large language models (LLMs) rapidly evolve, they are increasingly being customized through fine-tuning to suit the specific needs of various applications. A critical aspect of this advancement is the alignment process, which ensures that these models perform tasks in ways that align with human values and expectations. Current alignment methods, such as direct preference optimization (DPO) and reinforcement learning from human feedback (RLHF), focus primarily on alignment during training phase. However, these methods often involve complex and resource-intensive training processes, posing significant challenge for their implementation. Therefore, we propose \textbf{InferAligner}, a simple yet effective method for harmlessness alignment during inference phase. InferAligner decouples harmlessness from helpfulness. During the training phase, it focuses solely on enhancing the target model{'}s capabilities on downstream tasks. In the inference phase, it utilizes safety steering vectors extracted from the aligned model to guide the target model towards harmlessness alignment. Experimental results show that our method can be very effectively applied to domain-specific models in finance, medicine, and mathematics, as well as to multimodal large language models (MLLMs) such as LLaVA. It significantly diminishes the attack success rate (ASR) of both harmful instructions and jailbreak instructions, while maintaining almost unchanged performance in downstream tasks.",
}
| As large language models (LLMs) rapidly evolve, they are increasingly being customized through fine-tuning to suit the specific needs of various applications. A critical aspect of this advancement is the alignment process, which ensures that these models perform tasks in ways that align with human values and expectations. Current alignment methods, such as direct preference optimization (DPO) and reinforcement learning from human feedback (RLHF), focus primarily on alignment during training phase. However, these methods often involve complex and resource-intensive training processes, posing significant challenge for their implementation. Therefore, we propose \textbf{InferAligner}, a simple yet effective method for harmlessness alignment during inference phase. InferAligner decouples harmlessness from helpfulness. During the training phase, it focuses solely on enhancing the target model{'}s capabilities on downstream tasks. In the inference phase, it utilizes safety steering vectors extracted from the aligned model to guide the target model towards harmlessness alignment. Experimental results show that our method can be very effectively applied to domain-specific models in finance, medicine, and mathematics, as well as to multimodal large language models (MLLMs) such as LLaVA. It significantly diminishes the attack success rate (ASR) of both harmful instructions and jailbreak instructions, while maintaining almost unchanged performance in downstream tasks. | [
"Wang, Pengyu",
"Zhang, Dong",
"Li, Linyang",
"Tan, Chenkun",
"Wang, Xinghao",
"Zhang, Mozhi",
"Ren, Ke",
"Jiang, Botian",
"Qiu, Xipeng"
] | InferAligner: Inference-Time Alignment for Harmlessness through Cross-Model Guidance | emnlp-main.585 | Poster | 2401.11206 | [
"https://github.com/jihuai-wpy/inferaligner"
] | https://huggingface.co/papers/2401.11206 | 3 | 1 | 0 | 8 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.586.bib | https://aclanthology.org/2024.emnlp-main.586/ | @inproceedings{wilie-etal-2024-belief,
title = "Belief Revision: The Adaptability of Large Language Models Reasoning",
author = "Wilie, Bryan and
Cahyawijaya, Samuel and
Ishii, Etsuko and
He, Junxian and
Fung, Pascale",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.586",
pages = "10480--10496",
abstract = "The capability to reason from text is crucial for real-world NLP applications. Real-world scenarios often involve incomplete or evolving data. In response, individuals update their beliefs and understandings accordingly. However, most existing evaluations assume that language models (LMs) operate with consistent information. We introduce Belief-R, a new dataset designed to test LMs{'} belief revision ability when presented with new evidence. Inspired by how humans suppress prior inferences, this task assesses LMs within the newly proposed delta reasoning ($\Delta R$) framework. Belief-R features sequences of premises designed to simulate scenarios where additional information could necessitate prior conclusions drawn by LMs. We evaluate {\textasciitilde}30 LMs across diverse prompting strategies and found that LMs generally struggle to appropriately revise their beliefs in response to new information. Further, models adept at updating often underperformed in scenarios without necessary updates, highlighting a critical trade-off. These insights underscore the importance of improving LMs{'} adaptiveness to changing information, a step toward more reliable AI systems.",
}
| The capability to reason from text is crucial for real-world NLP applications. Real-world scenarios often involve incomplete or evolving data. In response, individuals update their beliefs and understandings accordingly. However, most existing evaluations assume that language models (LMs) operate with consistent information. We introduce Belief-R, a new dataset designed to test LMs{'} belief revision ability when presented with new evidence. Inspired by how humans suppress prior inferences, this task assesses LMs within the newly proposed delta reasoning ($\Delta R$) framework. Belief-R features sequences of premises designed to simulate scenarios where additional information could necessitate prior conclusions drawn by LMs. We evaluate {\textasciitilde}30 LMs across diverse prompting strategies and found that LMs generally struggle to appropriately revise their beliefs in response to new information. Further, models adept at updating often underperformed in scenarios without necessary updates, highlighting a critical trade-off. These insights underscore the importance of improving LMs{'} adaptiveness to changing information, a step toward more reliable AI systems. | [
"Wilie, Bryan",
"Cahyawijaya, Samuel",
"Ishii, Etsuko",
"He, Junxian",
"Fung, Pascale"
] | Belief Revision: The Adaptability of Large Language Models Reasoning | emnlp-main.586 | Poster | 2406.19764 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.587.bib | https://aclanthology.org/2024.emnlp-main.587/ | @inproceedings{liu-etal-2024-fisher,
title = "Fisher Information-based Efficient Curriculum Federated Learning with Large Language Models",
author = "Liu, Ji and
Ren, Jiaxiang and
Jin, Ruoming and
Zhang, Zijie and
Zhou, Yang and
Valduriez, Patrick and
Dou, Dejing",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.587",
pages = "10497--10523",
abstract = "As a promising paradigm to collaboratively train models with decentralized data, Federated Learning (FL) can be exploited to fine-tune Large Language Models (LLMs). While LLMs correspond to huge size, the scale of the training data significantly increases, which leads to tremendous amounts of computation and communication costs. The training data is generally non-Independent and Identically Distributed (non-IID), which requires adaptive data processing within each device. Although Low-Rank Adaptation (LoRA) can significantly reduce the scale of parameters to update in the fine-tuning process, it still takes unaffordable time to transfer the low-rank parameters of all the layers in LLMs. In this paper, we propose a Fisher Information-based Efficient Curriculum Federated Learning framework (FibecFed) with two novel methods, i.e., adaptive federated curriculum learning and efficient sparse parameter update. First, we propose a fisher information-based method to adaptively sample data within each device to improve the effectiveness of the FL fine-tuning process. Second, we dynamically select the proper layers for global aggregation and sparse parameters for local update with LoRA so as to improve the efficiency of the FL fine-tuning process. Extensive experimental results based on 10 datasets demonstrate that FibecFed yields excellent performance (up to 45.35{\%} in terms of accuracy) and superb fine-tuning speed (up to 98.61{\%} faster) compared with 17 baseline approaches).",
}
| As a promising paradigm to collaboratively train models with decentralized data, Federated Learning (FL) can be exploited to fine-tune Large Language Models (LLMs). While LLMs correspond to huge size, the scale of the training data significantly increases, which leads to tremendous amounts of computation and communication costs. The training data is generally non-Independent and Identically Distributed (non-IID), which requires adaptive data processing within each device. Although Low-Rank Adaptation (LoRA) can significantly reduce the scale of parameters to update in the fine-tuning process, it still takes unaffordable time to transfer the low-rank parameters of all the layers in LLMs. In this paper, we propose a Fisher Information-based Efficient Curriculum Federated Learning framework (FibecFed) with two novel methods, i.e., adaptive federated curriculum learning and efficient sparse parameter update. First, we propose a fisher information-based method to adaptively sample data within each device to improve the effectiveness of the FL fine-tuning process. Second, we dynamically select the proper layers for global aggregation and sparse parameters for local update with LoRA so as to improve the efficiency of the FL fine-tuning process. Extensive experimental results based on 10 datasets demonstrate that FibecFed yields excellent performance (up to 45.35{\%} in terms of accuracy) and superb fine-tuning speed (up to 98.61{\%} faster) compared with 17 baseline approaches). | [
"Liu, Ji",
"Ren, Jiaxiang",
"Jin, Ruoming",
"Zhang, Zijie",
"Zhou, Yang",
"Valduriez, Patrick",
"Dou, Dejing"
] | Fisher Information-based Efficient Curriculum Federated Learning with Large Language Models | emnlp-main.587 | Poster | 2410.00131 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.588.bib | https://aclanthology.org/2024.emnlp-main.588/ | @inproceedings{wang-etal-2024-bio,
title = "Bio-{RFX}: Refining Biomedical Extraction via Advanced Relation Classification and Structural Constraints",
author = "Wang, Minjia and
Liu, Fangzhou and
Li, Xiuxing and
Dong, Bowen and
Li, Zhenyu and
Pan, Tengyu and
Wang, Jianyong",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.588",
pages = "10524--10539",
abstract = "The ever-growing biomedical publications magnify the challenge of extracting structured data from unstructured texts. This task involves two components: biomedical entity identification (Named Entity Recognition, NER) and their interrelation determination (Relation Extraction, RE). However, existing methods often neglect unique features of the biomedical literature, such as ambiguous entities, nested proper nouns, and overlapping relation triplets, and underutilize prior knowledge, leading to an intolerable performance decline in the biomedical domain, especially with limited annotated training data. In this paper, we propose the Biomedical Relation-First eXtraction (Bio-RFX) model by leveraging sentence-level relation classification before entity extraction to tackle entity ambiguity. Moreover, we exploit structural constraints between entities and relations to guide the model{'}s hypothesis space, enhancing extraction performance across different training scenarios. Comprehensive experimental results on biomedical datasets show that Bio-RFX achieves significant improvements on both NER and RE tasks. Even under the low-resource training scenarios, it outperforms all baselines in NER and has highly competitive performance compared to the state-of-the-art fine-tuned baselines in RE.",
}
| The ever-growing biomedical publications magnify the challenge of extracting structured data from unstructured texts. This task involves two components: biomedical entity identification (Named Entity Recognition, NER) and their interrelation determination (Relation Extraction, RE). However, existing methods often neglect unique features of the biomedical literature, such as ambiguous entities, nested proper nouns, and overlapping relation triplets, and underutilize prior knowledge, leading to an intolerable performance decline in the biomedical domain, especially with limited annotated training data. In this paper, we propose the Biomedical Relation-First eXtraction (Bio-RFX) model by leveraging sentence-level relation classification before entity extraction to tackle entity ambiguity. Moreover, we exploit structural constraints between entities and relations to guide the model{'}s hypothesis space, enhancing extraction performance across different training scenarios. Comprehensive experimental results on biomedical datasets show that Bio-RFX achieves significant improvements on both NER and RE tasks. Even under the low-resource training scenarios, it outperforms all baselines in NER and has highly competitive performance compared to the state-of-the-art fine-tuned baselines in RE. | [
"Wang, Minjia",
"Liu, Fangzhou",
"Li, Xiuxing",
"Dong, Bowen",
"Li, Zhenyu",
"Pan, Tengyu",
"Wang, Jianyong"
] | Bio-RFX: Refining Biomedical Extraction via Advanced Relation Classification and Structural Constraints | emnlp-main.588 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.589.bib | https://aclanthology.org/2024.emnlp-main.589/ | @inproceedings{bao-etal-2024-decoding,
title = "Decoding Matters: Addressing Amplification Bias and Homogeneity Issue in Recommendations for Large Language Models",
author = "Bao, Keqin and
Zhang, Jizhi and
Zhang, Yang and
Huo, Xinyue and
Chen, Chong and
Feng, Fuli",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.589",
pages = "10540--10552",
abstract = "Adapting Large Language Models (LLMs) for recommendation requires careful consideration of the decoding process, given the inherent differences between generating items and natural language. Existing approaches often directly apply LLMs{'} original decoding methods. However, we find these methods encounter significant challenges: 1) amplification bias{---}where standard length normalization inflates scores for items containing tokens with generation probabilities close to 1 (termed ghost tokens), and 2) homogeneity issue{---}generating multiple similar or repetitive items for a user. To tackle these challenges, we introduce a new decoding approach named Debiasing-Diversifying Decoding ($D^3$). $D^3$ disables length normalization for ghost tokens to alleviate amplification bias, and it incorporates a text-free assistant model to encourage tokens less frequently generated by LLMs for counteracting recommendation homogeneity. Extensive experiments on real-world datasets demonstrate the method{'}s effectiveness in enhancing accuracy and diversity.",
}
| Adapting Large Language Models (LLMs) for recommendation requires careful consideration of the decoding process, given the inherent differences between generating items and natural language. Existing approaches often directly apply LLMs{'} original decoding methods. However, we find these methods encounter significant challenges: 1) amplification bias{---}where standard length normalization inflates scores for items containing tokens with generation probabilities close to 1 (termed ghost tokens), and 2) homogeneity issue{---}generating multiple similar or repetitive items for a user. To tackle these challenges, we introduce a new decoding approach named Debiasing-Diversifying Decoding ($D^3$). $D^3$ disables length normalization for ghost tokens to alleviate amplification bias, and it incorporates a text-free assistant model to encourage tokens less frequently generated by LLMs for counteracting recommendation homogeneity. Extensive experiments on real-world datasets demonstrate the method{'}s effectiveness in enhancing accuracy and diversity. | [
"Bao, Keqin",
"Zhang, Jizhi",
"Zhang, Yang",
"Huo, Xinyue",
"Chen, Chong",
"Feng, Fuli"
] | Decoding Matters: Addressing Amplification Bias and Homogeneity Issue in Recommendations for Large Language Models | emnlp-main.589 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.590.bib | https://aclanthology.org/2024.emnlp-main.590/ | @inproceedings{joshi-etal-2024-llms,
title = "{LLM}s Are Prone to Fallacies in Causal Inference",
author = "Joshi, Nitish and
Saparov, Abulhair and
Wang, Yixin and
He, He",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.590",
pages = "10553--10569",
abstract = "Recent work shows that causal facts can be effectively extracted from LLMs through prompting, facilitating the creation of causal graphs for causal inference tasks. However, it is unclear if this success is limited to explicitly-mentioned causal facts in the pretraining data which the model can memorize. Thus, this work investigates: Can LLMs infer causal relations from other relational data in text? To disentangle the role of memorized causal facts vs inferred causal relations, we finetune LLMs on synthetic data containing temporal, spatial and counterfactual relations, and measure whether the LLM can then infer causal relations. We find that: (a) LLMs are susceptible to inferring causal relations from the order of two entity mentions in text (e.g. X mentioned before Y implies X causes Y); (b) if the order is randomized, LLMs still suffer from the post hoc fallacy, i.e. X occurs before Y (temporal relation) implies X causes Y. We also find that while LLMs can correctly deduce the absence of causal relations from temporal and spatial relations, they have difficulty inferring causal relations from counterfactuals, questioning their understanding of causality.",
}
| Recent work shows that causal facts can be effectively extracted from LLMs through prompting, facilitating the creation of causal graphs for causal inference tasks. However, it is unclear if this success is limited to explicitly-mentioned causal facts in the pretraining data which the model can memorize. Thus, this work investigates: Can LLMs infer causal relations from other relational data in text? To disentangle the role of memorized causal facts vs inferred causal relations, we finetune LLMs on synthetic data containing temporal, spatial and counterfactual relations, and measure whether the LLM can then infer causal relations. We find that: (a) LLMs are susceptible to inferring causal relations from the order of two entity mentions in text (e.g. X mentioned before Y implies X causes Y); (b) if the order is randomized, LLMs still suffer from the post hoc fallacy, i.e. X occurs before Y (temporal relation) implies X causes Y. We also find that while LLMs can correctly deduce the absence of causal relations from temporal and spatial relations, they have difficulty inferring causal relations from counterfactuals, questioning their understanding of causality. | [
"Joshi, Nitish",
"Saparov, Abulhair",
"Wang, Yixin",
"He, He"
] | LLMs Are Prone to Fallacies in Causal Inference | emnlp-main.590 | Poster | 2406.12158 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.591.bib | https://aclanthology.org/2024.emnlp-main.591/ | @inproceedings{louie-etal-2024-roleplay,
title = "Roleplay-doh: Enabling Domain-Experts to Create {LLM}-simulated Patients via Eliciting and Adhering to Principles",
author = "Louie, Ryan and
Nandi, Ananjan and
Fang, William and
Chang, Cheng and
Brunskill, Emma and
Yang, Diyi",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.591",
pages = "10570--10603",
abstract = "Recent works leverage LLMs to roleplay realistic social scenarios, aiding novices in practicing their social skills. However, simulating sensitive interactions, such as in the domain of mental health, is challenging. Privacy concerns restrict data access, and collecting expert feedback, although vital, is laborious. To address this, we develop Roleplay-doh, a novel human-LLM collaboration pipeline that elicits qualitative feedback from a domain-expert, which is transformed into a set of principles, or natural language rules, that govern an LLM-prompted roleplay. We apply this pipeline to enable senior mental health supporters to create customized AI patients as simulated practice partners for novice counselors. After uncovering issues with basic GPT-4 simulations not adhering to expert-defined principles, we also introduce a novel principle-adherence prompting pipeline which shows a 30{\%} improvement in response quality and principle following for the downstream task. Through a user study with 25 counseling experts, we demonstrate that the pipeline makes it easy and effective to create AI patients that more faithfully resemble real patients, as judged by both creators and third-party counselors. We provide access to the code and data on our project website: https://roleplay-doh.github.io/.",
}
| Recent works leverage LLMs to roleplay realistic social scenarios, aiding novices in practicing their social skills. However, simulating sensitive interactions, such as in the domain of mental health, is challenging. Privacy concerns restrict data access, and collecting expert feedback, although vital, is laborious. To address this, we develop Roleplay-doh, a novel human-LLM collaboration pipeline that elicits qualitative feedback from a domain-expert, which is transformed into a set of principles, or natural language rules, that govern an LLM-prompted roleplay. We apply this pipeline to enable senior mental health supporters to create customized AI patients as simulated practice partners for novice counselors. After uncovering issues with basic GPT-4 simulations not adhering to expert-defined principles, we also introduce a novel principle-adherence prompting pipeline which shows a 30{\%} improvement in response quality and principle following for the downstream task. Through a user study with 25 counseling experts, we demonstrate that the pipeline makes it easy and effective to create AI patients that more faithfully resemble real patients, as judged by both creators and third-party counselors. We provide access to the code and data on our project website: https://roleplay-doh.github.io/. | [
"Louie, Ryan",
"N",
"i, Ananjan",
"Fang, William",
"Chang, Cheng",
"Brunskill, Emma",
"Yang, Diyi"
] | Roleplay-doh: Enabling Domain-Experts to Create LLM-simulated Patients via Eliciting and Adhering to Principles | emnlp-main.591 | Oral | 2407.00870 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.592.bib | https://aclanthology.org/2024.emnlp-main.592/ | @inproceedings{waldis-etal-2024-lou,
title = "The {L}ou Dataset - Exploring the Impact of Gender-Fair Language in {G}erman Text Classification",
author = "Waldis, Andreas and
Birrer, Joel and
Lauscher, Anne and
Gurevych, Iryna",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.592",
pages = "10604--10624",
abstract = "Gender-fair language, an evolving linguistic variation in German, fosters inclusion by addressing all genders or using neutral forms. However, there is a notable lack of resources to assess the impact of this language shift on language models (LMs) might not been trained on examples of this variation. Addressing this gap, we present Lou, the first dataset providing high-quality reformulations for German text classification covering seven tasks, like stance detection and toxicity classification. We evaluate 16 mono- and multi-lingual LMs and find substantial label flips, reduced prediction certainty, and significantly altered attention patterns. However, existing evaluations remain valid, as LM rankings are consistent across original and reformulated instances. Our study provides initial insights into the impact of gender-fair language on classification for German. However, these findings are likely transferable to other languages, as we found consistent patterns in multi-lingual and English LMs.",
}
| Gender-fair language, an evolving linguistic variation in German, fosters inclusion by addressing all genders or using neutral forms. However, there is a notable lack of resources to assess the impact of this language shift on language models (LMs) might not been trained on examples of this variation. Addressing this gap, we present Lou, the first dataset providing high-quality reformulations for German text classification covering seven tasks, like stance detection and toxicity classification. We evaluate 16 mono- and multi-lingual LMs and find substantial label flips, reduced prediction certainty, and significantly altered attention patterns. However, existing evaluations remain valid, as LM rankings are consistent across original and reformulated instances. Our study provides initial insights into the impact of gender-fair language on classification for German. However, these findings are likely transferable to other languages, as we found consistent patterns in multi-lingual and English LMs. | [
"Waldis, Andreas",
"Birrer, Joel",
"Lauscher, Anne",
"Gurevych, Iryna"
] | The Lou Dataset - Exploring the Impact of Gender-Fair Language in German Text Classification | emnlp-main.592 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.593.bib | https://aclanthology.org/2024.emnlp-main.593/ | @inproceedings{tong-etal-2024-generative,
title = "When Generative Adversarial Networks Meet Sequence Labeling Challenges",
author = "Tong, Yu and
Chen, Ge and
Zheng, Guokai and
Li, Rui and
Dazhi, Jiang",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.593",
pages = "10625--10635",
abstract = "The current framework for sequence labeling encompasses a feature extractor and a sequence tagger. This study introduces a unified framework named SLGAN, which harnesses the capabilities of Generative Adversarial Networks to address the challenges associated with Sequence Labeling tasks. SLGAN not only mitigates the limitation of GANs in backpropagating loss to discrete data but also exhibits strong adaptability to various sequence labeling tasks. Unlike traditional GANs, the discriminator within SLGAN does not discriminate whether data originates from the discriminator or the generator; instead, it focuses on predicting the correctness of each tag within the tag sequence. We conducted evaluations on six different tasks spanning four languages, including Chinese, Japanese, and Korean Word Segmentation, Chinese and English Named Entity Recognition, and Chinese Part-of-Speech Tagging. Our experimental results illustrate that SLGAN represents a versatile and highly effective solution, consistently achieving state-of-the-art or competitive performance results, irrespective of the specific task or language under consideration.",
}
| The current framework for sequence labeling encompasses a feature extractor and a sequence tagger. This study introduces a unified framework named SLGAN, which harnesses the capabilities of Generative Adversarial Networks to address the challenges associated with Sequence Labeling tasks. SLGAN not only mitigates the limitation of GANs in backpropagating loss to discrete data but also exhibits strong adaptability to various sequence labeling tasks. Unlike traditional GANs, the discriminator within SLGAN does not discriminate whether data originates from the discriminator or the generator; instead, it focuses on predicting the correctness of each tag within the tag sequence. We conducted evaluations on six different tasks spanning four languages, including Chinese, Japanese, and Korean Word Segmentation, Chinese and English Named Entity Recognition, and Chinese Part-of-Speech Tagging. Our experimental results illustrate that SLGAN represents a versatile and highly effective solution, consistently achieving state-of-the-art or competitive performance results, irrespective of the specific task or language under consideration. | [
"Tong, Yu",
"Chen, Ge",
"Zheng, Guokai",
"Li, Rui",
"Dazhi, Jiang"
] | When Generative Adversarial Networks Meet Sequence Labeling Challenges | emnlp-main.593 | Poster | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 1 |
||
https://aclanthology.org/2024.emnlp-main.594.bib | https://aclanthology.org/2024.emnlp-main.594/ | @inproceedings{ko-etal-2024-evidence,
title = "Evidence-Focused Fact Summarization for Knowledge-Augmented Zero-Shot Question Answering",
author = "Ko, Sungho and
Cho, Hyunjin and
Chae, Hyungjoo and
Yeo, Jinyoung and
Lee, Dongha",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.594",
pages = "10636--10651",
abstract = "Recent studies have investigated utilizing Knowledge Graphs (KGs) to enhance Quesetion Answering (QA) performance of Large Language Models (LLMs), yet structured KG verbalization remains challenging. Existing methods, like concatenation or free-form textual conversion of triples, have limitations, including duplicated entities or relations, reduced evidence density, and failure to highlight crucial evidence. To address these issues, we propose EFSum, an Evidence-focused Fact Summarization framework for enhanced QA with knowledge-augmented LLMs. We optimize an LLM as a fact summarizer through distillation and preference alignment. Our extensive expeirments show that EFSum improves LLM{'}s zero-shot QA performance with its helpful and faithful summaries, especially when noisy facts are retrieved.",
}
| Recent studies have investigated utilizing Knowledge Graphs (KGs) to enhance Quesetion Answering (QA) performance of Large Language Models (LLMs), yet structured KG verbalization remains challenging. Existing methods, like concatenation or free-form textual conversion of triples, have limitations, including duplicated entities or relations, reduced evidence density, and failure to highlight crucial evidence. To address these issues, we propose EFSum, an Evidence-focused Fact Summarization framework for enhanced QA with knowledge-augmented LLMs. We optimize an LLM as a fact summarizer through distillation and preference alignment. Our extensive expeirments show that EFSum improves LLM{'}s zero-shot QA performance with its helpful and faithful summaries, especially when noisy facts are retrieved. | [
"Ko, Sungho",
"Cho, Hyunjin",
"Chae, Hyungjoo",
"Yeo, Jinyoung",
"Lee, Dongha"
] | Evidence-Focused Fact Summarization for Knowledge-Augmented Zero-Shot Question Answering | emnlp-main.594 | Oral | 2403.02966 | [
"https://github.com/anon809/efsum"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.595.bib | https://aclanthology.org/2024.emnlp-main.595/ | @inproceedings{cho-etal-2024-speechworthy,
title = "Speechworthy Instruction-tuned Language Models",
author = "Cho, Hyundong Justin and
Jedema, Nicolaas Paul and
Ribeiro, Leonardo F. R. and
Sharma, Karishma and
Szekely, Pedro and
Moschitti, Alessandro and
Janssen, Ruben and
May, Jonathan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.595",
pages = "10652--10670",
abstract = "Current instruction-tuned language models are exclusively trained with textual preference data and thus may not be aligned to the unique requirements of other modalities, such as speech. To better align language models with the speech domain, we explore i) prompting strategies based on radio-industry best practices and ii) preference learning using a novel speech-based preference data of 20K samples collected by annotators who listen to response pairs. Both human and automatic evaluation show that both prompting and preference learning increase the speech-suitability of popular instruction tuned LLMs. More interestingly, we show that these methods are additive; combining them achieves the best win rates in head-to-head comparison, resulting in responses that are preferred or tied to the base model in 76.2{\%} of comparisons on average. Lastly, we share lexical, syntactical, and qualitative analyses that elicit how our studied methods differ with baselines in generating more speech-suitable responses.",
}
| Current instruction-tuned language models are exclusively trained with textual preference data and thus may not be aligned to the unique requirements of other modalities, such as speech. To better align language models with the speech domain, we explore i) prompting strategies based on radio-industry best practices and ii) preference learning using a novel speech-based preference data of 20K samples collected by annotators who listen to response pairs. Both human and automatic evaluation show that both prompting and preference learning increase the speech-suitability of popular instruction tuned LLMs. More interestingly, we show that these methods are additive; combining them achieves the best win rates in head-to-head comparison, resulting in responses that are preferred or tied to the base model in 76.2{\%} of comparisons on average. Lastly, we share lexical, syntactical, and qualitative analyses that elicit how our studied methods differ with baselines in generating more speech-suitable responses. | [
"Cho, Hyundong Justin",
"Jedema, Nicolaas Paul",
"Ribeiro, Leonardo F. R.",
"Sharma, Karishma",
"Szekely, Pedro",
"Moschitti, Aless",
"ro",
"Janssen, Ruben",
"May, Jonathan"
] | Speechworthy Instruction-tuned Language Models | emnlp-main.595 | Poster | 2409.14672 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.596.bib | https://aclanthology.org/2024.emnlp-main.596/ | @inproceedings{parmar-etal-2024-data,
title = "Data, Data Everywhere: A Guide for Pretraining Dataset Construction",
author = "Parmar, Jupinder and
Prabhumoye, Shrimai and
Jennings, Joseph and
Liu, Bo and
Jhunjhunwala, Aastha and
Wang, Zhilin and
Patwary, Mostofa and
Shoeybi, Mohammad and
Catanzaro, Bryan",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.596",
pages = "10671--10695",
abstract = "The impressive capabilities of recent language models can be largely attributed to the multi-trillion token pretraining datasets that they are trained on. However, model developers fail to disclose their construction methodology which has lead to a lack of open information on how to develop effective pretraining sets. To address this issue, we perform the first systematic study across the entire pipeline of pretraining set construction. First, we run ablations on existing techniques for pretraining set development to identify which methods translate to the largest gains in model accuracy on downstream evaluations. Then, we categorize the most widely used data source, web crawl snapshots, across the attributes of toxicity, quality, type of speech, and domain. Finally, we show how such attribute information can be used to further refine and improve the quality of a pretraining set. These findings constitute an actionable set of steps that practitioners can use to develop high quality pretraining sets.",
}
| The impressive capabilities of recent language models can be largely attributed to the multi-trillion token pretraining datasets that they are trained on. However, model developers fail to disclose their construction methodology which has lead to a lack of open information on how to develop effective pretraining sets. To address this issue, we perform the first systematic study across the entire pipeline of pretraining set construction. First, we run ablations on existing techniques for pretraining set development to identify which methods translate to the largest gains in model accuracy on downstream evaluations. Then, we categorize the most widely used data source, web crawl snapshots, across the attributes of toxicity, quality, type of speech, and domain. Finally, we show how such attribute information can be used to further refine and improve the quality of a pretraining set. These findings constitute an actionable set of steps that practitioners can use to develop high quality pretraining sets. | [
"Parmar, Jupinder",
"Prabhumoye, Shrimai",
"Jennings, Joseph",
"Liu, Bo",
"Jhunjhunwala, Aastha",
"Wang, Zhilin",
"Patwary, Mostofa",
"Shoeybi, Mohammad",
"Catanzaro, Bryan"
] | Data, Data Everywhere: A Guide for Pretraining Dataset Construction | emnlp-main.596 | Oral | 2407.06380 | [
""
] | https://huggingface.co/papers/2407.06380 | 2 | 0 | 0 | 9 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.597.bib | https://aclanthology.org/2024.emnlp-main.597/ | @inproceedings{soylu-etal-2024-fine,
title = "Fine-Tuning and Prompt Optimization: Two Great Steps that Work Better Together",
author = "Soylu, Dilara and
Potts, Christopher and
Khattab, Omar",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.597",
pages = "10696--10710",
abstract = "Natural Language Processing (NLP) systems are increasingly taking the form of sophisticated modular pipelines, e.g., Retrieval Augmented Generation (RAG), where each module may involve a distinct Language Model (LM) and an associated prompt template. These compound systems often lack intermediate labels or gradient flow to optimize each module, making their end-to-end optimization challenging. Here we seek strategies to optimize both the module-level LM weights and the associated prompt templates of such systems to maximize a downstream task metric. We propose for the first time combining the weight and prompt optimization strategies to optimize a modular LM pipeline by alternating between the two to get the same LM to teach itself. In experiments with multi-hop QA, mathematical reasoning, and feature-based classification using mistral-7b, llama-2-7b, and llama-3-8b, these BetterTogether strategies optimizing the weights and prompts of a pipeline together outperform directly optimizing weights alone and prompts alone by up to 60{\%} and 6{\%}, respectively, on average across LMs and tasks. Our BetterTogether optimizer is released in DSPy at [http://dspy.ai](http://dspy.ai).",
}
| Natural Language Processing (NLP) systems are increasingly taking the form of sophisticated modular pipelines, e.g., Retrieval Augmented Generation (RAG), where each module may involve a distinct Language Model (LM) and an associated prompt template. These compound systems often lack intermediate labels or gradient flow to optimize each module, making their end-to-end optimization challenging. Here we seek strategies to optimize both the module-level LM weights and the associated prompt templates of such systems to maximize a downstream task metric. We propose for the first time combining the weight and prompt optimization strategies to optimize a modular LM pipeline by alternating between the two to get the same LM to teach itself. In experiments with multi-hop QA, mathematical reasoning, and feature-based classification using mistral-7b, llama-2-7b, and llama-3-8b, these BetterTogether strategies optimizing the weights and prompts of a pipeline together outperform directly optimizing weights alone and prompts alone by up to 60{\%} and 6{\%}, respectively, on average across LMs and tasks. Our BetterTogether optimizer is released in DSPy at [http://dspy.ai](http://dspy.ai). | [
"Soylu, Dilara",
"Potts, Christopher",
"Khattab, Omar"
] | Fine-Tuning and Prompt Optimization: Two Great Steps that Work Better Together | emnlp-main.597 | Poster | 2407.10930 | [
""
] | https://huggingface.co/papers/2407.10930 | 0 | 0 | 0 | 3 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.598.bib | https://aclanthology.org/2024.emnlp-main.598/ | @inproceedings{huang-etal-2024-demystifying,
title = "Demystifying Verbatim Memorization in Large Language Models",
author = "Huang, Jing and
Yang, Diyi and
Potts, Christopher",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.598",
pages = "10711--10732",
abstract = "Large Language Models (LLMs) frequently memorize long sequences verbatim, often with serious legal and privacy implications. Much prior work has studied such verbatim memorization using observational data. To complement such work, we develop a framework to study verbatim memorization in a controlled setting by continuing pre-training from Pythia checkpoints with injected sequences. We find that (1) non-trivial amounts of repetition are necessary for verbatim memorization to happen; (2) later (and presumably better) checkpoints are more likely to verbatim memorize sequences, even for out-of-distribution sequences; (3) the generation of memorized sequences is triggered by distributed model states that encode high-level features and makes important use of general language modeling capabilities. Guided by these insights, we develop stress tests to evaluate unlearning methods and find they often fail to remove the verbatim memorized information, while also degrading the LM. Overall, these findings challenge the hypothesis that verbatim memorization stems from specific model weights or mechanisms. Rather, verbatim memorization is intertwined with the LM{'}s general capabilities and thus will be very difficult to isolate and suppress without degrading model quality.",
}
| Large Language Models (LLMs) frequently memorize long sequences verbatim, often with serious legal and privacy implications. Much prior work has studied such verbatim memorization using observational data. To complement such work, we develop a framework to study verbatim memorization in a controlled setting by continuing pre-training from Pythia checkpoints with injected sequences. We find that (1) non-trivial amounts of repetition are necessary for verbatim memorization to happen; (2) later (and presumably better) checkpoints are more likely to verbatim memorize sequences, even for out-of-distribution sequences; (3) the generation of memorized sequences is triggered by distributed model states that encode high-level features and makes important use of general language modeling capabilities. Guided by these insights, we develop stress tests to evaluate unlearning methods and find they often fail to remove the verbatim memorized information, while also degrading the LM. Overall, these findings challenge the hypothesis that verbatim memorization stems from specific model weights or mechanisms. Rather, verbatim memorization is intertwined with the LM{'}s general capabilities and thus will be very difficult to isolate and suppress without degrading model quality. | [
"Huang, Jing",
"Yang, Diyi",
"Potts, Christopher"
] | Demystifying Verbatim Memorization in Large Language Models | emnlp-main.598 | Poster | 2407.17817 | [
"https://github.com/explanare/verbatim-memorization"
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
|
https://aclanthology.org/2024.emnlp-main.599.bib | https://aclanthology.org/2024.emnlp-main.599/ | @inproceedings{niwa-iso-2024-ambignlg,
title = "{A}mbig{NLG}: Addressing Task Ambiguity in Instruction for {NLG}",
author = "Niwa, Ayana and
Iso, Hayate",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.599",
pages = "10733--10752",
}
| No abstract found | [
"Niwa, Ayana",
"Iso, Hayate"
] | AmbigNLG: Addressing Task Ambiguity in Instruction for NLG | emnlp-main.599 | Poster | 2402.17717 | [
"https://github.com/megagonlabs/ambignlg"
] | https://huggingface.co/papers/2402.17717 | 2 | 0 | 0 | 2 | [] | [] | [] | [] | [] | [] | 1 |
https://aclanthology.org/2024.emnlp-main.600.bib | https://aclanthology.org/2024.emnlp-main.600/ | @inproceedings{cognetta-etal-2024-distributional,
title = "Distributional Properties of Subword Regularization",
author = "Cognetta, Marco and
Zouhar, Vil{\'e}m and
Okazaki, Naoaki",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.600",
pages = "10753--10763",
abstract = "Subword regularization, used widely in NLP, improves model performance by reducing the dependency on exact tokenizations, augmenting the training corpus, and exposing the model to more unique contexts during training. BPE and MaxMatch, two popular subword tokenization schemes, have stochastic dropout regularization variants. However, there has not been an analysis of the distributions formed by them.We show that these stochastic variants are heavily biased towards a small set of tokenizations per word. If the benefits of subword regularization are as mentioned, we hypothesize that biasedness artificially limits the effectiveness of these schemes. Thus, we propose an algorithm to uniformly sample tokenizations that we use as a drop-in replacement for the stochastic aspects of existing tokenizers, and find that it improves machine translation quality.",
}
| Subword regularization, used widely in NLP, improves model performance by reducing the dependency on exact tokenizations, augmenting the training corpus, and exposing the model to more unique contexts during training. BPE and MaxMatch, two popular subword tokenization schemes, have stochastic dropout regularization variants. However, there has not been an analysis of the distributions formed by them.We show that these stochastic variants are heavily biased towards a small set of tokenizations per word. If the benefits of subword regularization are as mentioned, we hypothesize that biasedness artificially limits the effectiveness of these schemes. Thus, we propose an algorithm to uniformly sample tokenizations that we use as a drop-in replacement for the stochastic aspects of existing tokenizers, and find that it improves machine translation quality. | [
"Cognetta, Marco",
"Zouhar, Vil{\\'e}m",
"Okazaki, Naoaki"
] | Distributional Properties of Subword Regularization | emnlp-main.600 | Poster | 2408.11443 | [
""
] | -1 | -1 | -1 | -1 | [] | [] | [] | [] | [] | [] | 0 |
Subsets and Splits