@inproceedings{fu-etal-2023-large,
title = "Are Large Language Models Reliable Judges? A Study on the Factuality Evaluation Capabilities of {LLM}s",
author = "Fu, Xue-Yong and
Laskar, Md Tahmid Rahman and
Chen, Cheng and
Tn, Shashi Bhushan",
editor = "Gehrmann, Sebastian and
Wang, Alex and
Sedoc, Jo{\~a}o and
Clark, Elizabeth and
Dhole, Kaustubh and
Chandu, Khyathi Raghavi and
Santus, Enrico and
Sedghamiz, Hooman",
booktitle = "Proceedings of the Third Workshop on Natural Language Generation, Evaluation, and Metrics (GEM)",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.gem-1.25",
pages = "310--316",
abstract = "In recent years, large language models (LLMs) have drawn significant attention due to their impressive emergent capabilities that were not observed in earlier language models. One emerging area where LLMs have been widely used in recent times is the utilization of LLMs as the evaluator of the texts generated by various generative models. In this paper, we also explore the possibility of whether LLMs are reliable in assessing the factual consistency of summaries generated by text generation models. We first propose a new approach to evaluate the factuality score using LLMs by utilizing the same LLM to perform all steps in the question-answering-based factuality scoring pipeline. Subsequently, we study the performance of various LLMs to directly score the factuality. Our evaluation is conducted in traditional benchmarks by comparing their correlation with human annotations. Contrary to expectations, our findings revealed that none of the factuality metrics showed any significant correlations (e.g., coefficient scores greater than 0.3) to human evaluations of factuality for GPT-4, PaLM-2, and Claude-2, with the only exception being GPT-3.5 in two subcategories of factuality. Nonetheless, our findings are consistent across almost all factual error types, suggesting a fundamental limitation in the ability of current LLMs to assess factuality.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="fu-etal-2023-large">
<titleInfo>
<title>Are Large Language Models Reliable Judges? A Study on the Factuality Evaluation Capabilities of LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xue-Yong</namePart>
<namePart type="family">Fu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Md</namePart>
<namePart type="given">Tahmid</namePart>
<namePart type="given">Rahman</namePart>
<namePart type="family">Laskar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Cheng</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shashi</namePart>
<namePart type="given">Bhushan</namePart>
<namePart type="family">Tn</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-12</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Third Workshop on Natural Language Generation, Evaluation, and Metrics (GEM)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Gehrmann</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alex</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">João</namePart>
<namePart type="family">Sedoc</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Elizabeth</namePart>
<namePart type="family">Clark</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaustubh</namePart>
<namePart type="family">Dhole</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Khyathi</namePart>
<namePart type="given">Raghavi</namePart>
<namePart type="family">Chandu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Enrico</namePart>
<namePart type="family">Santus</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hooman</namePart>
<namePart type="family">Sedghamiz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Singapore</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>In recent years, large language models (LLMs) have drawn significant attention due to their impressive emergent capabilities that were not observed in earlier language models. One emerging area where LLMs have been widely used in recent times is the utilization of LLMs as the evaluator of the texts generated by various generative models. In this paper, we also explore the possibility of whether LLMs are reliable in assessing the factual consistency of summaries generated by text generation models. We first propose a new approach to evaluate the factuality score using LLMs by utilizing the same LLM to perform all steps in the question-answering-based factuality scoring pipeline. Subsequently, we study the performance of various LLMs to directly score the factuality. Our evaluation is conducted in traditional benchmarks by comparing their correlation with human annotations. Contrary to expectations, our findings revealed that none of the factuality metrics showed any significant correlations (e.g., coefficient scores greater than 0.3) to human evaluations of factuality for GPT-4, PaLM-2, and Claude-2, with the only exception being GPT-3.5 in two subcategories of factuality. Nonetheless, our findings are consistent across almost all factual error types, suggesting a fundamental limitation in the ability of current LLMs to assess factuality.</abstract>
<identifier type="citekey">fu-etal-2023-large</identifier>
<location>
<url>https://aclanthology.org/2023.gem-1.25</url>
</location>
<part>
<date>2023-12</date>
<extent unit="page">
<start>310</start>
<end>316</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Are Large Language Models Reliable Judges? A Study on the Factuality Evaluation Capabilities of LLMs
%A Fu, Xue-Yong
%A Laskar, Md Tahmid Rahman
%A Chen, Cheng
%A Tn, Shashi Bhushan
%Y Gehrmann, Sebastian
%Y Wang, Alex
%Y Sedoc, João
%Y Clark, Elizabeth
%Y Dhole, Kaustubh
%Y Chandu, Khyathi Raghavi
%Y Santus, Enrico
%Y Sedghamiz, Hooman
%S Proceedings of the Third Workshop on Natural Language Generation, Evaluation, and Metrics (GEM)
%D 2023
%8 December
%I Association for Computational Linguistics
%C Singapore
%F fu-etal-2023-large
%X In recent years, large language models (LLMs) have drawn significant attention due to their impressive emergent capabilities that were not observed in earlier language models. One emerging area where LLMs have been widely used in recent times is the utilization of LLMs as the evaluator of the texts generated by various generative models. In this paper, we also explore the possibility of whether LLMs are reliable in assessing the factual consistency of summaries generated by text generation models. We first propose a new approach to evaluate the factuality score using LLMs by utilizing the same LLM to perform all steps in the question-answering-based factuality scoring pipeline. Subsequently, we study the performance of various LLMs to directly score the factuality. Our evaluation is conducted in traditional benchmarks by comparing their correlation with human annotations. Contrary to expectations, our findings revealed that none of the factuality metrics showed any significant correlations (e.g., coefficient scores greater than 0.3) to human evaluations of factuality for GPT-4, PaLM-2, and Claude-2, with the only exception being GPT-3.5 in two subcategories of factuality. Nonetheless, our findings are consistent across almost all factual error types, suggesting a fundamental limitation in the ability of current LLMs to assess factuality.
%U https://aclanthology.org/2023.gem-1.25
%P 310-316
Markdown (Informal)
[Are Large Language Models Reliable Judges? A Study on the Factuality Evaluation Capabilities of LLMs](https://aclanthology.org/2023.gem-1.25) (Fu et al., GEM-WS 2023)
ACL