@inproceedings{xia-etal-2024-language,
title = "Language Models can Evaluate Themselves via Probability Discrepancy",
author = "Xia, Tingyu and
Yu, Bowen and
Wu, Yuan and
Chang, Yi and
Zhou, Chang",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.291",
pages = "4889--4901",
abstract = "In this paper, we begin by illustrating that, when presented with a query, Large Language Models (LLMs) capable of providing accurate responses tend to exhibit a more uniform probability distribution compared to their less proficient counterparts. Building upon this observation, we introduce a novel self-assessment criterion termed ProbDiff for evaluating the performance of diverse LLMs. This method eliminates the need for training an additional evaluation model or relying on external proprietary models such as GPT-4 as a judger. Instead, it solely relies on the LLMs under evaluation to compute the probability discrepancy between the original response generation and its revised versions. A higher discrepancy in two LLMs for the same query suggests a relatively weaker ability. We discover that ProbDiff yields comparable results to mainstream GPT-4-based evaluations on various scenarios including NLG tasks like translation and summarization, as well as LLM evaluation benchmarks such as AlignBench, MT-Bench, and AlpacaEval, across LLMs of different sizes.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="xia-etal-2024-language">
<titleInfo>
<title>Language Models can Evaluate Themselves via Probability Discrepancy</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tingyu</namePart>
<namePart type="family">Xia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bowen</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuan</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chang</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics ACL 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lun-Wei</namePart>
<namePart type="family">Ku</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andre</namePart>
<namePart type="family">Martins</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Srikumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Bangkok, Thailand and virtual meeting</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>In this paper, we begin by illustrating that, when presented with a query, Large Language Models (LLMs) capable of providing accurate responses tend to exhibit a more uniform probability distribution compared to their less proficient counterparts. Building upon this observation, we introduce a novel self-assessment criterion termed ProbDiff for evaluating the performance of diverse LLMs. This method eliminates the need for training an additional evaluation model or relying on external proprietary models such as GPT-4 as a judger. Instead, it solely relies on the LLMs under evaluation to compute the probability discrepancy between the original response generation and its revised versions. A higher discrepancy in two LLMs for the same query suggests a relatively weaker ability. We discover that ProbDiff yields comparable results to mainstream GPT-4-based evaluations on various scenarios including NLG tasks like translation and summarization, as well as LLM evaluation benchmarks such as AlignBench, MT-Bench, and AlpacaEval, across LLMs of different sizes.</abstract>
<identifier type="citekey">xia-etal-2024-language</identifier>
<location>
<url>https://aclanthology.org/2024.findings-acl.291</url>
</location>
<part>
<date>2024-08</date>
<extent unit="page">
<start>4889</start>
<end>4901</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Language Models can Evaluate Themselves via Probability Discrepancy
%A Xia, Tingyu
%A Yu, Bowen
%A Wu, Yuan
%A Chang, Yi
%A Zhou, Chang
%Y Ku, Lun-Wei
%Y Martins, Andre
%Y Srikumar, Vivek
%S Findings of the Association for Computational Linguistics ACL 2024
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand and virtual meeting
%F xia-etal-2024-language
%X In this paper, we begin by illustrating that, when presented with a query, Large Language Models (LLMs) capable of providing accurate responses tend to exhibit a more uniform probability distribution compared to their less proficient counterparts. Building upon this observation, we introduce a novel self-assessment criterion termed ProbDiff for evaluating the performance of diverse LLMs. This method eliminates the need for training an additional evaluation model or relying on external proprietary models such as GPT-4 as a judger. Instead, it solely relies on the LLMs under evaluation to compute the probability discrepancy between the original response generation and its revised versions. A higher discrepancy in two LLMs for the same query suggests a relatively weaker ability. We discover that ProbDiff yields comparable results to mainstream GPT-4-based evaluations on various scenarios including NLG tasks like translation and summarization, as well as LLM evaluation benchmarks such as AlignBench, MT-Bench, and AlpacaEval, across LLMs of different sizes.
%U https://aclanthology.org/2024.findings-acl.291
%P 4889-4901
Markdown (Informal)
[Language Models can Evaluate Themselves via Probability Discrepancy](https://aclanthology.org/2024.findings-acl.291) (Xia et al., Findings 2024)
ACL