@inproceedings{xu-etal-2025-progress,
title = "The Progress Illusion: Revisiting meta-evaluation standards of {LLM} evaluators",
author = "Xu, Tianruo Rose and
Gaur, Vedant and
Leqi, Liu and
Goyal, Tanya",
editor = "Christodoulopoulos, Christos and
Chakraborty, Tanmoy and
Rose, Carolyn and
Peng, Violet",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-emnlp.1036/",
doi = "10.18653/v1/2025.findings-emnlp.1036",
pages = "19033--19043",
ISBN = "979-8-89176-335-7",
abstract = "LLM judges have gained popularity as an inexpensive and performant substitute for human evaluation. However, we observe that the meta-evaluation setting in which the reliability of these LLM evaluators is established is substantially different from their use in model development. To address this, we revisit meta-evaluations of LLM evaluators under a setting that more closely aligns with practice by examining evaluators' ability to distinguish test system pairs that are closer in capability. Our fine-grained approach shows that all LLM evaluator{'}s correlations with human judgments are concerningly low when the models perform similarly, showcasing a key limitation of current norms. Equipped with this better methodology, we next analyze the impact that the choice of the reference model makes to LLM-as-a-judge evaluator performance. We show that single-reference evaluators only perform well at ranking test systems that fall within particular capability ranges, even if the standard meta-evaluation reports high overall correlation. Taken together, our analysis shows critical issues with current LLM meta-evaluation and recommend avenues for improvement."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="xu-etal-2025-progress">
<titleInfo>
<title>The Progress Illusion: Revisiting meta-evaluation standards of LLM evaluators</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tianruo</namePart>
<namePart type="given">Rose</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vedant</namePart>
<namePart type="family">Gaur</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liu</namePart>
<namePart type="family">Leqi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanya</namePart>
<namePart type="family">Goyal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Christos</namePart>
<namePart type="family">Christodoulopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanmoy</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carolyn</namePart>
<namePart type="family">Rose</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Violet</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-335-7</identifier>
</relatedItem>
<abstract>LLM judges have gained popularity as an inexpensive and performant substitute for human evaluation. However, we observe that the meta-evaluation setting in which the reliability of these LLM evaluators is established is substantially different from their use in model development. To address this, we revisit meta-evaluations of LLM evaluators under a setting that more closely aligns with practice by examining evaluators’ ability to distinguish test system pairs that are closer in capability. Our fine-grained approach shows that all LLM evaluator’s correlations with human judgments are concerningly low when the models perform similarly, showcasing a key limitation of current norms. Equipped with this better methodology, we next analyze the impact that the choice of the reference model makes to LLM-as-a-judge evaluator performance. We show that single-reference evaluators only perform well at ranking test systems that fall within particular capability ranges, even if the standard meta-evaluation reports high overall correlation. Taken together, our analysis shows critical issues with current LLM meta-evaluation and recommend avenues for improvement.</abstract>
<identifier type="citekey">xu-etal-2025-progress</identifier>
<identifier type="doi">10.18653/v1/2025.findings-emnlp.1036</identifier>
<location>
<url>https://aclanthology.org/2025.findings-emnlp.1036/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>19033</start>
<end>19043</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T The Progress Illusion: Revisiting meta-evaluation standards of LLM evaluators
%A Xu, Tianruo Rose
%A Gaur, Vedant
%A Leqi, Liu
%A Goyal, Tanya
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Findings of the Association for Computational Linguistics: EMNLP 2025
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-335-7
%F xu-etal-2025-progress
%X LLM judges have gained popularity as an inexpensive and performant substitute for human evaluation. However, we observe that the meta-evaluation setting in which the reliability of these LLM evaluators is established is substantially different from their use in model development. To address this, we revisit meta-evaluations of LLM evaluators under a setting that more closely aligns with practice by examining evaluators’ ability to distinguish test system pairs that are closer in capability. Our fine-grained approach shows that all LLM evaluator’s correlations with human judgments are concerningly low when the models perform similarly, showcasing a key limitation of current norms. Equipped with this better methodology, we next analyze the impact that the choice of the reference model makes to LLM-as-a-judge evaluator performance. We show that single-reference evaluators only perform well at ranking test systems that fall within particular capability ranges, even if the standard meta-evaluation reports high overall correlation. Taken together, our analysis shows critical issues with current LLM meta-evaluation and recommend avenues for improvement.
%R 10.18653/v1/2025.findings-emnlp.1036
%U https://aclanthology.org/2025.findings-emnlp.1036/
%U https://doi.org/10.18653/v1/2025.findings-emnlp.1036
%P 19033-19043
Markdown (Informal)
[The Progress Illusion: Revisiting meta-evaluation standards of LLM evaluators](https://aclanthology.org/2025.findings-emnlp.1036/) (Xu et al., Findings 2025)
ACL