@inproceedings{song-zheng-2026-beyond,
title = "Beyond the Illusion of Consensus: From Surface Heuristics to Knowledge-Grounded Evaluation in {LLM}-as-a-Judge",
author = "Song, Mingyang and
Zheng, Mao",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1552/",
pages = "35416--35433",
ISBN = "979-8-89176-459-0",
abstract = "Agreement among LLM judges is widely used as evidence that automated evaluation is reliable. Yet consensus may be stable without reflecting a shared understanding of quality. We call this \textbf{Evaluation Illusion}: judges can produce detailed critiques while relying on overlapping surface cues or a common evaluation instrument. In a fully crossed baseline study, three frontier judges conduct 105,600 evaluations of 32 models. First, for the two highest-scoring models, moving from a shared five-dimension rubric package to independently generated rubrics reduces agreement by 62{\%}, a joint contrast we term the \textbf{Rubric Commensurability Problem}. Second, nearly identical model-level rankings coexist with substantially weaker criterion-level agreement under non-equivalent metrics and sample sets, yielding the \textbf{Resolution Paradox}. We further introduce MERG, a four-stage protocol that elicits task knowledge and potential biases before scoring. In a controlled comparison on the same models at $t{=}0.0$, agreement decreases under independently generated MERG rubrics. These results do not establish that MERG scores are more accurate; they show that consensus depends on the evaluation instrument and analytical resolution at which it is measured."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="song-zheng-2026-beyond">
<titleInfo>
<title>Beyond the Illusion of Consensus: From Surface Heuristics to Knowledge-Grounded Evaluation in LLM-as-a-Judge</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mingyang</namePart>
<namePart type="family">Song</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mao</namePart>
<namePart type="family">Zheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Agreement among LLM judges is widely used as evidence that automated evaluation is reliable. Yet consensus may be stable without reflecting a shared understanding of quality. We call this Evaluation Illusion: judges can produce detailed critiques while relying on overlapping surface cues or a common evaluation instrument. In a fully crossed baseline study, three frontier judges conduct 105,600 evaluations of 32 models. First, for the two highest-scoring models, moving from a shared five-dimension rubric package to independently generated rubrics reduces agreement by 62%, a joint contrast we term the Rubric Commensurability Problem. Second, nearly identical model-level rankings coexist with substantially weaker criterion-level agreement under non-equivalent metrics and sample sets, yielding the Resolution Paradox. We further introduce MERG, a four-stage protocol that elicits task knowledge and potential biases before scoring. In a controlled comparison on the same models at t=0.0, agreement decreases under independently generated MERG rubrics. These results do not establish that MERG scores are more accurate; they show that consensus depends on the evaluation instrument and analytical resolution at which it is measured.</abstract>
<identifier type="citekey">song-zheng-2026-beyond</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1552/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>35416</start>
<end>35433</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Beyond the Illusion of Consensus: From Surface Heuristics to Knowledge-Grounded Evaluation in LLM-as-a-Judge
%A Song, Mingyang
%A Zheng, Mao
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F song-zheng-2026-beyond
%X Agreement among LLM judges is widely used as evidence that automated evaluation is reliable. Yet consensus may be stable without reflecting a shared understanding of quality. We call this Evaluation Illusion: judges can produce detailed critiques while relying on overlapping surface cues or a common evaluation instrument. In a fully crossed baseline study, three frontier judges conduct 105,600 evaluations of 32 models. First, for the two highest-scoring models, moving from a shared five-dimension rubric package to independently generated rubrics reduces agreement by 62%, a joint contrast we term the Rubric Commensurability Problem. Second, nearly identical model-level rankings coexist with substantially weaker criterion-level agreement under non-equivalent metrics and sample sets, yielding the Resolution Paradox. We further introduce MERG, a four-stage protocol that elicits task knowledge and potential biases before scoring. In a controlled comparison on the same models at t=0.0, agreement decreases under independently generated MERG rubrics. These results do not establish that MERG scores are more accurate; they show that consensus depends on the evaluation instrument and analytical resolution at which it is measured.
%U https://aclanthology.org/2026.emnlp-main.1552/
%P 35416-35433
Markdown (Informal)
[Beyond the Illusion of Consensus: From Surface Heuristics to Knowledge-Grounded Evaluation in LLM-as-a-Judge](https://aclanthology.org/2026.emnlp-main.1552/) (Song & Zheng, EMNLP 2026)
ACL