<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mukherjee-etal-2026-geometry">
    <titleInfo>
        <title>The Geometry of LLM-as-Judge: Why Inter-LLM Consensus Is Not Human Alignment</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Sourabrata</namePart>
        <namePart type="family">Mukherjee</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Hamna</namePart>
        <namePart type="family">Abid</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Kalika</namePart>
        <namePart type="family">Bali</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Sunayana</namePart>
        <namePart type="family">Sitaram</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
    </relatedItem>
    <abstract>LLM judges now score most open-ended NLP output, and their mutual agreement is routinely read as evidence that the scores can be trusted. That reading is unsafe: judges may agree because they capture quality, or because they share the same blind spots, and agreement statistics alone cannot tell these apart. We develop a geometric test that can. Treating each judge’s scores as a vector, we measure spread, effective rank, the angle to human scores, and the judge–judge, judge–human, and human–human agreement triple for 42 judges on two community-built Indic benchmarks covering four domains and eight languages. Every comparison is reference-matched: a judge and a held-out rater are scored against the same two-rater mean, since an averaged reference otherwise flatters judges by several degrees. On subjective rubrics, judges agree with one another as much as humans do yet reach only 58–66% of human agreement: consensus on an axis humans do not weight. On the one rubric with a verifiable answer, most of that gap closes. Ensembles converge on the judges’ shared axis rather than the human one, and training widens scores without rotating them. Inter-judge agreement is evidence of human alignment only after this check passes.</abstract>
    <identifier type="citekey">mukherjee-etal-2026-geometry</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1696/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>36237</start>
            <end>36266</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T The Geometry of LLM-as-Judge: Why Inter-LLM Consensus Is Not Human Alignment
%A Mukherjee, Sourabrata
%A Abid, Hamna
%A Bali, Kalika
%A Sitaram, Sunayana
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F mukherjee-etal-2026-geometry
%X LLM judges now score most open-ended NLP output, and their mutual agreement is routinely read as evidence that the scores can be trusted. That reading is unsafe: judges may agree because they capture quality, or because they share the same blind spots, and agreement statistics alone cannot tell these apart. We develop a geometric test that can. Treating each judge’s scores as a vector, we measure spread, effective rank, the angle to human scores, and the judge–judge, judge–human, and human–human agreement triple for 42 judges on two community-built Indic benchmarks covering four domains and eight languages. Every comparison is reference-matched: a judge and a held-out rater are scored against the same two-rater mean, since an averaged reference otherwise flatters judges by several degrees. On subjective rubrics, judges agree with one another as much as humans do yet reach only 58–66% of human agreement: consensus on an axis humans do not weight. On the one rubric with a verifiable answer, most of that gap closes. Ensembles converge on the judges’ shared axis rather than the human one, and training widens scores without rotating them. Inter-judge agreement is evidence of human alignment only after this check passes.
%U https://aclanthology.org/2026.findings-emnlp.1696/
%P 36237-36266

Markdown (Informal)

[The Geometry of LLM-as-Judge: Why Inter-LLM Consensus Is Not Human Alignment](https://aclanthology.org/2026.findings-emnlp.1696/) (Mukherjee et al., Findings 2026)

• The Geometry of LLM-as-Judge: Why Inter-LLM Consensus Is Not Human Alignment (Mukherjee et al., Findings 2026)

ACL

• Sourabrata Mukherjee, Hamna Abid, Kalika Bali, and Sunayana Sitaram. 2026. The Geometry of LLM-as-Judge: Why Inter-LLM Consensus Is Not Human Alignment. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 36237–36266, Budapest, Hungary. Association for Computational Linguistics.