<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhao-etal-2026-debiasing">
    <titleInfo>
        <title>Debiasing as a Measurement Intervention: Calibrated Ties and Resolution Loss in LLM-as-a-Judge Evaluation</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Liang</namePart>
        <namePart type="family">Zhao</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Yong</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jiangzhe</namePart>
        <namePart type="family">Chen</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-as-a-judge protocols are commonly debiased by instructing judges to ignore presentation cues such as citation formatting, source labels, and evidence-display style. We show that this intervention can suppress bias while damaging the resolution of the measurement instrument. We introduce TraceJudgeBench, a diagnostic benchmark for auditing citation-like artifacts in RAG and agent-workflow evaluation, covering content-equivalent pairs, citation ablations, correctness conflicts, human-validated soft and moderate quality gaps, prompt-strength ladders, decoupled judging, and a controlled workflow-ranking probe. Across GPT-5.5, Claude Sonnet 4.6, and DeepSeek V4-Flash, stronger anti-citation prompts reduce worse-cited wins from up to 50.5% to 0%; yet some operating points already convert validated moderate-gap decisions into Tie before the strict stress-test endpoint, while correctness-conflict accuracy remains at or above 93.0%. Open-weight Qwen2.5-14B-Instruct-AWQ and Gemma-3-12B-IT runs reproduce the central moderate-gap frontier, with TRACE-style decoupling recovering 96.5–98.5% better-plain resolution. Human validation separates three meanings of Tie: correct equivalence Tie, calibrated soft-boundary Tie, and resolution-destroying Tie on validated quality gaps. We frame debiasing as a measurement intervention whose bias suppression, resolution retention, Tie cost, and protocol cost must be reported jointly. We include TraceJudgeBench as a supplementary artifact with anonymized splits, prompts, raw judge outputs, validation summaries, and analysis scripts.</abstract>
    <identifier type="citekey">zhao-etal-2026-debiasing</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1880/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>40091</start>
            <end>40112</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Debiasing as a Measurement Intervention: Calibrated Ties and Resolution Loss in LLM-as-a-Judge Evaluation
%A Zhao, Liang
%A Wang, Yong
%A Chen, Jiangzhe
%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 zhao-etal-2026-debiasing
%X LLM-as-a-judge protocols are commonly debiased by instructing judges to ignore presentation cues such as citation formatting, source labels, and evidence-display style. We show that this intervention can suppress bias while damaging the resolution of the measurement instrument. We introduce TraceJudgeBench, a diagnostic benchmark for auditing citation-like artifacts in RAG and agent-workflow evaluation, covering content-equivalent pairs, citation ablations, correctness conflicts, human-validated soft and moderate quality gaps, prompt-strength ladders, decoupled judging, and a controlled workflow-ranking probe. Across GPT-5.5, Claude Sonnet 4.6, and DeepSeek V4-Flash, stronger anti-citation prompts reduce worse-cited wins from up to 50.5% to 0%; yet some operating points already convert validated moderate-gap decisions into Tie before the strict stress-test endpoint, while correctness-conflict accuracy remains at or above 93.0%. Open-weight Qwen2.5-14B-Instruct-AWQ and Gemma-3-12B-IT runs reproduce the central moderate-gap frontier, with TRACE-style decoupling recovering 96.5–98.5% better-plain resolution. Human validation separates three meanings of Tie: correct equivalence Tie, calibrated soft-boundary Tie, and resolution-destroying Tie on validated quality gaps. We frame debiasing as a measurement intervention whose bias suppression, resolution retention, Tie cost, and protocol cost must be reported jointly. We include TraceJudgeBench as a supplementary artifact with anonymized splits, prompts, raw judge outputs, validation summaries, and analysis scripts.
%U https://aclanthology.org/2026.findings-emnlp.1880/
%P 40091-40112

Markdown (Informal)

[Debiasing as a Measurement Intervention: Calibrated Ties and Resolution Loss in LLM-as-a-Judge Evaluation](https://aclanthology.org/2026.findings-emnlp.1880/) (Zhao et al., Findings 2026)

• Debiasing as a Measurement Intervention: Calibrated Ties and Resolution Loss in LLM-as-a-Judge Evaluation (Zhao et al., Findings 2026)

ACL

• Liang Zhao, Yong Wang, and Jiangzhe Chen. 2026. Debiasing as a Measurement Intervention: Calibrated Ties and Resolution Loss in LLM-as-a-Judge Evaluation. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 40091–40112, Budapest, Hungary. Association for Computational Linguistics.