@inproceedings{asgari-etal-2026-quantifying,
title = "Quantifying Metric and Model Agreement in Bias Evaluation of Large Language Models",
author = "Asgari, Arash and
Wu, Huan and
Naziri, Amirreza and
Kolahdouzi, Mojtaba and
Seyyed-Kalantari, Laleh",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-long.769/",
pages = "16868--16933",
ISBN = "979-8-89176-390-6",
abstract = "Bias evaluation in large language models (LLMs) uses many metrics and benchmarks, but lacks a systematic way to measure agreement across bias metrics and models. As a result, improvements observed under one metric may contradict another, and model rankings may reflect benchmark-specific artifacts rather than stable bias profiles. In this work, we introduce Metric Agreement Score (MeAS) and Model Agreement Score (MoAS), which quantify cross-metric and cross-model agreement in bias rankings, respectively. We apply these measures to eight LLMs, seven bias metrics, and nine corpora. Our results reveal disagreement among both metrics and models: Contrary to expectations, we find that metrics within the same category (generation-based and probabilistic) often behave independently of each other. For instance, HONEST shows independence with toxicity metrics, and the Context Association Test shows no correlation with Language Modeling Bias metric. At the model level, DeepSeek-family models invert bias rankings relative to most others, indicating that the model family strongly shapes specific bias profiles. These findings challenge the assumption that bias mitigation is universally transferable and highlight the need for agreement-aware evaluation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="asgari-etal-2026-quantifying">
<titleInfo>
<title>Quantifying Metric and Model Agreement in Bias Evaluation of Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Arash</namePart>
<namePart type="family">Asgari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Huan</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Amirreza</namePart>
<namePart type="family">Naziri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mojtaba</namePart>
<namePart type="family">Kolahdouzi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Laleh</namePart>
<namePart type="family">Seyyed-Kalantari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-390-6</identifier>
</relatedItem>
<abstract>Bias evaluation in large language models (LLMs) uses many metrics and benchmarks, but lacks a systematic way to measure agreement across bias metrics and models. As a result, improvements observed under one metric may contradict another, and model rankings may reflect benchmark-specific artifacts rather than stable bias profiles. In this work, we introduce Metric Agreement Score (MeAS) and Model Agreement Score (MoAS), which quantify cross-metric and cross-model agreement in bias rankings, respectively. We apply these measures to eight LLMs, seven bias metrics, and nine corpora. Our results reveal disagreement among both metrics and models: Contrary to expectations, we find that metrics within the same category (generation-based and probabilistic) often behave independently of each other. For instance, HONEST shows independence with toxicity metrics, and the Context Association Test shows no correlation with Language Modeling Bias metric. At the model level, DeepSeek-family models invert bias rankings relative to most others, indicating that the model family strongly shapes specific bias profiles. These findings challenge the assumption that bias mitigation is universally transferable and highlight the need for agreement-aware evaluation.</abstract>
<identifier type="citekey">asgari-etal-2026-quantifying</identifier>
<location>
<url>https://aclanthology.org/2026.acl-long.769/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>16868</start>
<end>16933</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Quantifying Metric and Model Agreement in Bias Evaluation of Large Language Models
%A Asgari, Arash
%A Wu, Huan
%A Naziri, Amirreza
%A Kolahdouzi, Mojtaba
%A Seyyed-Kalantari, Laleh
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-390-6
%F asgari-etal-2026-quantifying
%X Bias evaluation in large language models (LLMs) uses many metrics and benchmarks, but lacks a systematic way to measure agreement across bias metrics and models. As a result, improvements observed under one metric may contradict another, and model rankings may reflect benchmark-specific artifacts rather than stable bias profiles. In this work, we introduce Metric Agreement Score (MeAS) and Model Agreement Score (MoAS), which quantify cross-metric and cross-model agreement in bias rankings, respectively. We apply these measures to eight LLMs, seven bias metrics, and nine corpora. Our results reveal disagreement among both metrics and models: Contrary to expectations, we find that metrics within the same category (generation-based and probabilistic) often behave independently of each other. For instance, HONEST shows independence with toxicity metrics, and the Context Association Test shows no correlation with Language Modeling Bias metric. At the model level, DeepSeek-family models invert bias rankings relative to most others, indicating that the model family strongly shapes specific bias profiles. These findings challenge the assumption that bias mitigation is universally transferable and highlight the need for agreement-aware evaluation.
%U https://aclanthology.org/2026.acl-long.769/
%P 16868-16933
Markdown (Informal)
[Quantifying Metric and Model Agreement in Bias Evaluation of Large Language Models](https://aclanthology.org/2026.acl-long.769/) (Asgari et al., ACL 2026)
ACL