@inproceedings{zahraei-etal-2026-prior,
title = "Prior Beliefs Prejudice {LLM}-as-Judge: Evidence from Persuasion Evaluation",
author = {Zahraei, Pardis Sadat and
Wang, Xiaoning and
Bozdag, Nimet Beyza and
Tur, Gokhan and
Hakkani-T{\"u}r, Dilek},
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.2087/",
pages = "42049--42082",
ISBN = "979-8-89176-395-1",
abstract = "Large Language Models (LLMs) are increasingly used as judges to evaluate text quality, moderate content, and assess arguments. We investigate whether alignment-instilled prior beliefs bias LLM judgments, using persuasion evaluation as a representative task. We find a systematic failure: models conflate their trained beliefs with rhetorical quality, rating identical claims differently based on belief alignment rather than argumentative merit. A bare assertion aligned with training receives higher scores than a well-crafted counter-argument, even when explicitly instructed to judge rhetoric alone. We introduce ConvinceQA, a dataset of 27,756 persuasive arguments with controlled stance variation across subjective, harmful, and misinformation domains, and demonstrate this prior prejudice across models. We exploit this failure through persuasion-based probing: evaluating minimal pairs that differ only in the subject token bypasses learned refusals and reveals hidden biases. Analysis identifies three failure modes, with belief-conditioned rating inflation accounting for 88{\%} of cases. Cross-task validation on essay quality assessment and debate judging confirms this is a pervasive limitation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zahraei-etal-2026-prior">
<titleInfo>
<title>Prior Beliefs Prejudice LLM-as-Judge: Evidence from Persuasion Evaluation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pardis</namePart>
<namePart type="given">Sadat</namePart>
<namePart type="family">Zahraei</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaoning</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nimet</namePart>
<namePart type="given">Beyza</namePart>
<namePart type="family">Bozdag</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gokhan</namePart>
<namePart type="family">Tur</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dilek</namePart>
<namePart type="family">Hakkani-Tür</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>Findings of the Association for Computational Linguistics: ACL 2026</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">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Large Language Models (LLMs) are increasingly used as judges to evaluate text quality, moderate content, and assess arguments. We investigate whether alignment-instilled prior beliefs bias LLM judgments, using persuasion evaluation as a representative task. We find a systematic failure: models conflate their trained beliefs with rhetorical quality, rating identical claims differently based on belief alignment rather than argumentative merit. A bare assertion aligned with training receives higher scores than a well-crafted counter-argument, even when explicitly instructed to judge rhetoric alone. We introduce ConvinceQA, a dataset of 27,756 persuasive arguments with controlled stance variation across subjective, harmful, and misinformation domains, and demonstrate this prior prejudice across models. We exploit this failure through persuasion-based probing: evaluating minimal pairs that differ only in the subject token bypasses learned refusals and reveals hidden biases. Analysis identifies three failure modes, with belief-conditioned rating inflation accounting for 88% of cases. Cross-task validation on essay quality assessment and debate judging confirms this is a pervasive limitation.</abstract>
<identifier type="citekey">zahraei-etal-2026-prior</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.2087/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>42049</start>
<end>42082</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Prior Beliefs Prejudice LLM-as-Judge: Evidence from Persuasion Evaluation
%A Zahraei, Pardis Sadat
%A Wang, Xiaoning
%A Bozdag, Nimet Beyza
%A Tur, Gokhan
%A Hakkani-Tür, Dilek
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F zahraei-etal-2026-prior
%X Large Language Models (LLMs) are increasingly used as judges to evaluate text quality, moderate content, and assess arguments. We investigate whether alignment-instilled prior beliefs bias LLM judgments, using persuasion evaluation as a representative task. We find a systematic failure: models conflate their trained beliefs with rhetorical quality, rating identical claims differently based on belief alignment rather than argumentative merit. A bare assertion aligned with training receives higher scores than a well-crafted counter-argument, even when explicitly instructed to judge rhetoric alone. We introduce ConvinceQA, a dataset of 27,756 persuasive arguments with controlled stance variation across subjective, harmful, and misinformation domains, and demonstrate this prior prejudice across models. We exploit this failure through persuasion-based probing: evaluating minimal pairs that differ only in the subject token bypasses learned refusals and reveals hidden biases. Analysis identifies three failure modes, with belief-conditioned rating inflation accounting for 88% of cases. Cross-task validation on essay quality assessment and debate judging confirms this is a pervasive limitation.
%U https://aclanthology.org/2026.findings-acl.2087/
%P 42049-42082
Markdown (Informal)
[Prior Beliefs Prejudice LLM-as-Judge: Evidence from Persuasion Evaluation](https://aclanthology.org/2026.findings-acl.2087/) (Zahraei et al., Findings 2026)
ACL