@inproceedings{gao-chen-2026-learning,
title = "Learning to Elicit {LLM} Bias in Multi-Turn Conversations",
author = "Gao, Da and
Chen, Ziyu",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the {A}sia-Pacific Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 2: Short Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-short.52/",
pages = "816--829",
ISBN = "979-8-89176-440-8",
abstract = "Static bias benchmarks score a model on isolated questions, yet LLMs that answer safely at first can still reveal bias as a conversation unfolds. We frame bias evaluation as multi-turn conversational elicitation and learn an interviewer that probes a target LLM, conditioned on a demographic group, a bias category, and a conversational strategy. The interviewer is trained with supervised fine-tuning on synthetic dialogues and refined with preference optimization on conversation-level feedback. Across all models we test, bias grows as the conversation continues, exposing failures that single-turn evaluation misses. Training the interviewer matters: it surfaces substantially more bias than the same model prompted with identical instructions, and preference optimization pushes it further. Multi-turn elicitation and single-turn search turn out to be complementary: at matched query budgets they surface similar amounts of bias, but conversation uncovers more severe cases. Human annotations corroborate these judgments. Our results argue that bias evaluation of conversational AI should move beyond single-turn probes."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="gao-chen-2026-learning">
<titleInfo>
<title>Learning to Elicit LLM Bias in Multi-Turn Conversations</title>
</titleInfo>
<name type="personal">
<namePart type="given">Da</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ziyu</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-440-8</identifier>
</relatedItem>
<abstract>Static bias benchmarks score a model on isolated questions, yet LLMs that answer safely at first can still reveal bias as a conversation unfolds. We frame bias evaluation as multi-turn conversational elicitation and learn an interviewer that probes a target LLM, conditioned on a demographic group, a bias category, and a conversational strategy. The interviewer is trained with supervised fine-tuning on synthetic dialogues and refined with preference optimization on conversation-level feedback. Across all models we test, bias grows as the conversation continues, exposing failures that single-turn evaluation misses. Training the interviewer matters: it surfaces substantially more bias than the same model prompted with identical instructions, and preference optimization pushes it further. Multi-turn elicitation and single-turn search turn out to be complementary: at matched query budgets they surface similar amounts of bias, but conversation uncovers more severe cases. Human annotations corroborate these judgments. Our results argue that bias evaluation of conversational AI should move beyond single-turn probes.</abstract>
<identifier type="citekey">gao-chen-2026-learning</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-short.52/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>816</start>
<end>829</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Learning to Elicit LLM Bias in Multi-Turn Conversations
%A Gao, Da
%A Chen, Ziyu
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-440-8
%F gao-chen-2026-learning
%X Static bias benchmarks score a model on isolated questions, yet LLMs that answer safely at first can still reveal bias as a conversation unfolds. We frame bias evaluation as multi-turn conversational elicitation and learn an interviewer that probes a target LLM, conditioned on a demographic group, a bias category, and a conversational strategy. The interviewer is trained with supervised fine-tuning on synthetic dialogues and refined with preference optimization on conversation-level feedback. Across all models we test, bias grows as the conversation continues, exposing failures that single-turn evaluation misses. Training the interviewer matters: it surfaces substantially more bias than the same model prompted with identical instructions, and preference optimization pushes it further. Multi-turn elicitation and single-turn search turn out to be complementary: at matched query budgets they surface similar amounts of bias, but conversation uncovers more severe cases. Human annotations corroborate these judgments. Our results argue that bias evaluation of conversational AI should move beyond single-turn probes.
%U https://aclanthology.org/2026.aacl-short.52/
%P 816-829
Markdown (Informal)
[Learning to Elicit LLM Bias in Multi-Turn Conversations](https://aclanthology.org/2026.aacl-short.52/) (Gao & Chen, AACL-IJCNLP 2026)
ACL
- Da Gao and Ziyu Chen. 2026. Learning to Elicit LLM Bias in Multi-Turn Conversations. In Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 2: Short Papers), pages 816–829, Hengqin, China. Association for Computational Linguistics.