<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-balancing">
    <titleInfo>
        <title>Balancing Epistemic Value and Social Safety: PACR-Pro for False-Consensus Intervention in One Night Ultimate Werewolf</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Yonghui</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zhiliang</namePart>
        <namePart type="family">Tian</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Bo</namePart>
        <namePart type="family">Liu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jiarui</namePart>
        <namePart type="family">Tang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Wanlong</namePart>
        <namePart type="family">Yu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Shilong</namePart>
        <namePart type="family">Pan</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
    </relatedItem>
    <abstract>Large language model agents in social deduction games must reason and speak under hidden roles, shifting public beliefs, and social pressure. Existing methods mainly target final win rate through strategic prompting, belief modeling, or reinforcement learning, but they leave the utterance-level trade-off between belief correction and social risk largely implicit. In false-consensus intervention, useful private information can move the group toward the truth, while poorly timed or overly direct speech can trigger suspicion or rejection. We propose PACR-Pro, a risk-aware rhetorical decision framework for false-consensus intervention. PACR-Pro maps each speaking step to consensus-conflict encoding, structured candidate-action generation, Pareto-based selection, and risk-adaptive utterance realization. A dual-branch Epistemic-Social Critic scores candidate actions, and a state-conditioned łambda policy adjusts the trade-off between epistemic value and social safety. We train these decision modules offline from historical games and evaluate PACR-Pro on One Night Ultimate Werewolf (ONUW) as the test platform. Experiments on standard ONUW games and the Wolf-Pack Induction Benchmark show that PACR-Pro achieves the best mixed-agent win rate, the highest attack win rate, the lowest attack vote rate, and the smallest attack-control win-rate gap among all baselines.</abstract>
    <identifier type="citekey">wang-etal-2026-balancing</identifier>
    <location>
        <url>https://aclanthology.org/2026.emnlp-main.2587/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>58786</start>
            <end>58810</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Balancing Epistemic Value and Social Safety: PACR-Pro for False-Consensus Intervention in One Night Ultimate Werewolf
%A Wang, Yonghui
%A Tian, Zhiliang
%A Liu, Bo
%A Tang, Jiarui
%A Yu, Wanlong
%A Pan, Shilong
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F wang-etal-2026-balancing
%X Large language model agents in social deduction games must reason and speak under hidden roles, shifting public beliefs, and social pressure. Existing methods mainly target final win rate through strategic prompting, belief modeling, or reinforcement learning, but they leave the utterance-level trade-off between belief correction and social risk largely implicit. In false-consensus intervention, useful private information can move the group toward the truth, while poorly timed or overly direct speech can trigger suspicion or rejection. We propose PACR-Pro, a risk-aware rhetorical decision framework for false-consensus intervention. PACR-Pro maps each speaking step to consensus-conflict encoding, structured candidate-action generation, Pareto-based selection, and risk-adaptive utterance realization. A dual-branch Epistemic-Social Critic scores candidate actions, and a state-conditioned łambda policy adjusts the trade-off between epistemic value and social safety. We train these decision modules offline from historical games and evaluate PACR-Pro on One Night Ultimate Werewolf (ONUW) as the test platform. Experiments on standard ONUW games and the Wolf-Pack Induction Benchmark show that PACR-Pro achieves the best mixed-agent win rate, the highest attack win rate, the lowest attack vote rate, and the smallest attack-control win-rate gap among all baselines.
%U https://aclanthology.org/2026.emnlp-main.2587/
%P 58786-58810

Markdown (Informal)

[Balancing Epistemic Value and Social Safety: PACR-Pro for False-Consensus Intervention in One Night Ultimate Werewolf](https://aclanthology.org/2026.emnlp-main.2587/) (Wang et al., EMNLP 2026)

• Balancing Epistemic Value and Social Safety: PACR-Pro for False-Consensus Intervention in One Night Ultimate Werewolf (Wang et al., EMNLP 2026)

ACL

• Yonghui Wang, Zhiliang Tian, Bo Liu, Jiarui Tang, Wanlong Yu, and Shilong Pan. 2026. Balancing Epistemic Value and Social Safety: PACR-Pro for False-Consensus Intervention in One Night Ultimate Werewolf. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 58786–58810, Budapest, Hungary. Association for Computational Linguistics.