@inproceedings{hwang-etal-2026-marbo,
title = "{MARBO}: Relational Belief Grounding for {LLM} Agents in Social Deduction Games",
author = "Hwang, Yechan and
Bae, Sangjun and
Kim, Jeongmo and
Bang, Sangwoo and
Han, Seungyul",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1376/",
pages = "31414--31445",
ISBN = "979-8-89176-459-0",
abstract = "Social deduction games (SDGs) require agents to reason under partial observability by maintaining relational beliefs about hidden roles, team alignments, suspicion, and others' perceptions. While recent LLM-agent approaches improve gameplay through prompting, reflection, and preference optimization, they often optimize actions and in-game speech without explicitly grounding them in such beliefs. This frequently leads to strategically inconsistent behavior, especially for compact LLM agents. We introduce Multi-Agent Relational Belief Optimization (MARBO), a belief-grounded preference optimization framework that leverages relational beliefs to guide strategic decisions and in-game speech. MARBO selectively provides preference feedback only when behaviors are supported by reliable relational beliefs and lead to strategically favorable social outcomes, encouraging more consistent learning under uncertainty. Experiments on representative SDGs, including Werewolf and Among Us, show that MARBO enables compact LLM agents to consistently outperform existing baselines."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="hwang-etal-2026-marbo">
<titleInfo>
<title>MARBO: Relational Belief Grounding for LLM Agents in Social Deduction Games</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yechan</namePart>
<namePart type="family">Hwang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sangjun</namePart>
<namePart type="family">Bae</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeongmo</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sangwoo</namePart>
<namePart type="family">Bang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seungyul</namePart>
<namePart type="family">Han</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>Social deduction games (SDGs) require agents to reason under partial observability by maintaining relational beliefs about hidden roles, team alignments, suspicion, and others’ perceptions. While recent LLM-agent approaches improve gameplay through prompting, reflection, and preference optimization, they often optimize actions and in-game speech without explicitly grounding them in such beliefs. This frequently leads to strategically inconsistent behavior, especially for compact LLM agents. We introduce Multi-Agent Relational Belief Optimization (MARBO), a belief-grounded preference optimization framework that leverages relational beliefs to guide strategic decisions and in-game speech. MARBO selectively provides preference feedback only when behaviors are supported by reliable relational beliefs and lead to strategically favorable social outcomes, encouraging more consistent learning under uncertainty. Experiments on representative SDGs, including Werewolf and Among Us, show that MARBO enables compact LLM agents to consistently outperform existing baselines.</abstract>
<identifier type="citekey">hwang-etal-2026-marbo</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1376/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>31414</start>
<end>31445</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T MARBO: Relational Belief Grounding for LLM Agents in Social Deduction Games
%A Hwang, Yechan
%A Bae, Sangjun
%A Kim, Jeongmo
%A Bang, Sangwoo
%A Han, Seungyul
%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 hwang-etal-2026-marbo
%X Social deduction games (SDGs) require agents to reason under partial observability by maintaining relational beliefs about hidden roles, team alignments, suspicion, and others’ perceptions. While recent LLM-agent approaches improve gameplay through prompting, reflection, and preference optimization, they often optimize actions and in-game speech without explicitly grounding them in such beliefs. This frequently leads to strategically inconsistent behavior, especially for compact LLM agents. We introduce Multi-Agent Relational Belief Optimization (MARBO), a belief-grounded preference optimization framework that leverages relational beliefs to guide strategic decisions and in-game speech. MARBO selectively provides preference feedback only when behaviors are supported by reliable relational beliefs and lead to strategically favorable social outcomes, encouraging more consistent learning under uncertainty. Experiments on representative SDGs, including Werewolf and Among Us, show that MARBO enables compact LLM agents to consistently outperform existing baselines.
%U https://aclanthology.org/2026.emnlp-main.1376/
%P 31414-31445
Markdown (Informal)
[MARBO: Relational Belief Grounding for LLM Agents in Social Deduction Games](https://aclanthology.org/2026.emnlp-main.1376/) (Hwang et al., EMNLP 2026)
ACL