The Value of Variance: Mitigating Debate Collapse in Multi-Agent Systems via Uncertainty-Driven Policy Optimization

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Tang, Luoxi, Meng, Yuqiao, Costa, Joseph, Zhang, Yingxue, Ye, Muchao, Xi, Zhaohan
Format:	Preprint
Published:	2026
Subjects:	Multiagent Systems Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866912885779726336
author	Tang, Luoxi Meng, Yuqiao Costa, Joseph Zhang, Yingxue Ye, Muchao Xi, Zhaohan
author_facet	Tang, Luoxi Meng, Yuqiao Costa, Joseph Zhang, Yingxue Ye, Muchao Xi, Zhaohan
contents	Multi-agent debate (MAD) systems improve LLM reasoning through iterative deliberation, but remain vulnerable to debate collapse, a failure type where final agent decisions are compromised on erroneous reasoning. Existing methods lack principled mechanisms to detect or prevent such failures. To address this gap, we first propose a hierarchical metric that quantifies behavioral uncertainty at three levels: intra-agent (individual reasoning uncertainty), inter-agent (interactive uncertainty), and system-level (output uncertainty). Empirical analysis across several benchmarks reveals that our proposed uncertainty quantification reliably indicates system failures, which demonstrates the validity of using them as diagnostic metrics to indicate the system failure. Subsequently, we propose a mitigation strategy by formulating an uncertainty-driven policy optimization to penalize self-contradiction, peer conflict, and low-confidence outputs in a dynamic debating environment. Experiments demonstrate that our proposed uncertainty-driven mitigation reliably calibrates the multi-agent system by consistently improving decision accuracy while reducing system disagreement.
format	Preprint
id	arxiv_https___arxiv_org_abs_2602_07186
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	The Value of Variance: Mitigating Debate Collapse in Multi-Agent Systems via Uncertainty-Driven Policy Optimization Tang, Luoxi Meng, Yuqiao Costa, Joseph Zhang, Yingxue Ye, Muchao Xi, Zhaohan Multiagent Systems Machine Learning Multi-agent debate (MAD) systems improve LLM reasoning through iterative deliberation, but remain vulnerable to debate collapse, a failure type where final agent decisions are compromised on erroneous reasoning. Existing methods lack principled mechanisms to detect or prevent such failures. To address this gap, we first propose a hierarchical metric that quantifies behavioral uncertainty at three levels: intra-agent (individual reasoning uncertainty), inter-agent (interactive uncertainty), and system-level (output uncertainty). Empirical analysis across several benchmarks reveals that our proposed uncertainty quantification reliably indicates system failures, which demonstrates the validity of using them as diagnostic metrics to indicate the system failure. Subsequently, we propose a mitigation strategy by formulating an uncertainty-driven policy optimization to penalize self-contradiction, peer conflict, and low-confidence outputs in a dynamic debating environment. Experiments demonstrate that our proposed uncertainty-driven mitigation reliably calibrates the multi-agent system by consistently improving decision accuracy while reducing system disagreement.
title	The Value of Variance: Mitigating Debate Collapse in Multi-Agent Systems via Uncertainty-Driven Policy Optimization
topic	Multiagent Systems Machine Learning
url	https://arxiv.org/abs/2602.07186