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

TL;DR

This paper introduces a hierarchical uncertainty metric for multi-agent debate systems, enabling detection of failures and proposing an uncertainty-driven policy to improve decision accuracy and system reliability.

Contribution

It develops a novel hierarchical uncertainty quantification method and an uncertainty-driven policy optimization to mitigate debate collapse in multi-agent systems.

Findings

Uncertainty metrics reliably indicate system failures.

Mitigation improves decision accuracy.

Reduces system disagreement.

Abstract

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.