Group Perspective Matters: Regulating Debate Relationships Can Mitigate Blind Conformity in Multi-Agent Debate

2026-08-04

AI Summary

This work addresses the susceptibility of large language models to blind conformity in multi-agent debates, which undermines reasoning accuracy—a limitation inadequately mitigated by existing evaluation methods based on individual confidence or perplexity. To counter this, the authors propose DEAR, a novel framework that, for the first time, incorporates collective evidence—capturing both consensus and disagreement—into debate state modeling to dynamically regulate inter-agent “debate relationships” and suppress herd behavior. DEAR employs two reinforcement learning agents operating across three stages: perceiving group deliberation tendencies, dynamically selecting reference peers, and adaptively adjusting generation behavior. Experiments demonstrate that DEAR not only significantly reduces token consumption but also achieves superior reasoning performance compared to current approaches.

Abstract

Multi-Agent Debate (MAD) improves the reasoning performance of Large Language Models (LLMs) through multi-round interaction. However, LLMs in MAD are highly susceptible to blind conformity. Existing individual evaluation methods, typically based on confidence or perplexity, fail to reflect the correctness of reasoning and may even exacerbate blind conformity. To address this, we shift the perspective from individual evaluation to group interaction. We define mutual referencing among LLMs as \textbf{Debate Relationships} and recognize that regulating these relationships is the key to mitigating blind conformity. In this paper, we propose a novel framework for \textbf{D}ynamically r\textbf{E}gulating deb\textbf{A}te \textbf{R}elationships (DEAR) from the group perspective. At first, DEAR quantifies consensus and divergence as \textit{group evidence} to capture the debate state. Then, DEAR operates through three stages: 1) What: perceiving group consultation tendency and uncertainty; 2) Who: introducing a Selection RL-Agent to dynamically select reference peers; and 3) How: adopting a Behavior RL-Agent to adaptively adjust generation behaviors. Notably, we formulate the execution of the two RL-Agents as a sequential decision-making process, jointly optimizing via multi-agent reinforcement learning. Extensive experiments demonstrate that DEAR achieves superior performance while significantly reducing token consumption.

Problem

• Multi-Agent Debate
• Blind Conformity
• Large Language Models
• Group Interaction
• Debate Relationships

Innovation

• Multi-Agent Debate
• Blind Conformity
• Debate Relationships
• Reinforcement Learning
• Group Evidence