arXiv cs.MA· Hao Wu, Shoucheng Song, Chang Yao, Huaiyu Wan, Youfang Lin, Kai Lv·· 13 小时前AI 评分24
从群体视角调节辩论关系：DEAR 框架缓解多智能体辩论中的盲从
Group Perspective Matters: Regulating Debate Relationships Can Mitigate Blind Conformity in Multi-Agent Debate
AI 导读
针对多智能体辩论（MAD）中 LLM 易出现盲从的问题，研究者提出 DEAR 框架，从群体视角动态调节 LLM 间的辩论关系。DEAR 将共识与分歧量化为群体证据，并通过 Selection RL-Agent 选择参考对象、Behavior RL-Agent 调整生成行为，二者以多智能体强化学习联合优化。
正文 · 原文
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.
| Comments: | NeurIPS 2026 Accept |
| Subjects: | Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2608.03648 [cs.MA] |
| (or arXiv:2608.03648v2 [cs.MA] for this version) | |
| https://doi.org/10.48550/arXiv.2608.03648 arXiv-issued DOI via DataCite |
Submission history
From: Hao Wu [view email]
[v1]
Tue, 4 Aug 2026 13:31:24 UTC (2,298 KB)
[v2]
Wed, 21 Oct 2026 13:52:26 UTC (2,311 KB)
来源：arXiv cs.MA · arxiv.org