arXiv cs.AI· Dezheng Bao, Sihang Du, Renhong Huang, Zhizhang Yuan, Yifei Sun, Yang Yang·· 3 天前AI 评分24
TrueMAD：面向多智能体辩论中定向同伴互动的新方法
Beyond Answer Sharing: TrueMAD for Targeted Peer Engagement in Multi-Agent Debate
AI 导读
针对标准广播式多智能体辩论（MAD）中智能体无需回应特定同伴观点的问题，研究者提出 TrueMAD——一种与骨干模型无关的测试时插件，在答案广播与自我修订之间引入一对一、自由形式的同伴建议，源智能体通过支持、反驳、改进或追问回应目标智能体的当前推理。
正文 · 原文
Abstract:Multi-agent debate (MAD) aims to improve large language model reasoning by exposing agents to alternative solutions and enabling iterative revision. However, standard broadcast-style MAD makes peer answers visible without requiring agents to engage with a particular peer claim. As a result, reliable reasoning may be suppressed by conflicting or majority responses, limiting achievable performance, while erroneous reasoning may persist and degrade final predictions without actionable feedback. We propose TrueMAD, a backbone-agnostic test-time plug-in that introduces one-to-one, free-form peer advice between answer broadcasting and self-revision. Each source agent responds to a target's current reasoning through support, rebuttal, refinement, or further questions, while the target retains control over how to revise. The same mechanism can augment standard MAD and its improved variants without parameter updates. Across AIME, GPQA-Diamond, and TruthfulQA with Qwen3-30B-A3B, TrueMAD improves agent-level and strict-majority accuracy over MAD by 3.62 and 3.52 percentage points on average, while TrueDMAD improves DMAD by 4.72 and 4.06 points, respectively. Analyses at the utterance, agent-transition, and group levels further show more peer-specific deliberation, more correction-oriented revisions, and more favorable consensus dynamics.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.18240 [cs.AI] |
| (or arXiv:2610.18240v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.18240 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Dezheng Bao [view email]
[v1]
Tue, 13 Oct 2026 12:08:38 UTC (395 KB)
来源：arXiv cs.AI · arxiv.org