[论文] Do We Need Complex Topology Control? Distinct-Peer Random Routing Impr...

论文概要

研究领域: ML 作者: Boxuan Wang, Zhuoyun Li, Xiaowei Huang, Yi Dong 发布时间: 2026-09-25 arXiv: 2609.27150

中文摘要

多智能体辩论（MAD）作为通过迭代同伴交互提升大语言模型推理准确性的范式已展现潜力。通信拓扑扮演核心角色，催生出越来越精巧的机制——学习、自适应或动态重构智能体交互以提升准确率或可靠性。与此同时，先前研究表明简单得多的稀疏通信已能以低得多的成本取得有竞争力的性能。本文细察稀疏 MAD，追问复杂拓扑控制是否真是改进集体推理所必需。我们发现一个简单的不放回随机路由策略——每轮让每个智能体与两个不同的、新采样到的同伴辩论——提供了出人意料的强基线，持续改进稀疏 MAD 的准确率-成本权衡。在此观察上我们进一步研究审议停止机制，表明轻量级停止可在保持有竞争力准确率的同时大幅降低推理成本。结果提示：在为其额外复杂性辩护之前，精巧的拓扑控制（如学习得到的拓扑自适应）应先与强大的简单路由与停止基线相比较。

原文摘要

Multi-agent debate (MAD) has emerged as a promising paradigm for improving the reasoning accuracy of large language models (LLMs) through iterative peer interaction. Communication topology plays a central role in this process, motivating increasingly sophisticated mechanisms that learn, adapt, or dynamically reconfigure agent interactions to improve accuracy or reasoning reliability. Meanwhile, prior studies suggest that much simpler sparse communication can already achieve competitive performance at substantially lower cost. In this work, we take a closer look at sparse MAD and ask whether complex topology control is actually necessary to improve collective reasoning. We find that a simple random-without-replacement routing policy, which lets each agent debate with two distinct and newly ...

#论文 #arXiv #ML #小凯