[论文] Understanding Communication Failures in Multi-Agent LLM Systems

论文概要

研究领域: ML 作者: Priya Pitre, Jun Hou, Yi Fang, Naren Ramakrishnan, Xuan Wang 发布时间: 2026-10-27 arXiv: 2610.32060

中文摘要

多agent LLM系统旨在通过辩论、证据共享和相互纠正来提高推理能力，但交互也可能引入最终准确率无法单独解释的故障。agent可能过早收敛、忽视异议、压制正确的少数派答案，或跨轮次保留错误的框架。我们提出了一个基于语料库的通信故障分类法，该分类法从医疗问答、编码、金融、多跳推理、伦理和具身文本任务中3,211条失败的多轮轨迹构建而成。

利用该分类法，我们分析了跨数据集和系统设计的故障分布、轮次级错误动态、恢复难度、多数转换、少数信号抑制和自我报告的信心。我们发现，故障主要由恢复/纠正和共识/决策问题主导，而非孤立的首轮错误，其中许多故障在agent首次观察到同伴响应后出现。基于这些发现，我们建议在最终确定之前加强异议处理、证据验证、少数信号保留和恢复检查。

原文摘要

Multi-agent LLM systems aim to improve reasoning through debate, evidence sharing, and mutual correction, but interaction can also introduce failures that final accuracy alone cannot explain. Agents may converge too early, ignore dissent, suppress correct minority answers, or preserve a wrong framing across rounds. We present a corpus-grounded taxonomy of communication failures, built from 3,211 failed multi-round trajectories across medical QA, coding, finance, multi-hop reasoning, ethics, and embodied text tasks. Using this taxonomy, we analyze failure distributions across datasets and system designs, round-level error dynamics, recovery hardness, majority transitions, minority-signal suppression, and self-reported confidence. We find that failures are dominated by recovery/correction an...

#论文 #arXiv #ML #小凯