机构何时胜过智能？多智能体集体推理失败的结构性诊断

AI 摘要

该研究探讨了多智能体系统中，更强的智能体能力并不必然带来更强的集体表现，而机构结构（如信息路由、验证、状态维护等）在修复集体推理失败中起关键作用。通过构建四类失败场景（访问与路由、接纳与依赖、维护与激励、表征与行动）的人工生态，实验发现机构干预仅在修复公共状态构建失败时有效，而当信号无信息量、智能体可直接完成转换或状态无法支持行动时，其优势消失。研究将智能与机构的选择重新定义为对集体推理失败源的诊断。

正文节选

When Do Institutions Beat Intelligence?CCS: Computing methodologies Multi-agent systems Abstract. More capable agents do not necessarily form a more capable collective. A multi-agent system may jointly possess sufficient information yet fail because evidence is poorly routed, unreliable reports enter public belief, correlated claims masquerade as independent support, shared state becomes stale or strategically distorted, or useful evidence is exposed through an ineffective action interface. We a

发布时间：2026-08-13 12:00

抓取时间：2026-08-13 12:24

来源机构：arXiv