1
VerifyMAS: Hypothesis Verification for Failure Attribution in LLM Multi-Agent Systems
提出假设验证方法定位LLM多智能体系统故障原因，提升系统可靠性
arXiv:2605.17467v1 Announce Type: new Abstract: Large language model-driven multi-agent systems (LLM-MAS) excel at complex tasks, yet unreliable agent…