按拓扑诊断多智能体 LLM 协作失败
原标题：Know the Shape, Find the Fault: Topology-Conditioned Diagnosis of Multi-Agent LLM Failures
AI 摘要
该论文提出 MAScope 两阶段框架，用于诊断多智能体 LLM 系统的协作失败。第一阶段 TSE 从异构执行轨迹中恢复通信拓扑，第二阶段 TC-Judge 结合轨迹、预测拓扑、经验失败先验和拓扑特定失败模式进行分类。实验表明通信拓扑与失败类型存在统计显著关联，拓扑条件上下文可将 gpt-mini 的 Macro-F1 显著提升，并在固定编排结构下以约更低成本接近 gpt-5.4 基线。
正文节选
Know the Shape, Find the Fault: Topology-Conditioned Diagnosis of Multi-Agent LLM Failures Abstract. Multi-agent LLM systems coordinate task execution through exchanges of information among agents. When coordination breaks down, similar symptoms in execution traces can reflect different problems in how information is passed, used, or verified. Communication topology captures how agents exchange information and provides structural cues for distinguishing coordination failure modes. Using these cu