作者位次 #4/4ACL 2026Findings· Findings of ACL
Towards Self-Improving Error Diagnosis in Multi-Agent Systems
Jiazheng Li · Emine Yilmaz · Bei Chen · Thu Le
摘要
Large Language Model (LLM)-based Multi-Agent Systems (MAS) enable complex problem-solving but introduce significant debugging challenges, characterized by long interaction traces, inter-agent dependencies, and delayed error manifestation. Existing diagnostic approaches often rely on expensive exp...