New CoLFA Method Targets Failure Attribution in LLM Multi-Agent Systems
A paper introduces CoLFA, a collaboration-guided lightweight approach for attributing failures in LLM-based multi-agent systems by tracing error formation to the decisive step and responsible agent.
Read the full story at arXiv cs.CL ↗
Timeline · 1 report
- 2026-11-13 05:00 · arXiv cs.CL
CoLFA: Collaboration-Guided Lightweight Failure Attribution for LLM-Based Multi-Agent Systems