CoLFA Framework for Multi-Agent Failure Attribution
November 13, 2026
CoLFA uses collaboration cues to trace errors in multi-agent systems by organizing inputs based on how results are formed. This method aims to balance localization accuracy with the high token costs typically required to inspect lengthy agent execution records.
HOW THIS AFFECTS YOU
●
builderYou can debug multi-agent workflows more efficiently without paying for massive context windows.
●
researcherYou can use collaboration history to more accurately identify which agent in a chain caused a failure.