1 paper · 1 filter
Bingjie Li, Yumeng Song, Zhongming Yao +1
Failures in agentic AI systems can arise from interactions among messages exchanged by multiple large language model (LLM) agents. Pointwise attribution cannot distinguish a jointl…