Preprint
Jul 2026
Before Agents Speak: Pre-hoc Failure Risk Inference in Multi-Agent Systems
HalluProp, a Propagation-aware Hallucination inference framework that estimates individual agent failures and emergent system-level hallucination risks before inter-agent interaction, and effectively complements post-hoc methods, highlighting the potential of pre-hoc risk inference for building more reliable multi-agent systems.
Shi Lin, Chenpei Wang, Peng Qian et al.
· 0 citations