ResiliBench: Evaluating Agentic Workflow Adaptation in Stochastic Environments
Published in International Conference on Learning Representations (ICLR), 2026
ResiliBench is a benchmark designed to evaluate how well agentic workflows adapt in stochastic environments. The benchmark provides a systematic framework for assessing the resilience and adaptability of AI agent systems.
Authors: Ruicheng Ao, Zeping Min, Tingyu Zhu, Wotao Yin, Xinshang Wang (co-first author)