#AI SafetyStory
MemSyco-Bench: Benchmarking Sycophancy in Agent Memory
MemSyco-Bench is a new benchmark designed to evaluate memory-induced sycophancy in agent systems, addressing the issue where retrieved memories cause agents to over-align with users at the expense of factual accuracy. It includes five tasks to assess an agent's ability to handle memory appropriately.
Read original ↗Sign in to upvoteSign in to save