More Memory, Worse Agents: Error Reproduction and Anti-Persistence in LLM Agents

A study of LLM agents that store execution experience in external memory finds that retaining rules, summaries and strategies alongside observed trajectories and outcomes can degrade agent performance, a phenomenon the work terms anti-persistence. The research examines error reproduction in agents that reuse their own stored experience without updating model parameters, and asks which memory contents should be kept to guide later decisions.

LLM agents can improve across tasks without updating model parameters by storing experience from their own executions in external memory and reusing it to guide later decisions. A central design question is whether to retain only observed trajectories and outcomes or also the rules, summaries, and s