Berkeley Research Finds Periodic Memory Forgetting Boosts Agent Performance

A UC Berkeley, Stanford and Bespoke Labs paper argues that agents working on long-horizon tasks should periodically forget, as excessive memory causes optimization to stagnate; the approach improved results by up to 39%.

2026-11-01 ~ 2026-11-02 · 3 related posts

• Berkeley/Stanford paper: agents gain up to 39% by wiping memory and restarting from best result — alexcovo_eth · 2026-11-01
• Berkeley PhD on agent optimization: forgetting unlocks stalled discovery, GEPA successor extends to code and kernels — CShorten30 · 2026-11-02

1 near-duplicate retellings: CShorten30