DeepMind Paper: Self-Evolving Agents Learn Fastest but Forget the Most

A DeepMind and Mila paper finds that self-evolving agents learn new tasks fastest but forget the most old knowledge, revealing plasticity at the cost of stability, and examines mitigation approaches such as weight freezing.

2026-10-21 ~ 2026-10-21 · 2 related posts

• DeepMind paper: self-evolving agents forget as they learn — fastest learner GEPA also breaks old tasks most — rohanpaul_ai · 2026-10-21
• arXiv: Are Self-Evolving Agents Continual Learners? Plasticity comes at the cost of stability — rohanpaul_ai · 2026-10-21