University of Alberta's FAME Framework Tackles Catastrophic Forgetting in Continuous RL

University of Alberta researchers propose FAME, a principled continuous reinforcement learning framework with fast and meta learners that minimizes catastrophic forgetting via adaptive warm-start and knowledge consolidation, outperforming baselines on MinAtar, Atari, and Meta-World benchmarks.

FAME frameworkICRL 2026University of Alberta

0 likes · 16 min read

University of Alberta's FAME Framework Tackles Catastrophic Forgetting in Continuous RL