ISEP: Implicit Support Expansion for Offline Reinforcement Learning via Stochastic Policy Optimization

In three linesISEP proposes an offline reinforcement learning method that implicitly expands action support by interpolating between in-distribution data and policy samples. A stochastic mechanism alternates between conservative cloning and optimistic expansion signals, implemented via Conditional Flow Matching with classifier-free guidance.

Summary generated by Claude — human-verified