Towards Hyperparameter-free Policy Selection for Offline Reinforcement Learning

Nov-14-2025, 09:18:58 GMT–Neural Information Processing Systems 

While polynomial-variance estimators exist, either using TD (e.g., Fitted-Q