Offline Reinforcement Learning with Differentiable Function Approximation is Provably Efficient

Open in new window