D4RL benchmarks

New framework refines offline RL trajectories using counterfactual flows

Researchers have introduced a new framework called counterfactual transport flows for offline reinforcement learning. This method aims to improve decision-making policies using only logged historical data, without extra…