Here, we focus ourattention onthesecond case, which leads toscore improvement without anyre-training.

We propose RLlib Flow, a hybrid actor-dataflow programming model for distributed RL, and validate its practicality by porting the full suite of algorithms in RLlib, a widely adopted distributed RL library.

In offline RL, a critical challenge is distribution shift (also called "extrapolation error" in literature).

Generalization to new environments remains a major challenge in deep reinforcement learning (RL). Current methods fail to generalize to unseen environments even when trained on similar settings [19, 51, 71, 11, 21, 12, 60].

Theseoutcome-conditioned imitationlearningmethodsare appealing because of their simplicity, strong performance, and close ties with supervisedlearning.

This is challenging from an implementation perspective, as repeatedly differentiating policy gradient estimates may lead to biased Hessian estimates.