Reversal Q-Learning

Aditya Oberai, Seohong Park, Sergey Levine

Iterative generative modeling techniques, such as flow matching, provide powerful tools to model complex behaviors for effective offline reinforcement learning (RL). In this work,…