Bayesian Flow Networks for Offline Trajectory Planning

Ludvig Killingberg, Helge Langseth

Offline reinforcement learning (RL) leverages static datasets to learn decision policies without real-time environment interaction. While recent sequence-modeling approaches rely o…