Fetching the paper…

DreamerPro: Reconstruction-Free Model-Based Reinforcement Learning with Prototypical Representations · Around