Objects matter: object-centric world models improve reinforcement learning in visually complex environments figure
AlphaXiv 中文概览(可滚动查看)