1 paper
Zhekai Wang, Haoxiang Huang, Xiang Liu +6
Object-centric world models forecast future videos by evolving a set of entity slots, but the variables receiving dynamics supervision are often unconstrained visual features. We i…