1 paper
Tianheng Wang, Zhou Xie, Heng Jia +3
Generative visual models offer a foundation for learning representations of physical dynamics, yet their extension to continuous control raises a fundamental question: do visual pr…