1 paper
Yucen Wang, Fengming Zhang, De-Chuan Zhan +3
Adapting pretrained video generation models into controllable world models via latent actions is a promising step towards creating generalist world models. The dominant paradigm ad…