1 paper · 1 filter
Hongyu Wang, Jingquan Wang, Bocheng Zou +3
Video-based world models generate visually plausible rollouts, but since they infer dynamics in latent states, they enforce no explicit physical constraints: contacts drift, shapes…