1 paper
Zihan Liu, Yuzhe Zhuang, Yuanzu Li +4
JEPA-style visual world models offer an effective paradigm for visual goal planning by predicting future latent representations. Existing methods typically learn local transition c…