1 paper
Yuntian Gao, Xiangyu Xu
Joint-Embedding Predictive Architectures (JEPAs), including recent LeWorldModel (LeWM), have become a promising foundation for reconstruction-free visual world models. For visual p…