HUGSIM

WA-JEPA: Rethinking the Video JEPA Paradigm for World-Action Modeling in Autonomous Driving