VL-JEPA: Joint Embedding Predictive Architecture for Vision-language Paper • 2512.10942 • Published 25 days ago • 35 • 3