Unified World Models: Coupling Video and Action Diffusion for Pretraining on Large Robotic Datasets Paper • 2504.02792 • Published Apr 3 • 4 • 2