robot world model DreamDojo Platform for Scalable Robot Training DreamDojo is a generalist robot world model trained on 44,711 hours of human video. It learns interaction dynamics, enables zero-shot generalization, and supports model-based planning for scalable embodied AI.
humanoid robot learning VideoMimic: How Robots Learn Human Motion VideoMimic turns monocular human videos into deployable humanoid robot policies. By combining 4D reconstruction, scene geometry, and reinforcement learning, it enables context-aware robot control without motion capture or handcrafted rewards.
