Xindong He

Robotic world-action models do not need pretrained internet video backbones to scale: training native planning and dynamics modules from scratch across 30,000 hours unlocks strong zero-shot manipulation and drives a 17.7-point gain on an embodiment comprising under 2% of the data.

Decomposing robotic manipulation into coarse and fine-grained actions isn't just conceptually cleaner—it actually unlocks a sweet spot where learning difficulty is balanced, boosting performance.