نسخة أولية وصول مفتوح
What 30,000 Hours of Ego-centric Video Does Not Teach
World models offer a promising alternative to physics-based simulators, yet remain far from practical deployment. We ask how far scaling ego-centric human video takes them, using a dataset of 30,000 hours spanning over 1,000 scene types and 14,000 contributors. Rather than relying on opaque downstream metrics, we direc …