Ropedia W0 shows 10k hours of human video yields a robot learning scaling law
liuziwei7 · x · 2026-10-20
The Ropedia team released W0 plus the open-source Xp-Teleop dataset, presenting a "Human Experience Scaling Law": with post-training fixed, scaling human video pre-training from 180 to 10k hours steadily improves the world model — camera-following ATE drops from 2.21cm to 1.79cm, hand-following PCK@01 rises from 82.0% to 83.6%, and gains transfer to robot prediction (GR-1 PSNR up to 21.74, Unitree G1 LPIPS down to 0.365).
- Data: 10k hours of human video + 3k hours of robot data with synchronized vision, motion and language
- Xp-Teleop: long-horizon teleoperation on Unitree G1 with Inspire hands — 27 episodes, 543 segments, 96k frames with RGB, keypoints, IMU, tactile and action annotations (CC-BY-4.0, on Hugging Face)
Related event: Ropedia Open-Sources W0 World Model Trained on 10k Hours of Human Video(5 posts)→
More from Embodied
- Surgeons Complete 100+ Cataract Surgeries Using Apple Vision Pro and ClearSphere App — asymco · 2026-10-20
- What makes human video generation good for robot manipulation? A constrained-optimization analysis — mangahomanga · 2026-10-20
- Figure's F.03 humanoid walks 348 miles and 1.2M steps over 9 straight days on a treadmill — adcock_brett · 2026-10-20
- NVIDIA joins Gemma hardware hackathon with 30 Jetson kits at AGI House on Oct 24 — agihouse_org · 2026-10-20
- Tesla shows Optimus V3 render in investor letter, Musk touts an 'Amazing Abundance' era — JOBhakdi · 2026-10-20
- UCSD PhD Nicklas Hansen joins NYU as postdoc with LeCun and Pinto on world models + robots — LerrelPinto · 2026-10-20