Ropedia W0: a world model learning loco-manipulation from 10k hours of human egocentric video
_akhaliq · x · 2026-10-22
Ropedia W0 is a world model that learns loco-manipulation from 10,000 hours of human egocentric video, then transfers across embodiments — from robot arms to humanoids, and from grippers to dexterous hands. It's a notable example of using large-scale human video to train generalist robot skills without heavy teleoperation data.
More from Embodied
- Open-source robotics harness reproduces Gemini zero-shot bimanual handover — berkeley_ai · 2026-10-22
- SCOUT Adapts Sim-Trained Robot Policies to the Real World in a Few Trials — davheld · 2026-10-22
- XDOF launches real-robot leaderboard benchmarking frontier models on 16 physical tasks — pabbeel · 2026-10-22
- OpenTac's shear-sensing trick: 3D-printed bumps that tilt under lateral force — mangahomanga · 2026-10-22
- OpenTac: JHU and MIT open-source an ecosystem for custom robot tactile skins sensing shear force — mangahomanga · 2026-10-22
- BiDex-100: a 100+ task simulation benchmark for bimanual dexterous manipulation from Zhejiang U & Ant Group — Robbyant-Research · 2026-10-22