Light-O1: Human-Action Pretraining Lets One Robot Brain Drive Different Humanoids
haukejung · x · 2026-09-21
Origins introduces Light-O1, a whole-body intelligence model pretrained on structured human actions recovered from internet videos. Scaling human-action pretraining yields power-law error reductions across embodiments, letting the same brain coordinate movement, posture and manipulation on different humanoids to wipe tables, pick up trash and tidy shoes on command.
Related event: Light-O1: Humanoid Robot Brain Pretrained on Human Videos(2 posts)→
More from Embodied
- Phone video in, navigable 3D room out: K3-powered splat pipeline runs in the browser — willeastcott · 2026-09-21
- World's first human vs. humanoid robot fight held, a 6ft robot steps into the ring — Admirable-Cell-2658 · 2026-09-21
- Sim training in 4.2 GPU-hours teaches robotic hand pen-spinning and cube solving — FinanceYF5 · 2026-09-21
- RoboHarm robot safety benchmark: GPT-6 Astra attempts dangerous acts in 97% of tests — 量子位 · 2026-09-21
- Overlaying object tracking on OpenStreetMap with GNSS/IMU trajectories and depth estimation — rsasaki0109 · 2026-09-21
- Asking for the best way to run MiniMax H3 video generation on a DGX Spark — Shady-Dragon · 2026-09-21