GR00T 2 deep dive: 40% zero-shot success on unseen tasks, human-to-robot transfer
DrJimFan · x · 2026-11-11
The GR00T 2 project page adds technical depth: a new generation of world-action models (WAMs) that predict future video and robot actions jointly, supporting parallel grippers, dexterous hands, and humanoids.
Highlights:
- Emergent X-Adaptability: skills shown only by humans transfer to robots — a two-five-fingered-hand robot lines a lab funnel in 65% of trials from human-only demos; YAM assembles a toy dump truck end-to-end in 60% of trials (up to 4 minutes); wearable DexUMI demos enable banana yogurt prep and match-striking
- Zero-/few-shot: 40% zero-shot success on 50 tasks never seen on the target robot; five demos and under 3 minutes of robot data per task substantially boost performance
Code available in December.
Related event: NVIDIA Open-Sources GR00T 2, a World-Action Model for Robots(18 posts)→
More from Embodied
- US robotaxi deployments are closing the gap with China, new data shows — skorusARK · 2026-11-11
- OpenTac debuts at CoRL as researchers call tactile the missing half of robot data — anand_bhattad · 2026-11-11
- Fully DIY in-wheel-motor autonomous RC car runs SLAM on a ¥2000 LiDAR — ___Mufasaa · 2026-11-11
- AHA 2026: AI reads smartwatch ECGs for heart failure and speeds echo interpretation — AmiBhattMD · 2026-11-11
- SpaceX moves Super Heavy booster to the launch pad at Starbase — DimaZeniuk · 2026-11-11
- SkeLAM Uses Skeleton-Grounded Latent Actions to Decouple Motion from Appearance in Robot World Models — jiqizhixin · 2026-11-11