Liber-0 hits 36% zero-shot success across 60 robot manipulation tasks without task-specific demos
heyshrutimishra · x · 2026-11-04
Robotics startup Liber AI introduced Liber-0, a World Action Model for zero-shot robot manipulation including dexterous, multi-step tasks.
- Across 60 tasks proposed by GPT-6 and 600 trials, it achieved 36% overall success without robot demonstrations collected for those tasks: pick-and-place 63.5%, other manipulation 34.0%, challenging multi-step tasks 10.5%.
- The model was pretrained on 2.2M+ hours of human experience (still including robot data), using glove-recorded motion and embodiment alignment to turn human experience into robot skills by jointly predicting future observations and actions.
- If this transfer scales, new capabilities could require far fewer task-specific robot demonstrations.
Related event: Liber AI Releases Liber-0 World Action Model for Zero-shot Robotics(3 posts)→
More from Embodied
- This robot runs an entire 3D printing farm unattended, print after print — lukas_m_ziegler · 2026-11-04
- Tsinghua & NVIDIA propose PrismWAM, splitting semantics, physics and actions in world models — jiqizhixin · 2026-11-04
- 30 days, six teams: lessons from delivering a full humanoid robot system on a tight timeline — yongqianme · 2026-11-04
- SOMA, the Claude Code CAD app, passes 1,000 users with a 9-minute bookshelf design walkthrough — Pr8tham · 2026-11-04
- Nano-texture iPhone Duo display makes Apple Pencil feel like writing on paper — rudrank · 2026-11-04
- NEURA launches Neuron AI, but Europe's Physical AI test is whether the stack matures together — ingliguori · 2026-11-04