Asmov-1 robot policy 85% trained with 0.007% error rate, open source next week
andrew_n_carr · x · 2026-11-01
Andrew Carr announced that the Asmov-1 policy is 85% done training using Cartwheel's cross-embodiment, control, and scaling-laws dataset, sitting at a 0.007% error rate across all eval tracks. An open source release is planned for next week, including the finished RoboParty RP1 policy.
More from Embodied
- Firefighting robot survives 30 minutes inside a 1,000°C furnace and keeps working — LexiLove · 2026-11-01
- Neuralink to start vision implants within months, aiming to restore sight for the blind — DimaZeniuk · 2026-11-01
- Robotics practitioner: data is far from solved — zero-shot and real-hardware gaps remain — paigeinsf · 2026-11-01
- Archon Robotics, a Shanghai embodied-AI startup from UniAD and Huawei ADS veterans, wants to be China's OpenAI for humanoids — Scobleizer · 2026-11-01
- Why the pinky's position matters for robot dexterity: thumb as fulcrum — chris_j_paxton · 2026-11-01
- Princeton's Mengdi Wang to unveil LabOS, an AI-XR co-scientist that works in real labs — lihua_lei_stat · 2026-11-01