Best policy hits only 29.1% on HandsOn; tool use drops to 2.9%, transfer near zero
stingning · x · 2026-10-27
On Core-50, the best multi-task checkpoint (OpenWAM) reaches just 29.1% task-macro success, with huge variance by hand skill. Pooled across eight trained checkpoints: grasp–transport–release 34.9%, using a held tool on another object 2.9%, precision fitting 2.2%. On Transfer-16, overall success ranges from 0.0% to 0.9%. The results locate a substantial gap in functional tool use, precise contacts, and task transfer, with stage diagnostics helping explain what contributes to it.
More from Embodied
- UniAction: unified action conditioning lets one world model handle mixed action types — ZeYanjie · 2026-10-27
- StackChan robot performs an AI-generated skit, orchestrated by Hermes Agent via JSON script — alexcovo_eth · 2026-10-27
- DONIC hits 86.79% on bimanual dexterous tracking, crushing prior SOTA's 9.43% — liuziwei7 · 2026-10-27
- Bedrock Robotics raises $600M as its fully autonomous excavators do paid work on live job sites — Sethwinterroth · 2026-10-27
- AW-0 debuts: one generative world action model plans in code, predicts, and controls robots — trevordarrell · 2026-10-27
- Skyfall AI's SKY-001 world model turns sketches into CAD builds in 3.9s, vs 199s for GPT-6 Astra — rohanpaul_ai · 2026-10-27