HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation
stingning · x · 2026-10-27
HandsOn is a new simulation platform and benchmark for generalizable dexterous manipulation, with 66 human-verified tasks (50 core + 16 transfer), 18,000 curated human demonstrations, and 56 hours of data across tool use, precise interaction, articulated objects, and bimanual coordination. It evaluates diffusion policies, VLAs, world-action models, and LLMs, analyzing failure modes like contact establishment and goal maintenance. Core-50 success rates: OpenWAM leads at 29.1%, GPT-6 (moveeef) trails at 5.2%. Paper, code, and dataset are public.
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