TacClaw-Infinity: 10,000-Hour Vision-Tactile Dataset Lifts π0.5 Success from 50% to 74%
CyberRobooo · x · 2026-11-09
Daimon Robotics released TacClaw-Infinity, a large-scale Vision-Tactile-Language-Action (VTLA) dataset for dexterous manipulation, targeting the data gap in handling soft bags, rigid cartons and slippery bottles — vision tells robots what they're picking up, tactile feedback teaches them how.
- Scale: 10,000+ hours, 1,000+ task types, 1M+ trajectories, collected by 20+ institutions across 50+ sites.
- Collected with handheld tactile grippers independent of robot platforms, with contact-centered annotation.
- Across 3 robot embodiments and 2 policy models, adding tactile data raised π0.5's average success rate from 50% to 74%, with a max task-level gain of 44.7 points.
Part of Daimon's broader tactile stack: sensors, data infrastructure, and the Daimon-TWM world model.
Related event: TacClaw-Infinity: Massive Tactile Dataset Lifts Robot Grasping(2 posts)→
More from Embodied
- RoboMoCap dataset released: 200+ hours of 120Hz whole-body mocap for humanoid robot learning — DJiafei · 2026-11-09
- Tesla cuts Optimus AI6 chip memory to 144GB, a 33% reduction amid DRAM shortage — JOBhakdi · 2026-11-09
- Tesla expands Austin Robotaxi zone to roughly 3x Waymo's coverage area — Polymarket · 2026-11-09
- Waymo explains at Hot Chips why camera-only autonomy plateaus at 99.9% — 4310sy · 2026-11-09
- 'The future is here, just unevenly distributed': humanoid robot spotted at a conference — technollama · 2026-11-09
- OriginFlow CEO on physical AI data: real moats are systems engineering and economic efficiency, not data volume — 创业邦 · 2026-11-09