Berkeley's DexTacWAM adds multi-finger touch to world models, wins all 6 dexterous tasks
berkeley_ai · x · 2026-09-30
Researchers from UC Berkeley, UIUC and Northwestern present DexTacWAM, a visuo-tactile world-action model for dexterous manipulation.
- Core idea: treats touch as part of the predicted world state, extending a pretrained video world model to multi-finger tactile dynamics via lightweight tactile-encoder adaptation and continual vision-to-touch learning; a single forward pass feeds visuo-tactile features directly to the action expert.
- Compression: a finger- and pose-aware tactile compressor maps 10 fingertip streams into 2 hand-level latents, retaining 89.4% contact recall, yielding 2.26× faster training and 1.29× faster inference.
- Data efficiency: adapts with only 4 hours of tactile data, no video-model tactile midtraining required.
- Results: best on all 6 tasks with an average score of 70.6 vs 38.0 for the strongest baseline; demos include tool-mediated force control, bowl unstacking, bottle-cap unscrewing and two-hand wiping.
Code, weights and data are released on the project website.
More from Embodied
- Investor explains backing Figure AI at ~$2B: betting on $1T+ outcomes — markjeffrey · 2026-09-30
- Dyna Robotics ships Dyna-2.1, pivoting from task mastery to 'whole employee' robots paid by role — JasonMa2020 · 2026-09-30
- HG-DAgger Pitfall: Annealing With Clean Demonstrations Only Lifts Robot Task Success to 88% — DominiqueCAPaul · 2026-09-30
- Wayve supervised ride in San Francisco and highway completed with zero interventions — alexgkendall · 2026-09-30
- Raspberry Pi's $30 Smart Display Module Ships With Optional Hailo AI Acceleration — JeremyCMorgan · 2026-09-30
- Dyna Robotics' Taku masters new whole-body tasks from just 30 minutes of data — charlieholtz · 2026-09-30