Robbyant open-sources LingBot-Touch, a tactile foundation model trained on 4,983 hours of data
rohanpaul_ai · x · 2026-11-10
Robbyant released LingBot-Touch (Apache 2.0), an open-source tactile foundation model giving robots a sense of touch that cameras can't provide.
- Architecture: JEPA-based representation model with a shared Haptic Transformer and modality-specific tokenizers, unifying piezoresistive, magnetic and visuotactile signals.
- Scale: pretrained on 4,983 hours / 1.33B frames across 464K episodes; the smallest HaT-Tiny encoder is just 5.46M params and supports high-frequency inference.
- Transfer: generalizes to unseen tactile sensors and real dexterous hands; evaluated on 26 real + 12 sim tasks with continued gains as data scales.
- Project page, tech report, paper, GitHub and Hugging Face all available.
More from Embodied
- Binary-reward RL hits 94% success on hard robotics task, recipe looks clear — kastnerkyle · 2026-11-11
- UMA team at CoRL2026 to discuss latent world models and sim2real for humanoids — RemiCadene · 2026-11-11
- World's first fighting teleop: S.T.E.E.L. brings VR and mocap-driven real-time sparring — cixliv · 2026-11-11
- EEG can't match implanted BCIs: physics caps scalp signals at ~70Hz — EricMTrautmann · 2026-11-11
- Dyna Robotics' CoRL 2026 demo blends large-scale navigation with dexterous manipulation — JasonMa2020 · 2026-11-11
- MBZUAI releases PAN 2, an open-weights foundation world model for physical intelligence — Scobleizer · 2026-11-11