LingBot-Touch: open-source tactile foundation model pretrained on ~5,000 hours of real data
HeyAmit_ · x · 2026-11-10
- robbyantbrain released LingBot-Touch, an open-source tactile foundation model for dexterous manipulation, pretrained on nearly 5,000 hours of real tactile data, with unified heterogeneous tactile modeling enabling transfer across sensors, hands, tasks and action models.
- Reposter HeyAmit highlights a striking detail: XHand1's magnetic sensor model wasn't in the pretraining data, yet the frozen LingBot-Touch encoder worked without any sensor-specific adapter—only robot policies needed post-training—showing real generalization in its tactile representations.
Related event: Ant's Robbyant Open-Sources Tactile Foundation Model LingBot-Touch(13 posts)→
More from Embodied
- Amazon staffs up home robot team: field support engineers for Fauna's Sprout humanoid — xmercury_one · 2026-11-10
- Stanford BEHAVIOR Challenge: robot full-task success jumps from 12% to 48% in one year — yining_hong · 2026-11-10
- Amazon FAR hiring researchers and PhD interns in SF for human-level open-world navigation — iamfakhrealam · 2026-11-10
- Apple is building a world-class robotics team, recruiting at CoRL — iamfakhrealam · 2026-11-10
- Using open models to reverse-engineer a Surface Pro Linux kernel build, almost done — vincent_koc · 2026-11-10
- Robotics-at-Scale Pipeline Chart Sparks Debate: Almost All Bitter Lesson — KyleMorgenstein · 2026-11-10