LingBot-Touch Tactile Tokens Boost Robot Policy Success by 14–25 Points
LingBot-Touch, trained on 5,000 hours of hard-to-crawl tactile data, lifts success rates of four robot policies by 14–25 points over raw signals, and its single encoder generalizes across magnetic, visuotactile, and piezoresistive sensors, cutting contact error from 5.89mm to 0.16mm on unseen hardware.
2026-11-11 ~ 2026-11-11 · 3 related posts
- Episode 1: Ant's Robbyant Open-Sources LingBot-Touch, a Tactile Foundation Model with Real-Robot Gains(2026-11-10, 27 posts)
- Episode 2: LingBot-Touch Tactile Tokens Boost Robot Policy Success by 14–25 Points(2026-11-11, 3 posts)
- Learned touch tokens add 14–25 points over raw signals across four policies — kimmonismus · 2026-11-11
- 5,000 hours of paid touch data: robotics' scarcest, unscrapeable training input — kimmonismus · 2026-11-11
- One touch encoder generalizes across sensors, cutting contact error to 0.16mm — kimmonismus · 2026-11-11