LingBot-Touch lifts VLA success rate from 45.6% to 72.5%, pretrained on 4,983 hours of tactile data
CyberRobooo · x · 2026-11-10
Key results for LingBot-Touch: on Sharpa Wave with LingBot-VLA, holding everything else fixed, mean success rate went 45.60% (no touch) → 56.29% (raw touch) → 72.51% (LingBot-Touch).
- Pretrained on 4,983 hours of real tactile recordings
- Tested on 26 real-world tasks with two dexterous hands on dual-arm setups (Sharpa Wave and XHand1): twisting caps, inserting straps, closing zippers, moving simulated eggs
- The 5.46M-parameter LingBot-Touch-Tiny encoder adds only 0.13% to LingBot-VLA's parameter count
- With the encoder frozen and no sensor-specific adaptation, it transferred to an unseen XHand1 magnetic sensor setup
More from Embodied
- Claude outperforms specialist robot-control AI on RoboDojo two-arm benchmark — burny_tech · 2026-11-10
- How LingBot-Touch works: one shared Haptic Transformer unifies magnetic, piezoresistive and visuotactile signals — CyberRobooo · 2026-11-10
- Robbyant releases LingBot-Touch, a tactile model giving robot hands human-like touch sensing — CyberRobooo · 2026-11-10
- Dyna Robotics shows RL vs IK head-to-head on wheeled robot, unveils workflow agent Dyna-2.1 — chris_j_paxton · 2026-11-10
- AI researcher Chris Paxton comes away positive from a Tesla Cybercab ride — chris_j_paxton · 2026-11-10
- Two CoRL papers speed up VLAs 2.4x and 4.7x, fast enough for real-time robot control — songhan_mit · 2026-11-10