# Ant Lingbo open-sources LingBot-Touch, a tactile foundation model for dexterous robot hands

> Original title: Token卷到触觉了！机器人终于有了通用手感

- Source: QbitAI
- Published: 2026-11-10T07:26:53.000Z
- Bad Influence: https://badinfluence.news/items/lk4dbuktku0bpkn4j20x5h8iz
- Original: https://www.qbitai.com/2026/11/508919.html

## Summary

Ant Lingbo has open-sourced LingBot-Touch, a tactile pretraining foundation model that translates signals from three sensor types into a shared token format for action models. In its tests, feeding raw tactile readings to pi0.5 lowered the zipper task's success rate from 39% to 15%, while adding LingBot-Touch raised it to 59%. The project page, technical report, code, and weights are published on GitHub, Hugging Face, and ModelScope.

## Why it matters

The article explains why raw tactile signals made robot grasping worse and how a translation layer fixes it, a useful reference for sensor-policy integration.