# Tsinghua team proposes SimpleTouch: fine manipulation learned from 50 demonstrations without tactile policy pretraining

- Source: Xbotics具身智能实验室 · 微信公众号
- Published: 2026-10-27T03:00:00.000Z
- RoboHot: https://robohot.news/en/items/ta3qv7g8eijisxk0r4ty6gv9c
- Original: https://mp.weixin.qq.com/s?__biz=MzkyMTc1NTI3Ng==&mid=2247509513&idx=1&sn=25d5a7219414656c2b42bc84f1ce131a&chksm=c00d27d77951394842c5be1d02fc72b230ad2a8718971698fff718d1d1c8cf40a6e153c34ab6&scene=126&sessionid=1793099785#rd

## Summary

A Tsinghua University research team proposes SimpleTouch, which adds a tactile expert to the pretrained VLA π0.5 and uses single-stage training with only 50 demonstrations per task, without additional tactile policy pretraining or a separate alignment stage.