LingBot-Touch gives robot touch a shared representation
Robbyant released model resources for learning from different types of tactile sensors.

Robots encounter touch through sensors that measure contact in different ways.
What happened
Robbyant released LingBot-Touch, a tactile representation model pretrained on nearly 5,000 hours of magnetic, piezoresistive and visual-tactile recordings.
The release provides code and model resources for turning different sensor signals into reusable representations.
Why it matters
Manipulation depends on information such as contact, pressure and movement at the point where a hand meets an object. A shared representation gives downstream systems a common way to process those signals.
The bigger picture
Reusable robot learning involves more than vision and language. Tactile models add a route for sharing experience across sensing hardware and manipulation tasks.
