Researchers introduce LingBot-Touch tactile model for robot manipulation

A paper posted to arXiv describes LingBot-Touch, a tactile foundation model built to give robots a single way to interpret touch data from different kinds of sensors. The authors say the model produces a unified representation across piezoresistive, magnetic and visuotactile sensors, and can be extended to new sensor types. The resulting tactile tokens are designed to plug into policy models without modification, from diffusion policies to vision-language-action models including π0.5 and LingBot-VLA. Experiments covered simulation and real-robot deployments on several hardware platforms, and the authors report gains on complex manipulation tasks with efficient computation. The abstract does not state whether model weights or code are being released.