蚂蚁灵波科技
Embodied AI unit of Ant Group developing the LingBot family of robot foundation models spanning spatial perception, interaction
Ant Group's Lingbo Technology has open-sourced LingBot-Touch, a tactile foundation model pre-trained on nearly 5,000 hours of real-world tactile data that unifies magnetic, visuotactile, and piezoresistive sensing signals. In 26 real-robot tests, integrating the model boosted average task success rates of four action models by 14 to 26 percentage points, extending embodied AI perception from vision to touch.
Ant Group's Lingbo Technology has open-sourced LingBot-Touch, a tactile foundation model pretrained on nearly 5,000 hours of real-world tactile data. It unifies signals from magnetic, vision-based and piezoresistive sensors into a single tactile representation reusable across sensors, dexterous hands, models and tasks. Across 26 real-robot tests, plugging in the representation lifted average task success for four action models by roughly 14 to 26 percentage points, with bottle-cap twisting rising from 56.25% to 90%.