AI CURATED HOT STORY · 11/10/2026
Ant Group's Lingbo Technology has open-sourced LingBot-Touch, a general-purpose tactile foundation model pre-trained on nearly 5,000 hours of real-world touch data. It unifies three heterogeneous sensing modalities—magnetic, visuotactile, and piezoresistive—into a single tactile representation. In 26 real-robot tests, integrating LingBot-Touch boosted the average task success rate of four action models (LingBot-VLA, π0.5, Diffusion Policy, and T-Rex) by 14 to 26 percentage points, with the bottle-cap twisting task jumping from 56.25% to 90%.
The model learns a reusable tactile representation that generalizes across sensors, dexterous hands, models, and tasks, and performance continues to improve with data scale. The model, code, and technical report are available on Hugging Face, GitHub, and ModelScope, providing a reusable tactile perception foundation for embodied AI research.
This open-source release extends embodied intelligence from vision to touch, filling the gap in general tactile foundation models. It is expected to lower the barrier to tactile perception R&D and accelerate iteration in dexterous manipulation and humanoid robotics, marking a significant step for open collaboration in the embodied AI ecosystem.