Ant Group's Lingbo Open-Sources LingBot-Touch Tactile Foundation Model

On November 10, Ant Lingbo Technology open-sourced LingBot-Touch, a tactile foundation model, filling in the critical tactile perception capability for robot brains. The model is pretrained on nearly 5,000 hours of real tactile data, covering three mainstream sensing principles: magnetic tactile, visuotactile, and piezoresistive. It can convert different signals into a unified tactile representation for use by different dexterous hands and action models, enabling robots to "see clearly and think clearly" while further completing the tactile perception required for fine manipulation.

Hugging Face: https://huggingface.co/collections/robbyant/lingbot-touch

GitHub: https://github.com/Robbyant/lingbot-touch

Technical report: https://technology.robbyant.com/lingbot-touch

At a time when general-purpose large models are accelerating, choosing to release a tactile foundation model oriented toward the physical manipulation loop conveys Ant Lingbo's continued investment in and commitment to the embodied-native route.

Previously, Ant Lingbo's general-purpose brain already covered high-precision spatial perception, highly dynamic environment interaction, and generalizable dexterous manipulation, enabling robots to see more clearly, think more clearly, and work more efficiently. LingBot-Touch has now been integrated into LingBot-VLA, extending perception capability to the "fingertips" so robots can "feel more sensitively," further enhancing the capabilities of the embodied-native general-purpose brain.

26 Real-Robot Tests: Task Success Rates Improve by 14 to 26 Percentage Points

As general-purpose large models continue to advance, robots' ability to understand complex tasks and reason about manipulation steps is strengthening. However, in the real physical world, every time a robot executes an action, it must perceive changes in the object and environment and then correct the next action. In many fine manipulation tasks, the robot can see where the object is and can move its hand to the correct position, yet it may not successfully complete the task, because after the fingers touch the object, whether the force is appropriate and whether the object slips are difficult to judge from vision alone and require tactile supplementation.

Take tightening a bottle cap with a dexterous hand as an example. Depth information can help the fingers align with the cap. Once rotation begins, the cap may already be loose, or the fingertip pads may be slipping along the cap wall. These two situations look similar in the image but require different action adjustments. At this point, the contact position at the fingertips, the magnitude of force, and their changes become important bases for the action model's judgment, yet such information is often occluded by the fingers and the object and is difficult for vision to obtain.

To verify the help of tactile representations for real manipulation, Ant Lingbo conducted 26 real-robot tests on two dexterous hands, Sharpa Wave and XHand1, covering fine manipulation tasks such as turning book pages, tightening bottle caps, and recognizing mahjong tiles. After integrating LingBot-Touch, the average task success rates of four action models—LingBot-VLA, π0.5, Diffusion Policy, and T-Rex—improved by about 14 to 26 percentage points compared with the no-tactile setting.

Taking T-Rex controlling Sharpa Wave to tighten a bottle cap as an example, the task success rate was 56.25% without touch, 62.5% with raw sensor signals directly input, and 90% after integrating the LingBot-Touch tactile representation.

(Caption: After integrating LingBot-Touch, the real-robot task success rates of the four action models on both dexterous hands all reached their respective highest levels.)

Unified Tactile Representation Enables Cross-Hardware Reuse of Robot "Touch"

Equipping a robot with tactile sensors does not mean it can use touch well. At present, commercial tactile sensors are still at a relatively early stage of development overall. Mainstream routes such as visuotactile, magnetic tactile, and piezoresistive each have advantages and limitations, and multiple technical routes may coexist for a long time for different costs and use cases.

More challenging is that the data collected by these sensors is difficult to use directly across systems. The same contact, under different sensing principles and contact layouts, produces different raw signals. If a model learns to interpret only one of them, then after changing the sensor or dexterous hand, existing data and tactile experience may be difficult to reuse, requiring new data collection, training, and adaptation. Amid rapid hardware iteration, this binding causes redundant development and increases adaptation costs among hardware vendors, full-system teams, and model developers.

Therefore, the key to scaling tactile technology is to free tactile understanding from dependence on the physical signals of specific sensors, so that data from different sources can be used together for training and accumulated tactile capabilities can be reused across sensors, dexterous hands, models, and tasks.

To this end, LingBot-Touch sets up separate encoding entrances according to sensing principles, first processing different raw signals, then mapping contact points of different layouts into a unified fingertip coordinate space, where a shared model learns contact position, force, and their changes over time and outputs unified tactile tokens. As a result, data collected by humans wearing tactile gloves and by teleoperated dexterous hands can be used together for training, and different action models can also use the trained model to extract contact features. This architecture also leaves room for extension to new types of tactile sensors in the future.

(Caption: LingBot-Touch encodes magnetic tactile, visuotactile, and piezoresistive tactile signals into unified tactile tokens for use by different action models.)

On XHand1, the team further verified cross-hardware reuse of tactile capability. The magnetic tactile sensor model used by XHand1 was not in the pretraining data; all the magnetic tactile experience learned by the model came from data collected by humans wearing gloves. After switching to this dexterous hand, the team kept LingBot-Touch parameters unchanged and trained the action model only with XHand manipulation demonstrations.

At the same time, LingBot-Touch was integrated with four different action models in this real-robot test and is not bound to Ant Lingbo's self-developed models. It provides general perception capability between tactile sensors and action models, rather than a dedicated module for a particular hardware or policy. For the industry, sensor and dexterous hand vendors can choose hardware based on cost and task requirements, and full-system teams can retain their own action policies, thereby reducing the repeated investment of redeveloping tactile capability every time a hardware and algorithm combination is changed.

To balance perception capability and on-device compute, Ant Lingbo offers models in two sizes. All 26 real-robot tests this time were completed using the lightweight LingBot-Touch-Tiny with approximately 5.3 million parameters. After integration with LingBot-VLA, π0.5, and T-Rex, the parameter count increased by only about 0.13% to 0.16%.

The larger version, LingBot-Touch-Small, showed potential for further improvement in perception tests—when pretraining data increased to nearly 5,000 hours, the model still benefited from the additional data, indicating that tactile pretraining still has room for continued scaling and improvement.

Committed to Embodied-Native, Advancing in Tandem with General-Purpose Large Models

Regarding recent evolution in the technical路线 of embodied large models, Ant Lingbo CEO Zhu Xing stated: "Progress in digital-world large language models represented by GPT-6 Astra and Claude Fable 5.2 has indeed made robots' complex task reasoning and physical manipulation capabilities stronger. We are paying close attention to this and actively absorbing it. But frontline industry practice tells us that to truly close the manipulation loop, the hard nut of physical manipulation generalization remains unavoidable, and to a large extent still depends on breakthroughs in embodied foundation models themselves."

In his view, general-purpose large models and embodied foundation models are closer to a 1+1>2 collaborative relationship: the former enhances understanding and reasoning about tasks and environments, while the latter improves physical perception and closed-loop manipulation capabilities. Each achieves its own breakthroughs and promotes the other, jointly accelerating the maturation of the embodied intelligence industry.

"The quality of modality input largely determines the quality of action output—in other words, no good input means no good output," said Zhu Xing. Ant Lingbo will adhere to the embodied-native路线, continue exploring new upper limits of physical manipulation generalization, build a universal brain for the robotics industry across configurations, promote the alignment of more physical modalities—including vision, depth, touch, and even temperature—within a unified model, and continue to open-source and open up to the community while advancing commercialization.

Currently, the technical report, model, and code for LingBot-Touch have been open-sourced, and developers can access and download them on Hugging Face, GitHub, and the ModelScope community.