Ant Lingbo Open-Sources Tactile Foundation Model LingBot-Touch, Improving Real-Robot Task Success Rate by 14 to 26 Percentage Points

On November 10, the same day Claude Fable 5.2 was released, Ant Lingbo Technology open-sourced the tactile foundation model LingBot-Touch. One is a general-purpose large model for the digital world, the other an embodied foundation model for the physical manipulation loop; both entered public view on the same day.

The relationship between these two technical routes is precisely the focus of discussion in the embodied intelligence industry today. In Ant Lingbo's 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 fills in physical perception and closed-loop manipulation capabilities. Only by advancing both in tandem can the embodied industry mature faster.

The open-sourced LingBot-Touch 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 single unified tactile representation for use by different dexterous hands and action models, allowing robots to "see clearly and think clearly" while further filling in the tactile perception needed for fine manipulation.

Previously, Ant Lingbo's general-purpose brain already covered high-precision spatial perception, high-dynamic environment interaction, and generalizable dexterous manipulation. Now LingBot-Touch has been integrated into LingBot-VLA, extending perception to the "fingertips" so robots can "feel more sensitively," taking the hand-eye coordination of the embodied native general-purpose brain to the next level.

26 real-robot tests: task success rates improved 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. But in the real physical world, every step a robot executes requires perceiving changes in the object and environment, then correcting 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 may still fail to complete the task—once the fingers touch the object, whether the force is appropriate and whether the object is slipping are difficult to judge by vision alone, and tactile sensing is needed to fill the gap.

Take unscrewing 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 fingertips may be slipping along the cap wall. These two situations look similar in the image but require different action adjustments. The contact position at the fingertip, the magnitude of force and its changes are often occluded by the fingers and the object and are difficult for vision to capture, yet they are precisely important judgment cues for action models.

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, unscrewing bottle caps, and identifying 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 no tactile sensing. Taking T-Rex controlling Sharpa Wave to unscrew a bottle cap as an example, the task success rate was 56.25% without tactile sensing, 62.5% with raw sensor signals fed directly, and reached 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 lets "feel" be reused across hardware, models, and tasks

Equipping a robot with tactile sensors does not mean it can use touch well. Currently, commercial tactile sensors as a whole are still at an early stage of development; mainstream routes such as visuotactile, magnetic tactile, and piezoresistive each have strengths and limitations, and multiple technical routes may coexist for a long time for different costs and use scenarios. More troublesome is that the data collected by these sensors is hard to use directly across systems: the same contact produces different raw signals under different sensing principles and contact layouts. Once a model learns to interpret only one of them, after changing sensors or dexterous hands, the existing data and tactile experience may be difficult to reuse, requiring new data collection, training, and adaptation.

Therefore, the key to scaling tactile sensing into real-world deployment 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 directly call the trained model to extract contact features. This architecture also leaves room for expansion to integrate new 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 validated the cross-hardware reuse of tactile capabilities: the magnetic tactile sensor model used by XHand1 was not in the pretraining data, and all the magnetic tactile experience the model learned came from data collected by humans wearing gloves. After switching to this dexterous hand, the team kept the LingBot-Touch parameters unchanged and only needed to train the action model with XHand's manipulation demonstrations to complete the transfer. At the same time, in this real-robot test, LingBot-Touch was connected to four different action models and is not tied to Ant Lingbo's self-developed models; what it provides is a general perception capability situated between tactile sensors and action models, rather than a dedicated module for a particular hardware or policy. For the industry, sensor and dexterous-hand manufacturers 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 capabilities 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 version LingBot-Touch-Tiny with about 5.3 million parameters, and after connecting to LingBot-VLA, π0.5, and T-Rex, the model parameter count increased by only about 0.13% to 0.16%; the larger LingBot-Touch-Small showed potential for further improvement in perception tests—when pretraining data increased to nearly 5,000 hours, the model could still benefit from the newly added data, indicating that tactile pretraining still has room for continuous scaling and improvement.

Adhering to embodied nativity, advancing in coordination with general large models

Fable 5.2 and LingBot-Touch appeared on the same day, which is to some extent a microcosm of the current evolution of embodied intelligence technology routes. Regarding the impact of the accelerated evolution of general large models on the robot "brain" route, Ant Lingbo CEO Zhu Xing said: "The progress of digital-world large language models represented by Claude Fable 5.2 and GPT-6 Astra has indeed made robots' complex task reasoning and physical manipulation capabilities stronger, and we pay close attention to this and are actively absorbing it. But frontline industry practice tells us that to truly close the manipulation loop, the hard nut of physical manipulation generalization is still unavoidable, and to a large extent it still depends on breakthroughs in embodied foundation models themselves."

In his view, general 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; the two make breakthroughs separately and promote each 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, without good input there is no good output," Zhu Xing said. Ant Lingbo will adhere to the embodied-native route, continue exploring new upper limits of physical manipulation generalization, build a general brain for the robotics industry across configurations, promote the alignment of more physical modalities such as vision, depth, touch, and even temperature in a unified model, and continue to open source 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.

Leiphone Leiphone