en.Wedoany.com Reported - On May 26, Ant Lingbo announced that it has recently entered into a deep strategic partnership with Jianzhi Xinchuang (Beijing) Robotics Technology Co., Ltd. The two parties will collaborate on data around Ant Lingbo's full suite of embodied intelligence model matrices and jointly develop proprietary data collection equipment to enhance the precision and dimensionality of human data, accelerating the large-scale expansion of high-quality physical real-world data.
This collaboration targets the data bottleneck in embodied intelligence training. Unlike large language models that primarily rely on text, image, and code data, embodied intelligence models need to understand actions, objects, force feedback, tactile changes, spatial relationships, and task causal chains in the real physical world. For a robot to complete tasks such as grasping, transporting, organizing, cleaning, and operating tools, the training data must not only record "what is seen" but also, as much as possible, record how humans judge, move, touch, adjust force, and the physical feedback after an action is completed. Ant Lingbo and Jianzhi Robotics are collaborating around model matrices and data collection equipment, with the core goal of improving the authenticity, precision, and reusability of data collection.
The direction Jianzhi Robotics is taking is precisely the data infrastructure for embodied intelligence. Public information shows that Jianzhi Xinchuang (Beijing) Robotics Technology Co., Ltd. was established in July 2025, focusing on one-stop embodied intelligence data solutions, emphasizing serving embodied intelligence research and development based on real-world data.
The value of high-quality physical real-world data is mainly reflected in model generalization capabilities. Currently, many robot demonstrations can complete tasks in limited scenarios, but once the environment changes to a new space, new objects, new lighting, new human action habits, or new task sequences, problems such as recognition deviations, action errors, or inability to transfer easily occur. For an embodied intelligence model to move from "executable single actions" to "generalizable across multiple scenarios," it must rely on larger-scale, higher-precision, and stronger spatiotemporally aligned data. Jointly developing proprietary data collection equipment helps integrate human behavior, visual information, tactile feedback, joint movements, and task outcomes into a single collection chain, providing a more complete data loop for model training.
Ant Lingbo is an embodied intelligence company under Ant Group. Its official website shows it focuses on humanoid robot research and development. The R1/R1-S intelligent service robots support applications in commercial, home, and office scenarios, and possess capabilities such as multimodal interaction and autonomous control. Previous reports from Beijing Daily also mentioned that Ant Group responded by stating that Shanghai Ant Lingbo Technology Co., Ltd. focuses on the research and development of embodied intelligence technologies and products, and continues to strategically deploy and increase AI investment.
For Ant Lingbo, data collaboration can support the continuous iteration of its full suite of embodied intelligence model matrices. A model matrix implies that different tasks, different embodiments, and different scenarios may require the collaboration of multiple types of model capabilities, such as perception models, action models, planning models, interaction models, and control models. If data sources are scattered, dimensions are insufficient, or action chains are incomplete, it is difficult for model training to form stable transfer capabilities; if data collection equipment can continuously generate high-quality, multi-dimensional, alignable human operation data, models can more easily learn the subtle actions and decision-making logic in real scenarios.
This type of cooperation also indicates that competition in the embodied intelligence industry is expanding from the robot body itself to the systemic capabilities of "data—model—equipment—scenario." The hardware body determines whether a robot can execute actions, the model determines whether the robot can understand tasks and plan behaviors, and the data determines whether the model can continuously improve in the real world. By combining Jianzhi Robotics' capabilities in data collection and data infrastructure with Ant Lingbo's embodied intelligence model matrices, both parties have the opportunity to form a tighter engineering loop around high-quality data production, model training, equipment verification, and scenario implementation.
Subsequent project milestones include the research and development progress of the proprietary data collection equipment, data collaboration mechanisms, types of data collected, the range of compatible robot bodies, model training effectiveness, and whether both parties disclose specific scenario verification results. What can be confirmed at this stage is that Ant Lingbo and Jianzhi Robotics have reached a strategic cooperation and plan to collaborate around the embodied intelligence model matrix, data collaboration, and data collection equipment; public information has not yet disclosed the cooperation amount, equipment models, collection scale, model parameters, customer lists, or commercialization timeline. Therefore, it should not be extrapolated that the relevant equipment is already in mass production or that the two parties have formed confirmed orders.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
