Jiangxing Intelligence Takes Part in the National Key R&D Program to Drive the Deployment of Causal World Models in Industrial Scenarios
On June 22, the mid-term exchange meeting for the project Causal Learning Theory and Key Technologies for Large Model Robustness, a special project under the National Key R&D Program for Strategic Science and Technology Innovation Cooperation, was hosted by Jiangxing Intelligence. Teams from Harbin Institute of Technology, the Hong Kong University of Science and Technology, Zhejiang University, iFLYTEK and Jiangxing Intelligence conducted an in-depth review of research progress, agreeing to accelerate the industrial deployment of causal world models in real-world industrial scenarios in the next phase.
From left: Cai Bibo, Kuang Kun, Fan Xiaoyi, Zhang Qian, Ding Xiao, Zhang Cong, Chen Long
Since the project was launched in 2025, it has centered on causal representation, robust algorithms and evaluation systems, targeting a core pain point of large models in complex physical scenarios: these models can only identify data correlations rather than causal relationships. Traditional models rely on superficial data correlations and tend to malfunction in industrial environments featuring changing lighting conditions, fluctuating equipment status and other disturbances. By explicitly learning causal structures and dynamic mechanisms between variables, causal world models enable large models to truly comprehend physical logic such as the root causes of anomalies and the consequences of operational actions.
Professor Zhang Qian from the Hong Kong University of Science and Technology at the Physical AI Workshop of Jiangxing Intelligence
As the sole industrial implementation partner of this project, Jiangxing Intelligence is tasked with integrating research achievements into its self-developed JX-Phi Physical AI system. Through collaborative optimization covering the full chain of perception and recognition, logical reasoning, as well as decision-making and planning, the company is helping transform theoretical breakthroughs of causal world models into operable and iterable system capabilities for practical scenarios including power inspection and industrial quality inspection.
Going forward, Jiangxing Intelligence will further deepen industry-university-research collaboration, scale up the industrial application of causal world models, and deliver core technical support for the large-scale rollout of Physical AI in open industrial sites.