As part of the World Digital Education Conference, held in Hangzhou on May 12, Parallel Session 5 was convened under the theme “Building New Scientific Research Capabilities: Transforming Research Paradigms with AI”. At the close of the session, six world-class China-developed AI foundation models and intelligent agents supporting scientific research were released, representing the nation’s approach to integrating artificial intelligence (AI) into this field. The event was hosted by Jiang Peixue, an academician of the Chinese Academy of Sciences and deputy chair of the Tsinghua University Council.
The models and agents covered many interdisciplinary areas, including life sciences, chemistry and materials, wireless communications, smart healthcare, automated research, and traditional Chinese medicine (TCM) education.
Beijing Zhongguancun Academy introduced Carbon, a generative genomic foundation model designed for DNA sequence generation and genome annotation.
Shanghai Innovation Institute and East China Normal University presented an intelligent agent infrastructure for self-driving laboratories, aimed at advancing automated and autonomous experimentation.
Shenzhen Loop Area Institute and Shenzhen Research Institute of Big Data jointly launched a high-precision multimodal wireless dataset and the AI4Net foundation model, designed to sense and predict real-world physical wireless environments.
Shanghai Jiao Tong University unveiled DeepRare, the world’s first agent-based evidence-driven reasoning system for rare disease diagnosis, built to improve transparency and reliability in clinical reasoning.
Westlake University released DeepScientist, an open-source automated scientific research system capable of carrying out the full workflow from autonomous reasoning to experiment execution.
Beijing University of Chinese Medicine released China’s first officially registered large AI model for TCM education, establishing a benchmark for compliant AI applications in the field.
The session drew considerable interest from participants, with many describing the showcased projects as not merely algorithmic advancements, but practical research toolkits capable of accelerating scientific discovery. Attendees said the rapid development of scientific AI models and intelligent agents would allow researchers to focus more on creative scientific thinking.
The conference highlighted a broader shift in research paradigms—from AI as a supporting instrument toward intelligent systems, deep human-machine collaboration, and automated closed-loop research workflows. The newly released scientific AI models and agents are expected to significantly improve research quality and efficiency, shorten discovery cycles, and accelerate the transition from laboratory exploration to large-scale application. The open-source release and wider adoption of these tools are also expected to support the development of a more open, collaborative, and intelligent research ecosystem.
