Xiamen University (XMU) announced that its proposal for a national mega-project under the Ministry of Science and Technology’s New Generation Artificial Intelligence initiative—titled “Construction and Application of a Knowledge-Enhanced Embodied Scientific Agent Platform”—has been officially approved. A kickoff meeting was held in Shanghai on January 22, marking the project’s transition into full-scale implementation.

Led by XMU, the project brings together top-tier institutions like Peking University, the University of Science and Technology of China, and Zhengzhou University, alongside industry leaders such as AGIBOT, Wantai BioPharm (Beijing Wantai & UMIC), Hithium, Yealink Network, and Mintou Technology. It aims to deepen integration between scientific research and industrial development, establish a next-generation research ecosystem, pioneer AI-driven scientific paradigms, and secure a global competitive edge in scientific intelligence.

Aligned with China’s “AI+” national strategy, the three-year project will develop a world-leading, hardware-software-integrated embodied intelligence platform with universal applications across key industries. The platform seeks to automate the entire research workflow—from hypothesis generation and experimental planning to execution—potentially overcoming the inefficiencies of traditional trial-and-error methods. It is expected to accelerate the shift toward intelligent, autonomous, and standardized research practices while setting benchmarks for cross-industry advancement.

To address global challenges like “expertise hallucination” in general AI models and the “reality-simulation gap” in scientific applications, the project will integrate cutting-edge technologies such as large language models, knowledge graphs, and embodied intelligence. Efforts will focus on five core areas: developing a foundational embodied scientific model, establishing trust mechanisms and knowledge-updating frameworks for AI agents, overcoming training and deployment barriers for virtual-physical integration, building a unified experimental platform, and creating an interdisciplinary open-source validation system.

The project targets complex research scenarios, enhancing scientific agents’ specialization, digital twin reliability, and hardware-software synergy. It will also establish a robust evaluation framework and promote applications in medicine, biology, materials science, and energy.

By building a self-sufficient “embodied scientific intelligence” system, the project will not only foster interdisciplinary innovation but also deliver replicable “future lab” solutions for strategic sectors like smart manufacturing and life sciences in China.