MatOS: The Intelligent Agent for Materials Science Officially Goes Open Source, Markedly Boosting Success Rates in Materials R&D
2 day ago / Read about 0 minute
Author:小编   

In September of this year, the Beijing Zhongguancun Academy and the Zhongguancun Institute of Artificial Intelligence released MatOS, a scientific research agent system tailored for materials R&D. The system uniformly organizes six types of elements—including data and knowledge—into scientific assets. Through three key design features—a hierarchical scientific asset graph, stateful adaptive workflows, and evidence-driven asset evolution—it enables scientific research capabilities to continuously evolve with use. Currently, its asset library has integrated a wealth of resources, including datasets, skills, and workflows, along with multiple materials computing software packages. The related technical report has been published on arXiv, and the code has been open-sourced. Multiple benchmark tests and research cases demonstrate that MatOS outperforms comparable systems in tasks such as end-to-end autonomous scientific research and the application of materials software. It can handle complex materials research tasks—such as screening tens of millions of candidate structures and achieving high-precision prediction of alloy melting points—and can adapt to different foundation models, thereby expanding the cost-score Pareto frontier. Future versions of the system will further integrate instruments and experimental workflows.