Translation. Region: Russian Federal
Source: People's Republic of China in Russian – People's Republic of China in Russian –
An important disclaimer is at the bottom of this article.
Source: People's Republic of China – State Council News
BEIJING, July 29 (Xinhua) — The Chinese Academy of Sciences (CAS) unveiled the ScienceOne model at the 2025 World Conference on Artificial Intelligence (AI), marking a major breakthrough in AI innovation.
Designed to analyze complex scientific data (including waveforms, spectra, and fields), ScienceOne integrates literature mining, knowledge-based reasoning, and tool coordination capabilities, aiming to become the intelligent foundation for technological innovation across multiple fields.
Current approaches rely heavily on retraining generic models with industry-specific data and developing isolated tools for specific domains. This paradigm faces three key limitations: fragmented scientific data systems, a lack of specialized inferential analysis capabilities, and closed research ecosystems.
To overcome these problems, 12 ANC institutes, including the Institute of Automation, the Center for Information Networks and Computing, the National Science Library, and the Institute of High Energy Physics, jointly developed ScienceOne.
The model has achieved a comprehensive understanding of complex scientific data (including waveforms, spectra and fields) by integrating key capabilities of literature mining, logical knowledge analysis and computational tool coordination. Researchers can easily use the capabilities of this model at various stages of scientific work.
Built on open-source Chinese models with deep scientific adaptation, ScienceOne combines a suite of unique models designed for common scientific data types, as well as specialized tools such as AlphaFold and MatterGen.
It demonstrates a systematic mastery of the fundamental principles, laws, and specialized knowledge in fundamental disciplines, including mathematics, physics, chemistry, astronomy, earth sciences, and biology.
The assessments confirmed its leading performance in mathematics, physics, chemistry, materials science, and biology, with leading capabilities in scientific professional tool recall and scientific reasoning. ScienceOne also demonstrated outstanding results in the Humanity's Last Exam (HLE).
Based on ScienceOne, the research team developed two scientific intelligent agents.
One is designed to help researchers analyze articles in depth, write reviews, and evaluate research topics and technical approaches. It has access to 170 million scientific publications and open source data in real time. With its support, literature review tasks that previously took 3-5 days are now completed in 20 minutes.
The second agent lowers the threshold for using scientific tools by providing autonomous planning of over 300 scientific computing tools. It automatically identifies research tasks and intelligently selects the optimal tools, increasing the efficiency of the scientific process.
Accelerating interdisciplinary discoveries, ScienceOne currently powers the X-Cell digital cell platform in the field of automatic identification of biological targets, improves the efficiency of particle modeling at the Beijing Electron-Positron Collider, improves the accuracy of molecular predictions in chemistry, optimizes the global coordination of telescopes for astronomy, and promotes innovation in the design of high-speed rail systems. -0-
Please note: This information is raw content obtained directly from the source of the information. It is an accurate report of what the source claims and does not necessarily reflect the position of MIL-OSI or its clients.
.