文献类型：期刊文献

英文题名：Coevolution of large language models with physical models boosts advanced battery research

作者：Li, Jia-hui[1,2];Hu, Yue[3];Xia, Guangyu[4];Mo, Wendi[4];Li, Baorong[1];Jia, Yingzhen[5];Gao, Yang[3];Xuan, Fuzhen[3];Liu, Honglai[1,3];Lian, Cheng[1,3]

机构：[1]East China Univ Sci & Technol, Sch Chem & Mol Engn, Shanghai 200237, Peoples R China;[2]Shenzhen Univ Adv Technol, Fac Synthet Biol, Shenzhen 518107, Peoples R China;[3]East China Univ Sci & Technol, Sch Mech & Power Engn, Shanghai Key Lab Intelligent Sensing & Detect Tech, Shanghai 200237, Peoples R China;[4]East China Univ Sci & Technol, Sch Chem Engn, State Key Lab Chem Engn, Shanghai 200237, Peoples R China;[5]Boston Univ, Coll Arts & Sci, Boston, MA 02215 USA

年份：2025

卷号：6

期号：5

外文期刊名：CELL REPORTS PHYSICAL SCIENCE

收录：;WOS:【SCI-EXPANDED(收录号:WOS:001497599200024)】；

基金：This work was supported by the National Natural Science Foundation of China (nos. 22308096, 22278127, and 22378112) , Shanghai Pilot Program for Basic Research (22T01400100-18) , and the Fundamental Research Funds for the Central Universities (no. 2022ZFJH004) .

语种：英文

摘要：With the advent of large language models (LLMs), deep artificial intelligence methods have begun to approach the field of scientific research. Nevertheless, the pre-trained LLMs face problems such as information fabrication and the inability to handle simulations in specific research fields, which restricts the further integration of artificial intelligence with professional scientific research. Here, we present a pioneering framework that harnesses the textual comprehension of LLMs alongside the physical models, thereby advancing the field of battery research. Our coevolution blueprint not only refines the LLMs' ability to conduct domain-specific simulations but also generates data, steering AI toward broader intelligence. The integration of LLMs and physical models provides a powerful tool for researchers, streamlining the transition from data analysis to experimental validation, and inspires more researchers to adopt this paradigm as a prototype to promote the integration between data, models, and information, innovating the research patterns in related fields.