文献类型：期刊文献

英文题名：Generative transformer decoder-only model polyGT: Toward polymer informatics comprehension and target polymer inverse design

作者：Xu, Hao[1,2,3];Zhong, Weimin[1,2,3];Peng, Xin[3,4];Wang, Yiming[1]

机构：[1]East China Univ Sci & Technol, State Key Lab Chem Engn, Meilong Rd 130, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, Sch Informat Sci & Engn, Meilong Rd 130, Shanghai 200237, Peoples R China;[3]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Meilong Rd 130, Shanghai 200237, Peoples R China;[4]East China Univ Sci & Technol, State Key Lab Ind Control Technol, Meilong Rd 130, Shanghai 200237, Peoples R China

年份：2026

卷号：324

外文期刊名：CHEMICAL ENGINEERING SCIENCE

收录：;EI(收录号：20260319935750);WOS:【SCI-EXPANDED(收录号:WOS:001662706800001)】；

基金：This work was supported by National Key Research and Development Program of China (2023YFB3307800) , Fundamental and Interdisciplinary Disciplines Breakthrough Plan of the Ministry of Education of China (JYB2025XDXM402) , the Shanghai Pilot Program for Basic Research (22TQ1400100-16) , National Natural Science Foundation of China (62173145, 62403201) , the Open Research Project of the State Key Laboratory of Industrial Control Technology of China (ICT2024A23) . and Fundamental Research Funds for the Central Universities (222202517006) .

语种：英文

外文关键词：Polymer; Generative artificial intelligence; Inverse design; Transformer model; Polymer generation; Machine learning

摘要：The scarcity of polymer data poses an obstacle to the application of data-driven artificial intelligence methods in polymer research. Computer-assisted polymer property prediction has become increasingly mature, while research on polymer design remains in its early stages. Polymer design models are plagued with weak robustness and complex model structures. In this paper, inspired by natural language processing models, we propose a polymer inverse design approach based on a generative Transformer decoder-only model. Through a self-supervised training pipeline, the model learns the chemical structure rules of polymers, thereby generating a broader polymer chemical space. Furthermore, using a supervised training pipeline, the model establishes mappings between customized properties and polymer structures to design candidate polymers. This research highlights the potential of artificial intelligence in polymer data generation and inverse design, serving as a starting point for future innovation cycles in polymer materials science.