Time: 10:00-12:00, April 1st, 2026

Location: Meeting Room of Kavli ITS, 3rd Floor, No. 7 Building, UCAS

Speaker: Zeyu Li (Peking Univ.)

Abstract:   Transformers are the foundational architecture of large language models, with powerful capabilities in processing sequences. In this talk, I will briefly review the basic principles of Transformers and their existing applications in theoretical physics research. Then, I will introduce how to apply them to exactly solvable two-dimensional lattice gauge theories. Specifically, we convert Wilson loops and their analytical solutions on the lattice into sequences and use Transformers to learn the underlying mapping between them. In the designed tasks, the Transformer achieves high accuracy and exhibits a certain degree of generalization.