Mathematics meets Artificial Intelligence
MathonAI Team brings together faculty and students mainly from Tsinghua University and BIMSA. We study how mathematical structure can make modern AI and foundation models more principled, reliable, and useful for scientific discovery, with active work spanning generative modeling, agent systems, scientific machine learning, and AI for molecules, materials, and physics.
Primary Research Areas
PDEs, numerical analysis, inverse problems, and dynamical-systems viewpoints for machine learning and foundation models.
Discrete diffusion, autoregressive diffusion, flow-based generation, and controllable generative modeling across modalities.
Agentic model design, long-horizon memory, reasoning systems, and the infrastructure needed to support reliable agents.
Data-driven discovery for PDEs and stochastic dynamics, together with machine learning models informed by scientific structure.
Computational physics, topological materials, quantum systems, and scientific modeling at the interface of math and AI.
Molecular structure generation, science foundation models, and AI for materials discovery including metal-organic frameworks.
Recent papers, courses, and milestones from MathonAI Team
Recent papers by our faculty, students, and collaborators, grouped into published and preprint work
Selected systems and products developed by our team
Selected BIMSA courses taught by our team
Faculty and students mainly from Tsinghua University and BIMSA