Project Grant 2608503

The National Science Foundation Directorate for Mathematical and Physical Sciences awarded $250,000 to the University of Wisconsin – Madison on September 1, 2026, under the Mathematical and Physical Sciences program (CFDA 47.049) to develop high-fidelity generative mean flow models that advance scientific machine learning through a theory-validation feedback loop.

The project addresses critical limitations in applying generative AI to physical simulations by building a new class of AI models designed specifically for scientific and engineering applications with rigorous accuracy guarantees and reduced computational costs. Research goals focus on flow-based generative models, targeting three key challenges: closing the fidelity gap relative to diffusion models, reducing computational complexity during sampling, and establishing application-driven validation loops. The work aims to accelerate complex computations in fields of federal strategic interest, including nanoscale material design for electronic devices and plasma control for nuclear fusion energy production.

The award runs through August 31, 2029, with place of performance in Madison, Wisconsin. Beyond scientific advancement, the project supports training and education of three to four doctoral students at the University of Wisconsin – Madison and the University of Wisconsin – Milwaukee, funds curriculum development in data science and stochastic computation, and disseminates open-source software and research findings through international workshops to develop a highly skilled STEM workforce.