News

PhD Student Benjamin Valpey Wins Best Paper Award

Published
July 28, 2025
Benjamin holding up paper award on conference stage

Congratulations to computer science PhD Student Benjamin Valpey, who won the “Best Paper Award” at the 17th NASA Formal Methods Symposium held in June in Williamsburg, Virginia.

Benjamin wrote the paper, “An SMT Formalization of Mixed-Precision Matrix Multiplication (Modeling Three Generations of Tensor Cores),” with co-authors Xinyi Li of Pacific Northwest National Laboratory, Ganesh Gopalakrishnan of University of Utah, and Associate Professor Sreepathi Pai.

The work studies and formalizes GPU functional units called tensor cores used primarily in AI applications, enabling automated and rigorous analysis of algorithms that use these cores. Benjamin and his co-authors show that a recently proposed algorithm using these cores and designed to be more accurate than previous proposals but only tested empirically is actually less accurate on some inputs. The prize is accompanied by a €1,500 award sponsored by Springer. Terrific work!