Frank Cole (UMN): In-Context Learning for Scientific Computing
Please note this event occurred in the past.
May 20, 2026 11:00 am - 12:00 pm ET
LGRT 1685
Abstract
Abstract:
Transformer-based foundation models have demonstrated the ability to perform in-context learning: given a prompt containing a few examples of an unseen task, they can make relevant predictions without parameter updates. While its origins lie in language modeling, in-context learning has recently emerged as a tool for scientific computing, exhibiting success in data-scarce settings. In this talk, I will give a mathematical overview of in-context learning and present recent advances in its theoretical foundations. I will focus on the application of transformer neural networks to three representative problems in scientific computing: solution operator learning for PDEs, dynamical system prediction, and optimal transport.