Statistics and Data Science Seminar - Dylan Foster
January 26, 2026 4:00PM—5:00PM
Location:
In Person
-
Baker Hall A53
Speaker:
DYLAN FOSTER,
Principal ResearcherMicrosoft Research, New England and New York City
https://www.microsoft.com/en-us/research/people/dylanfoster/
The prevailing recipe of ever-larger models trained on passively collected data is showing diminishing returns and faces growing constraints on data. The next phase of progress will increasingly rely on interactive learning: systems that actively collect data through experimentation, from clinical trials and A/B testing to recommendation systems and scientific discovery. This talk presents a research program developing statistical foundations for interactive learning with modern deep models.
Part I introduces the Decision-Estimation Coefficient, a unifying framework for understanding when interactive learning is statistically tractable, analogous to empirical process theory in supervised learning. This theory directly yields practical, industry-deployed algorithms that transform any off-the-shelf estimator into an optimal sequential decision-making method.
Part II addresses the modern paradigm of adapting pre-trained foundation models for sequential decision making. I introduce the coverage profile—the probability mass a model places on high-quality outputs—as a key statistical quantity governing post-training success, leading to new interventions that connect pre-training objectives, post-training signals, and downstream performance.
—
Dylan Foster is a principal researcher at Microsoft Research, New England and New York City. Previously, he was a postdoctoral fellow at IDSS at MIT and received his PhD from Cornell University, advised by Karthik Sridharan. His research develops statistical foundations for interactive learning---spanning reinforcement learning, bandits, imitation learning, and causal inference---with a focus on understanding when complex decision-making problems can be solved using standard estimation primitives. His recent work applies this perspective to understand and improve the foundation model training pipeline, particularly post-training. His work has received several awards, including COLT Best Paper (2019), COLT Best Student Paper (2018, 2019), and the Cornell CS doctoral dissertation award.
3:30 pm → Refreshments will be offered prior to the seminar
For More Information:
jpaschke@andrew.cmu.edu