I am a principal researcher at Microsoft Research, New England (and New York City), where I am a member of the Reinforcement Learning Group. Previously, I was a postdoctoral fellow at the MIT Institute for Foundations of Data Science in IDSS, and prior to this I received my PhD from the Department of Computer Science at Cornell University (2019), advised by Karthik Sridharan. I received my BS and MS in electrical engineering from the University of Southern California in 2014.
Research
I am interested in the mathematical foundations—algorithm design principles and fundamental limits—necessary to develop intelligent agents that learn from experience. I am currently most excited about:
The statistical and computational foundations of interactive decision making, including reinforcement learning and imitation learning.
Understanding and improving foundation models, from pre-training to post-training and test-time—particularly as a basis upon which to build interactive decision making agents.
More broadly, I like to dabble in almost all theoretical aspects of machine learning and adjacent topics (generalization, optimization, information theory, statistics, …).
News
8/15/25: We will teach our course 9.522: Statistical Reinforcement Learning and Decision Making for the third time at MIT this Fall, with updated content on RL for language models.
8/1/25: Adam Block, Max Simchowitz, and I will be presenting a NeurIPS 2025 tutorial, Foundations of Imitation Learning: From Language Modeling to Continuous Control.
7/4/25: We are organizing a workshop on Foundations of Reasoning in Language Models at NeurIPS 2025! The submission deadline is Sept 3, 2025.
7/1/25: Upcoming talks: July: EXAIT workshop at ICML, Aug: IAIFI Summer Workshop at Harvard, IFDS Summer Workshop at UW, Sept: Harvard Stats Colloquium
6/29/25: I was elected to the board of directors of the Association for Computational Learning for a four-year term.
5/10/25: We are organizing a workshop on Foundations of Post-Training at COLT 2025! The submission deadline is May 19, 2025.
Internships and Postdocs
I have been fortunate to work with the following amazing interns and postdocs at MSR:
Phil Amortila (2023), Adam Block (2023, PD ‘24-25), Fan Chen (2025), Noah Golowich (2022, PD ‘25-26), Audrey Huang (2024, 2025), Qinghua Liu (PD ‘24-25), Sadhika Malladi (PD ‘25-26), Nived Rajaraman (PD ‘25-), Dhruv Rohatgi (2024), Clayton Sanford (2023), Anikait Singh (2025), Yuda Song (2023), Jens Tuyls (2025), Andrew Wagenmaker (2022), Tengyang Xie (PD ‘23-24), Yunzong Xu (2021, PD ‘23-24), and Yinglun Zhu (2021).
Check back in Fall 2025 for internship and postdoc postings for 2026.
Selected Recent Papers
Is Best-of-N the Best of Them? Coverage, Scaling, and Optimality in Inference-Time Alignment
Audrey Huang, Adam Block, Qinghua Liu, Nan Jiang, Akshay Krishnamurthy, and Dylan J. Foster.
ICML, 2025.
Is a Good Foundation Necessary for Efficient Reinforcement Learning? The Computational Role of the Base Model in Exploration
Dylan J. Foster, Zakaria Mhammedi, and Dhruv Rohatgi.
COLT, 2025.
Computational-Statistical Tradeoffs at the Next-Token Prediction Barrier: Autoregressive and Imitation Learning under Misspecification
Dhruv Rohatgi, Adam Block, Audrey Huang, Akshay Krishnamurthy, and Dylan J. Foster.
COLT, 2025.
Self-Improvement in Language Models: The Sharpening Mechanism
Audrey Huang*, Adam Block*, Dylan J. Foster*, Dhruv Rohatgi, Cyril Zhang, Max Simchowitz, Jordan T. Ash, and Akshay Krishnamurthy.
ICLR, 2025. Oral presentation.
Correcting the Mythos of KL-Regularization: Direct Alignment without Overoptimization via Chi-Squared Preference Optimization
Audrey Huang, Wenhao Zhan, Tengyang Xie, Jason D. Lee, Wen Sun, Akshay Krishnamurthy, and Dylan J. Foster.
ICLR, 2025. Spotlight presentation.
Exploratory Preference Optimization: Harnessing Implicit Q*-Approximation for Sample-Efficient RLHF
Tengyang Xie*, Dylan J. Foster*, Akshay Krishnamurthy, Corby Rosset, Ahmed Awadallah, and Alexander Rakhlin.
ICLR, 2025.
Reinforcement Learning Under Latent Dynamics: Toward Statistical and Algorithmic Modularity
Philip Amortila, Dylan J. Foster, Nan Jiang, Akshay Krishnamurthy, and Zakaria Mhammedi.
NeurIPS, 2024. Oral presentation.
Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning
Dylan J. Foster, Adam Block, and Dipendra Misra.
NeurIPS, 2024. Spotlight presentation.
Tight Guarantees for Interactive Decision Making with the Decision-Estimation Coefficient
Dylan J. Foster, Noah Golowich, and Yanjun Han.
COLT, 2023.
The Role of Coverage in Online Reinforcement Learning
Tengyang Xie*, Dylan J. Foster*, Yu Bai, Nan Jiang, and Sham M. Kakade.
ICLR, 2023. Oral presentation.
Offline Reinforcement Learning: Fundamental Barriers for Value Function Approximation
Dylan J. Foster, Akshay Krishnamurthy, David Simchi-Levi, and Yunzong Xu.
COLT, 2022.
The Statistical Complexity of Interactive Decision Making
Dylan J. Foster, Sham M. Kakade, Jian Qian, and Alexander Rakhlin.
Preprint (under review), 2021.
Efficient First-Order Contextual Bandits: Prediction, Allocation, and Triangular Discrimination
Dylan J. Foster and Akshay Krishnamurthy.
NeurIPS, 2021. Oral presentation.
Selected Older Papers
Lower Bounds for Non-Convex Stochastic Optimization
Yossi Arjevani, Yair Carmon, John C. Duchi, Dylan J. Foster, Nathan Srebro, and Blake Woodworth.
Mathematical Programming, Series A, 2022.
Beyond UCB: Optimal and Efficient Contextual Bandits with Regression Oracles
Dylan J. Foster and Alexander Rakhlin.
ICML, 2020.
Naive Exploration is Optimal for Online LQR
Max Simchowitz and Dylan J. Foster.
ICML, 2020.
Orthogonal Statistical Learning
Dylan J. Foster and Vasilis Syrgkanis.
COLT, 2019. Best Paper Award. Journal version in Annals of Statistics (2023).
The Complexity of Making the Gradient Small in Stochastic Convex Optimization
Dylan J. Foster, Ayush Sekhari, Ohad Shamir, Nathan Srebro, Karthik Sridharan, and Blake Woodworth.
COLT, 2019. Best Student Paper Award.
Logistic Regression: The Importance of Being Improper
Dylan J. Foster, Satyen Kale, Haipeng Luo, Mehryar Mohri, and Karthik Sridharan.
COLT, 2018. Best Student Paper Award.
Online Learning: Sufficient Statistics and the Burkholder Method
Dylan J. Foster, Alexander Rakhlin, and Karthik Sridharan.
COLT, 2018.
Spectrally-Normalized Margin Bounds for Neural Networks
Peter L. Bartlett, Dylan J. Foster, and Matus J. Telgarsky.
NeurIPS, 2017. Spotlight presentation.
Selected Awards
Best Paper Award (Orthogonal Statistical Learning)
Conference on Learning Theory (COLT), 2019.
Best Student Paper Award (
Conference on Learning Theory (COLT), 2019.
Best Student Paper Award (Logistic Regression: The Importance of Being Improper)
Conference on Learning Theory (COLT), 2018.
Facebook PhD Fellowship, 2018.
NDSEG PhD Fellowship, 2016.