Profiles2021-2026
Dylan J Foster
Microsoft Research
18Papers
2Last author
24Unique co-authors
6First author
0Solo
First and last author come from byline position, which some venues assign alphabetically.
Recent papers18
- The Coverage Principle: How Pre-Training Enables Post-TrainingAuthor 8 of 8ICLR 2026 · 28 citations
- Exploratory Preference Optimization: Harnessing Implicit Q*-Approximation for Sample-Efficient RLHFAuthor 2 of 3ICLR 2025
- Self-Improvement in Language Models: The Sharpening MechanismAuthor 3 of 6ICLR 2025
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation LearningAuthor 1 of 3NeurIPS 2024 · 112 citations
- Assouad, Fano, and Le Cam with Interaction: A Unifying Lower Bound Framework and Characterization for Bandit LearnabilityAuthor 2 of 5NeurIPS 2024 · 15 citations
- The Power of Resets in Online Reinforcement LearningAuthor 2 of 3NeurIPS 2024 · 15 citations
- Online Estimation via Offline Estimation: An Information-Theoretic FrameworkAuthor 1 of 3NeurIPS 2024 · 13 citations
- Butterfly Effects of SGD Noise: Error Amplification in Behavior Cloning and AutoregressionAuthor 2 of 6ICLR 2024 · 12 citations
- Scalable Online Exploration via CoverabilityAuthor 2 of 2ICML 2024 · 10 citations
- Rich-Observation Reinforcement Learning with Continuous Latent DynamicsAuthor 3 of 4ICML 2024 · 2 citations
- Model-Free Reinforcement Learning with the Decision-Estimation CoefficientAuthor 1 of 3NeurIPS 2023 · 16 citations
- Hardness of Independent Learning and Sparse Equilibrium Computation in Markov GamesAuthor 1 of 2ICML 2023 · 14 citations
- The Role of Coverage in Online Reinforcement LearningAuthor 2 of 4ICLR 2023 · 1 citation
- On the Complexity of Adversarial Decision MakingAuthor 1 of 2NeurIPS 2022 · 37 citations
- Contextual Bandits with Large Action Spaces: Made PracticalAuthor 2 of 5ICML 2022 · 34 citations
- Understanding the Eluder DimensionAuthor 3 of 4NeurIPS 2022 · 22 citations
- Interaction-Grounded Learning with Action-Inclusive FeedbackAuthor 3 of 8NeurIPS 2022 · 12 citations
- Efficient First-Order Contextual Bandits: Prediction, Allocation, and Triangular DiscriminationAuthor 1 of 2NeurIPS 2021 · 62 citations
