Publications & Preprints

The publications and preprints are sorted chronologically. For the most updated list, see my Google Scholar.

Publications

  1. Is Best-of-N the Best of Them? Coverage, Scaling, and Optimality in Inference-Time Alignment
    Audrey Huang, Adam Block, Qinghua Liu, Nan Jiang, Dylan J. Foster, Akshay Krishnamurthy. arxiv
    ICML 2025

  2. GaussMark: A Practical Approach for Structural Watermarking of Language Models
    Adam Block, Ayush Sekhari, and Alexander Rakhlin. arxiv
    ICML 2025

  3. A Theory of Learning with Autoregressive Chain of Thought
    Nirmit Joshi, Gal Vardi, Adam Block, Surbhi Goel, Zhiyuan Li, Theodor Misiakiewicz, Nathan Srebro. arxiv
    COLT 2025

  4. 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 Foster. arxiv
    COLT 2025

  5. Self-Improvement in Language Models: The Sharpening Mechanism N Audrey Huang, Adam Block, Dylan J. Foster, Dhruv Rohatgi, Cyril Zhang, Max Simchowitz, Jordan T. Ash, and Akshay Krishnamurthy [https:arxiv.orgabs2412.01951 ICLR 2025 (Oral)

  6. Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning
    Dylan Foster, Adam Block, and Dipendra Misra arxiv
    Neurips 2024 (Spotlight)

  7. Oracle-Efficient Differentially Private Learning with Public Data
    Adam Block, Mark Bun, Rathin Desai, Abhishek Shetty, and Zhiwei Steven Wu. arxiv
    Neurips 2024

  8. On the Performance of Empirical Risk Minimization with Smoothed Data
    Adam Block, Alexander Rakhlin, and Abhishek Shetty. arxiv
    COLT 2024

  9. Butterfly Effects of SGD Noise: Error Amplification in Behavior Cloning and Autoregression
    Adam Block, Dylan J. Foster, Akshay Krishnamurthy, Max Simchowitz, and Cyril Zhang. arxiv
    ICLR 2024

  10. Smoothed Online Learning for Prediction in Piecewise Affine Systems
    Adam Block, Max Simchowitz, and Russ Tedrake. arxiv
    Neurips 2023 (Spotlight)

  11. Efficient Model-Free Exploration in Low-Rank MDPs
    Zak Mhammedi, Adam Block, Dylan Foster, and Alexander Rakhlin. arxiv
    NeurIPS 2023

  12. On the Imitation of Non-Markovian Demonstrations: From Low-Level Stability to High Level Planning
    Adam Block, Ali Jadbabaie, Daniel Pfrommer, Max Simchowitz, and Russ Tedrake. arxiv
    NeurIPS 2023
    Workshop on Optimal Transport in Learning, Control, and Dynamical Systems, ICML 2023

  13. Oracle-Efficient Smoothed Online Learning for Piecewise Continuous Decision Making
    Adam Block, Max Simchowitz, and Alexander Rakhlin. COLT 2023. arxiv

  14. The Sample Complexity of Approximate Rejection Sampling With Applications to Smoothed Online Learning
    Adam Block and Yury Polyanskiy. COLT 2023. arxiv

  15. Efficient and Near-Optimal Smoothed Online Learning for Generalized Linear Functions
    Adam Block and Max Simchowitz. NeurIPS 2022. arxiv

  16. Intrinsic Dimension Estimation using Wasserstein Distance
    Adam Block, Zeyu Jia, Yury Polyanskiy, and Alexander Rakhlin. Journal of Machine Learning Research (Accepted 2022). arxiv

  17. Smoothed Online Learning is as Easy as Statistical Learning
    Adam Block, Yuval Dagan, Noah Golowich, and Alexander Rakhlin. COLT 2022. arxiv

  18. Counterfactual Learning To Rank for Utility-Maximizing Query Autocompletion
    Adam Block, Rahul Kidambi, Thorsten Joachims, Daniel N. Hill, and Inderjit S. Dhillon. SIGIR 2022. arxiv

  19. Majorizing Measures, Sequential Complexities, and Online Learning
    Adam Block, Yuval Dagan, and Alexander Rakhlin. COLT 2021. arxiv

Preprints

  1. Small Loss Bounds for Online Learning Separated Function Classes: A Gaussian Process Perspective
    Adam Block and Abhishek Shetty. arxiv

  2. Rate of Convergence of the Smoothed empirical Wasserstein Distance
    Adam Block, Zeyu Jia, Yury Polyanskiy, and Alexander Rakhlin. arxiv

  3. Fast mixing of multi-scale langevin dynamics underthe manifold hypothesis
    Adam Block, Youssef Mroueh, Jerret Ross, and Alexander Rakhlin. arxiv

  4. Generative modeling with denoising auto-encoders and Langevin sampling
    Adam Block, Youssef Mroueh, and Alexander Rakhlin. arxiv