Yunbei Xu

Research

Papers under Review and Revision

• Pointwise Generalization in Deep Neural Networks

  with Shaojie Li.

  Under review at Journal of the ACM.

• Upper Counterfactual Confidence Bounds: a New Optimism Principle for Contextual Bandits

  with Assaf Zeevi.

  Under revision at Journal of Machine Learning Research.

Papers in Preparation for Submission

• Statistical Properties of Robust Learning under Distribution Shifts

  with Zhiyi Li, Xiaojia Mao, and Ruohan Zhan.

  To be submitted to Operations Research.

• On the Blessing of Pre-training in Weak-to-Strong Generalization

  with Wei Yao, Wang Zhaoyang, Gengze Xu, Chen Qian, Dongrui Liu, Ziqiao Wang, and Yong Liu.

  To be submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence.

• Local Search with Correlated Randomness

  In preparation for submission to ACM Symposium on Theory of Computing (STOC) 2027.

Working Papers and Preprints

• Pointwise Complexity for Gaussian Fields: Upper Envelopes, Algorithmic Lower Bounds, and Separation

  Preprint.

• Bellman-sufficient Information Complexity

  Preprint.

• In-Context Learning for Data-Driven Censored Inventory Control

  with Sohom Mukherjee, Anh-Duy Pham, and Richard Pibernik.

  Preprint.

• Thompson Sampling for Repeated Newsvendor

  with Weizhou Zhang, Chen Li, Hanzhang Qin, and Ruihao Zhu.

  Preprint.

Published Papers

• Sampling Allocation of LinUCB: Optimal Design Limits in the Small-Gap Regime

  with Yujie Liu and Vincent Tan.

  ACM SIGMETRICS 2027.

• Autoregressive Learning in Joint KL: Sharp Oracle Bounds and Lower Bounds

  with Yuzhe Yuan and Ruohan Zhan.

  Conference on Neural Information Processing Systems (NeurIPS) 2026.

  Spotlight (top 4.3%) in NeurIPS 2025 ML×OR Workshop.

• Finite-Time Minimax Bounds and an Optimal Lyapunov Policy in Queueing Control

  with Yujie Liu and Vincent Tan.

  Operations Research, 2026.

  Stochastic Networks Conference Outstanding Poster Prize

• On the Power of Adaptivity for ε-Best Arm Identification in Linear Bandits

  with Arnab Maiti and Kevin Jamieson.

  Conference on Learning Theory (COLT) 2026.

• Bayesian Design Principles for Frequentist Sequential Learning

  with Assaf Zeevi.

  Journal of the ACM, 2025. Code

  International Conference on Machine Learning (ICML) 2023.

  ICML Outstanding Paper Award

  INFORMS George Nicholson Student Paper Competition, First Place

  Applied Probability Society Best Student Paper Award, Finalist

  Special Plenary Session: Highlights Beyond EC

• Towards Optimal Problem Dependent Generalization Error Bounds in Statistical Learning Theory

  with Assaf Zeevi.

  Mathematics of Operations Research, 2025.

  Applied Probability Society Best Student Paper Award, Finalist

• Assouad, Fano, and Le Cam with Interaction: A Unifying Lower Bound Framework and Characterization for Bandit Learnability

  with Fan Chen, Dylan Foster, Yanjun Han, Jian Qian, and Alexander Rakhlin.

  Conference on Neural Information Processing Systems (NeurIPS) 2024.

  Spotlight (top 2.5%)

  Journal version under preparation.

• Statistical Properties of Robust Satisficing

  with Zhiyi Li and Ruohan Zhan.

  International Conference on Machine Learning (ICML) 2024.

  INFORMS Undergraduate Operations Research Prize, Finalist

• Acceleration of Primal-Dual Methods by Preconditioning and Simple Subproblem Procedures

  with Yanli Liu and Wotao Yin.

  Journal of Scientific Computing, 2021. Code

• Towards Problem-dependent Optimal Learning Rates

  with Assaf Zeevi.

  Conference on Neural Information Processing Systems (NeurIPS) 2020.

  Spotlight (top 4.1%)