Kaizheng Wang

I am an assistant professor in the Department of Industrial Engineering and Operations Research and a core member of the Data Science Institute at Columbia University. My research interests lie at the intersection of machine learning, statistics and optimization. E-mail: kaizheng.wang [@] columbia [DOT] edu Google Scholar Resume

Openings

I'm seeking highly motivated Ph.D. students with strong mathematical backgrounds and interests in machine learning, statistics and optimization.

News

• PhD student Chengpiao Huang received the Outstanding Student Service Award at Columbia IEOR. Congratulations!
  

• New papers “Adaptive Querying with AI Persona Priors” (link) and “Estimating Continuous Treatment Effects with Two-Stage Kernel Ridge Regression” (link) accepted by ICML 2026. Congratulations to PhD students Yuhang Wu and Seok-Jin Kim!
  

• New paper “SYN-DIGITS: A Synthetic Control Framework for Calibrated Digital Twin Simulation” available on arXiv and SSRN.
  

• Honored to receive the 2026 IMS Tweedie New Researcher Award.
  

• New paper “Pseudo-Labeling for Unsupervised Domain Adaptation with Kernel GLMs” available on arXiv.
  

• Paper “Adaptive Transfer Clustering: A Unified Framework” accepted by Journal of the American Statistical Association.
  

• New paper “The Nonstationarity-Complexity Tradeoff in Return Prediction” available on arXiv and SSRN.
  

• Paper “A Particle Algorithm for Mean-Field Variational Inference” accepted by SIAM Journal on Optimization.
  

• I will be an Area Chair of ICML 2026.
  

Representative Publications

(\(\alpha\)-\(\beta\): author names are sorted alphabetically; \(\dagger\): student/postdoc supervised.)

• How Many Human Survey Respondents is a Large Language Model Worth? An Uncertainty Quantification Perspective.
  Chengpiao Huang\(\dagger\), Yuhang Wu\(\dagger\), Kaizheng Wang
  Preliminary version “Uncertainty Quantification for LLM-Based Survey Simulations” appeared in International Conference on Machine Learning, 2025.
  arXiv      Talk      Code
  
  

• A Minimalist Bayesian Framework for Stochastic Optimization.
  Kaizheng Wang
  Preliminary version “MINTS: Minimalist Thompson Sampling” appeared in NeurIPS 2025 ML×OR Workshop: Mathematical Foundations and Operational Integration of Machine Learning for Uncertainty-Aware Decision-Making.
  arXiv preprint arXiv:2509.07030, 2025.      Code
  
  

• A Stability Principle for Learning under Non-Stationarity.
  Chengpiao Huang\(\dagger\), Kaizheng Wang \((\alpha\)-\(\beta)\)
  Operations Research 73 (6): 3044-3064, 2025.
  arXiv      Code
  
  

• Pseudo-Labeling for Kernel Ridge Regression under Covariate Shift.
  Kaizheng Wang
  Annals of Statistics 54(1): 252-276, 2026.
  Junior Researcher Award, 2024 ICSA China Conference.
  arXiv      Talk      Code
  
  

• Adaptive and Robust Multi-Task Learning.
  Yaqi Duan, Kaizheng Wang \((\alpha\)-\(\beta)\)
  Annals of Statistics 51(5): 2015-2039, 2023.
  arXiv      Code
  
  

• Clustering a Mixture of Gaussians with Unknown Covariance.
  Damek Davis, Mateo Díaz, Kaizheng Wang \((\alpha\)-\(\beta)\)
  Bernoulli 31 (3): 2105-2126, 2025.
  arXiv
  
  

• An \(\ell_p\) Theory of PCA and Spectral Clustering.
  Emmanuel Abbe, Jianqing Fan, Kaizheng Wang \((\alpha\)-\(\beta)\)
  Annals of Statistics 50(4): 2359-2385, 2022.
  Frontiers of Science Award in Mathematics, 2024 International Congress of Basic Science.
  arXiv
  
  

• Entrywise Eigenvector Analysis of Random Matrices with Low Expected Rank.
  Emmanuel Abbe, Jianqing Fan, Kaizheng Wang, Yiqiao Zhong \((\alpha\)-\(\beta)\)
  Annals of Statistics 48(3): 1452-1474, 2020.
  arXiv
  
  

• Implicit Regularization in Nonconvex Statistical Estimation: Gradient Descent Converges Linearly for Phase Retrieval, Matrix Completion and Blind Deconvolution.
  Cong Ma, Kaizheng Wang, Yuejie Chi, Yuxin Chen
  Foundations of Computational Mathematics 20 (3): 451–632, 2020.
  SIAM Activity Group on Imaging Science Best Paper Prize, 2024.
  arXiv
  
  

• Distributed Estimation of Principal Eigenspaces.
  Jianqing Fan, Dong Wang, Kaizheng Wang, Ziwei Zhu \((\alpha\)-\(\beta)\)
  Annals of Statistics 47 (6): 3009-3031, 2019.
  arXiv
  
  

• Spectral Method and Regularized MLE are both Optimal for Top-K Ranking.
  Yuxin Chen, Jianqing Fan, Cong Ma, Kaizheng Wang \((\alpha\)-\(\beta)\)
  Annals of Statistics 47 (4): 2204-2235, 2019.
  arXiv
  
  

Funding

Support from the following funding sources is gratefully acknowledged:

• Columbia University SEAS GenAI in Education Initiative Grant. Duration: 2025 – 2026. Role: PI.

• National Science Foundation (NSF) Grant DMS-2515679. Title: “Adaptive Data Integration: Harnessing Commonality amidst Heterogeneity”. Duration: 2025 – 2028. Role: PI.

• National Institutes of Health (NIH) Grant 5R01AG087496. Title: “Statistical Framework for Unraveling Age-Dependent Genetic Landscape of Alzheimer's Disease and Related Dementias: Harnessing Large-Scale EHR and DNA-Biobank Integration”. Duration: 2024 – 2029. Role: Co-PI.

• National Science Foundation (NSF) Grant DMS-2210907. Title: “Statistical and Computational Tools for Analyzing High-Dimensional Heterogeneous Data”. Duration: 2022 – 2026. Role: PI.

• Columbia University Data Science Institute Seed Fund. Title: “Policy Evaluation with Transfer Learning: How to assess safety performance of self-driving cars in NYC?”. Duration: 2024 - 2025. Role: Co-PI.

Miscellaneous

My Erdös number is 3: Kaizheng Wang - Emmanuel Abbe - Noga Alon - Paul Erdös.