Research

All papers are available on Google Scholar

2025

• Statistical Impossibility and Possibility of Aligning LLMs with Human Preferences: From Condorcet Paradox to Nash Equilibrium. K. Liu, Q. Long, Z. Shi, W. Su, and J. Xiao.

• An Overview of Large Language Models for Statisticians. W. Ji, W. Yuan, E. Getzen, K. Cho, M. Jordan, S. Mei, J. Weston, W. Su, J. Xu, and L. Zhang.

2024

• Robust Detection of Watermarks for Large Language Models under Human Edits. X. Li, F. Ruan, H. Wang, Q. Long, and W. Su.

• The 2020 United States Decennial Census Is More Private Than You (Might) Think. B. Su, W. Su, and C. Wang. [code]

• Debiasing Watermarks for Large Language Models via Maximal Coupling. Y. Xie, X. Li, T. Mallick, W. Su, and R. Zhang.

• Magnetic Preference Optimization: Achieving Last-Iterate Convergence for Language Model Alignment. M. Wang, C. Ma, Q. Chen, L. Meng, Y. Han, J. Xiao, Z. Zhang, J. Huo, W. Su, and Y. Yang. International Conference on Learning Representations (ICLR), 2025.

• A Statistical Viewpoint on Differential Privacy: Hypothesis Testing, Representation and Blackwell’s Theorem. W. Su. Annual Review of Statistics and Its Application, 2025.

• A Law of Next-Token Prediction in Large Language Models. H. He and W. Su.

• Analysis of the ICML 2023 Ranking Data: Can Authors’ Opinions of Their Own Papers Assist Peer Review in Machine Learning? B. Su, J. Zhang, N. Collina, Y. Yan, D. Li, K. Cho, J. Fan, A. Roth, and W. Su.

• A Statistical Framework of Watermarks for Large Language Models: Pivot, Detection Efficiency and Optimal Rules. X. Li, F. Ruan, H. Wang, Q. Long, and W. Su. The Annals of Statistics, 2025.

• Fine-Tuning Attention Modules Only: Enhancing Weight Disentanglement in Task Arithmetic. R. Jin, B. Hou, J. Xiao, W. Su, and L. Shen. International Conference on Learning Representations (ICLR), 2025.

• Bridging the Gap: Rademacher Complexity in Robust and Standard Generalization. J. Xiao, R. Sun, Q. Long, and W. Su. Conference on Learning Theory (COLT), 2024.

• Tackling GenAI Copyright Issues: Originality Estimation and Genericization. H. Chiba-Okabe and W. Su. Scientific Reports, 2025.

• Towards Rationality in Language and Multimodal Agents: A Survey. B. Jiang, Y. Xie, X. Wang, Y. Yuan, Z. Hao, X. Bai, W. Su, C. Taylor, and T. Mallick. Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL), 2025.

• On the Algorithmic Bias of Aligning Large Language Models with RLHF: Preference Collapse and Matching Regularization. J. Xiao, Z. Li, X. Xie, E. Getzen, C. Fang, Q. Long, and W. Su.

• Neural Collapse Meets Differential Privacy: Curious Behaviors of NoisyGD with Near-Perfect Representation Learning. C. Wang, Y. Zhu, W. Su, and Y. Wang. International Conference on Machine Learning (ICML) (oral), 2024.

• An Economic Solution to Copyright Challenges of Generative AI. J. Wang, Z. Deng, H. Chiba-Okabe, B. Barak, and W. Su.

• Eliciting Honest Information from Authors Using Sequential Review. Y. Zhang, G. Schoenebeck, and W. Su. AAAI Conference on Artificial Intelligence (AAAI), 2024.

• Provable Multi-Party Reinforcement Learning with Diverse Human Feedback. H. Zhong, Z. Deng, W. Su, Z. Wu, and L. Zhang.

• WildfireGPT: Tailored Large Language Model for Wildfire Analysis. Y. Xie, B. Jiang, T. Mallick, J. Bergerson, J. Hutchison, D. Verner, J. Branham, M. Alexander, R. Ross, Y. Feng, L. Levy, W. Su, and C. Taylor.

• A Peek into Token Bias: Large Language Models Are Not Yet Genuine Reasoners. B. Jiang, Y. Xie, Z. Hao, X. Wang, T. Mallick, W. Su, C. Taylor, and D. Roth. Conference on Empirical Methods in Natural Language Processing (EMNLP), 2024.

• Shifted Interpolation for Differential Privacy. J. Bok, W. Su, and J. Altschuler. International Conference on Machine Learning (ICML), 2024.

2023

• What Should Data Science Education Do with Large Language Models. X. Tu, J. Zou, W. Su, and L. Zhang. Harvard Data Science Review, 2024.

• Isotonic Mechanism for Exponential Family Estimation in Machine Learning Peer Review. Y. Yan, W. Su, and J. Fan. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2025+.

• DP-HyPO: An Adaptive Private Framework for Hyperparameter Optimization. H. Wang, S. Gao, H. Zhang, W. Su, and M. Shen. Neural Information Processing Systems (NeurIPS), 2023.

• Advancing Differential Privacy: Where We Are Now and Future Directions for Real-World Deployment. R. Cummings, D. Desfontaines, D. Evans, R. Geambasu, Y. Huang, M. Jagielski, P. Kairouz, G. Kamath, S. Oh, O. Ohrimenko, N. Papernot, R. Rogers, M. Shen, S. Song, W. Su, A. Terzis, A. Thakurta, S. Vassilvitskii, Y. Wang, L. Xiong, S. Yekhanin, D. Yu, H. Zhang, and W. Zhang. Harvard Data Science Review, 2024.

• A Truth Serum for Eliciting Self-Evaluations in Scientific Reviews. J. Wu, H. Xu, Y. Guo, and W. Su.

• Reward Collapse in Aligning Large Language Models. Z. Song, T. Cai, J. Lee, and W. Su.

• Unified Enhancement of Privacy Bounds for Mixture Mechanisms via $f$-Differential Privacy. C. Wang, B. Su, J. Ye, R. Shokri, and W. Su. Neural Information Processing Systems (NeurIPS), 2023.

• The Implicit Regularization of Dynamical Stability in Stochastic Gradient Descent. L. Wu and W. Su. International Conference on Machine Learning (ICML), 2023.

2022

• A Law of Data Separation in Deep Learning. H. He and W. Su. Proceedings of the National Academy of Sciences (direct submission), 2023.

• On Quantum Speedups for Nonconvex Optimization via Quantum Tunneling Walks. Y. Liu, W. Su, and T. Li. Quantum, 2023.

• FIFA: Making Fairness More Generalizable in Classifiers Trained on Imbalanced Data. Z. Deng, J. Zhang, L. Zhang, T. Ye, Y. Coley, W. Su, and J. Zou. International Conference on Learning Representations (ICLR), 2023.

• Rock: Causal Inference Principles for Reasoning about Commonsense Causality. J. Zhang, H. Zhang, W. Su, and D. Roth. International Conference on Machine Learning (ICML), 2022.

• You Are the Best Reviewer of Your Own Papers: The Isotonic Mechanism. W. Su.

• Analytical Composition of Differential Privacy via the Edgeworth Accountant. H. Wang, S. Gao, H. Zhang, M. Shen, and W. Su.

• The Alignment Property of SGD Noise and How It Helps Select Flat Minima: A Stability Analysis. L. Wu, M. Wang, and W. Su. Neural Information Processing Systems (NeurIPS), 2022.

2021

• Envisioning Future Deep Learning Theories: Some Basic Concepts and Characteristics. W. Su. Science China Information Sciences, 2024.

• You Are the Best Reviewer of Your Own Papers: An Owner-Assisted Scoring Mechanism. W. Su. Neural Information Processing Systems (NeurIPS), 2021.

• Imitating Deep Learning Dynamics via Locally Elastic Stochastic Differential Equations. J. Zhang, H. Wang, and W. Su. Neural Information Processing Systems (NeurIPS), 2021.

• An Unconstrained Layer-Peeled Perspective on Neural Collapse. W. Ji, Y. Lu, Y. Zhang, Z. Deng, and W. Su. International Conference on Learning Representations (ICLR), 2022.

• Characterizing the SLOPE Trade-off: A Variational Perspective and the Donoho-Tanner Limit. Z. Bu, J. Klusowski, C. Rush, and W. Su. The Annals of Statistics, 2023.

• Weighted Training for Cross-Task Learning. S. Chen, K. Crammer, H. He, D. Roth, and W. Su. International Conference on Learning Representations (ICLR) (oral), 2022.

• A Central Limit Theorem for Differentially Private Query Answering. J. Dong, W. Su, and L. Zhang. Neural Information Processing Systems (NeurIPS), 2021.

• Minimax Estimation for Personalized Federated Learning: An Alternative Between FedAvg and Local Training? S. Chen, Q. Zheng, Q. Long, and W. Su. Journal of Machine Learning Research, 2023.

• Rejoinder: Gaussian Differential Privacy. J. Dong, A. Roth, and W. Su. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2022.

• Exploring Deep Neural Networks via Layer-Peeled Model: Minority Collapse in Imbalanced Training. C. Fang, H. He, Q. Long, and W. Su. Proceedings of the National Academy of Sciences (direct submission), 2021.

• Oneshot Differentially Private Top-$k$ Selection. G. Qiao, W. Su, and L. Zhang. International Conference on Machine Learning (ICML), 2021.

• Federated $f$-Differential Privacy. Q. Zheng, S. Chen, Q. Long, and W. Su. International Conference on Artificial Intelligence and Statistics (AISTATS), 2021.

2020

• A Power Analysis for Model-X Knockoffs with $\ell_p$-Regularized Statistics. A. Weinstein, W. Su, M. Bogdan, R. Barber, and E. Candès. The Annals of Statistics, 2023.

• Precise High-Dimensional Asymptotics for Quantifying Heterogeneous Transfers. F. Yang, H. Zhang, S. Wu, C. Ré, and W. Su.

• Label-Aware Neural Tangent Kernel: Toward Better Generalization and Local Elasticity. S. Chen, H. He, and W. Su. Neural Information Processing Systems (NeurIPS), 2020.

• Benign Overfitting and Noisy Features. Z. Li, W. Su, and D. Sejdinovic. Journal of the American Statistical Association, 2023.

• The Complete Lasso Tradeoff Diagram. H. Wang, Y. Yang, Z. Bu, and W. Su. Neural Information Processing Systems (NeurIPS), 2020.

• The Price of Competition: Effect Size Heterogeneity Matters in High Dimensions. H. Wang, Y. Yang, and W. Su. IEEE Transactions on Information Theory, 2022.

• On Learning Rates and Schrödinger Operators. B. Shi, W. Su, and M. Jordan. Journal of Machine Learning Research, 2023.

• Sharp Composition Bounds for Gaussian Differential Privacy via Edgeworth Expansion. Q. Zheng, J. Dong, Q. Long, and W. Su. International Conference on Machine Learning (ICML), 2020.

• Towards Understanding the Dynamics of the First-Order Adversaries. Z. Deng, H. He, J. Huang, and W. Su. International Conference on Machine Learning (ICML), 2020.

• Toward Better Generalization Bounds with Locally Elastic Stability. Z. Deng, H. He, and W. Su. International Conference on Machine Learning (ICML), 2021.

2019

• The Local Elasticity of Neural Networks. H. He and W. Su. International Conference on Learning Representations (ICLR), 2020.

• Algorithmic Analysis and Statistical Estimation of SLOPE via Approximate Message Passing. Z. Bu, J. Klusowski, C. Rush, and W. Su. IEEE Transactions on Information Theory, 2020.

• Gaussian Differential Privacy. J. Dong, A. Roth, and W. Su. Journal of the Royal Statistical Society: Series B (Statistical Methodology) (with Discussion), 2022. [full version]

• Deep Learning with Gaussian Differential Privacy. Z. Bu, J. Dong, Q. Long, and W. Su. Harvard Data Science Review, 2020. [code]

• Robust Learning Rate Selection for Stochastic Optimization via Splitting Diagnostic. M. Sordello, N. Dalmasso, H. He, and W. Su. Transactions on Machine Learning Research (TMLR), 2024.

• Acceleration via Symplectic Discretization of High-Resolution Differential Equations. B. Shi, S. Du, W. Su, and M. Jordan. Neural Information Processing Systems (NeurIPS), 2019.

2018

• Understanding the Acceleration Phenomenon via High-Resolution Differential Equations. B. Shi, S. Du, M. Jordan, and W. Su. Mathematical Programming, 2022.

• The FDR-Linking Theorem. W. Su.

• Differentially Private False Discovery Rate Control. C. Dwork, W. Su, and L. Zhang. Journal of Privacy and Confidentiality, 2021.

• Robust Inference Under Heteroskedasticity via the Hadamard Estimator. E. Dobriban, W. Su, Y. Yang, and Z. Zhang.

• HiGrad: Uncertainty Quantification for Online Learning and Stochastic Approximation. W. Su and Y. Zhu. Journal of Machine Learning Research, 2023. [code]

• Assumption Lean Regression. R. Berk, A. Buja, L. Brown, E. George, A. Kuchibhotla, W. Su, and L. Zhao. The American Statistician, 2021.

2017

• Statistical Inference for the Population Landscape via Moment Adjusted Stochastic Gradients. T. Liang and W. Su. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 2019.

• Multiple Testing When Many $p$-Values Are Uniformly Conservative, with Application to Testing Qualitative Interaction in Educational Interventions. Q. Zhao, D. Small, and W. Su. Journal of the American Statistical Association, 2019.

• When Is the First Spurious Variable Selected by Sequential Regression Procedures? W. Su. Biometrika, 2018.

2016

• Detecting Multiple Replicating Signals Using Adaptive Filtering Procedures. J. Wang, L. Gui, W. Su, C. Sabatti, and A. Owen. The Annals of Statistics, 2022.

2015

• Group SLOPE—Adaptive Selection of Groups of Predictors. D. Brzyski, A. Gossmann, W. Su, and M. Bogdan. Journal of the American Statistical Association, 2019. [code]

• False Discoveries Occur Early on the Lasso Path. W. Su, M. Bogdan, and E. Candès. The Annals of Statistics, 2017. [appendix]

• Familywise Error Rate Control via Knockoffs. L. Janson and W. Su. Electronic Journal of Statistics, 2016.

• A Differential Equation for Modeling Nesterov’s Accelerated Gradient Method: Theory and Insights. W. Su, S. Boyd, and E. Candès. Journal of Machine Learning Research, 2016.

• SLOPE Is Adaptive to Unknown Sparsity and Asymptotically Minimax. W. Su and E. Candès. The Annals of Statistics, 2016. [appendix]

2014

• SLOPE—Adaptive Variable Selection via Convex Optimization. M. Bogdan, E. van den Berg, C. Sabatti, W. Su, and E. Candès. The Annals of Applied Statistics, 2015. [code]