Selected papers (full list)

(* denotes alphabetical ordering or co-first author)

Theory of AI and deep learning

Licong Lin, Jingfeng Wu, Peter L. Bartlett. Improved Scaling Laws in Linear Regression via Data Reuse. Preprint, 2025+. arXiv

Licong Lin, Song Mei. A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics. Preprint, 2025+. arXiv

Kazusato Oko*, Licong Lin*, Yuhang Cai*, Song Mei*. A Statistical Theory of Contrastive Pre-training and Multimodal Generative AI. Preprint, 2025+. arXiv

Licong Lin, Jingfeng Wu, Sham M. Kakade, Peter L. Bartlett, Jason D. Lee. Scaling Laws in Linear Regression: Compute, Parameters, and Data. Advances in Neural Information Processing Systems (NeurIPS), 2024. arXiv

Licong Lin, Yu Bai, Song Mei. Transformers as Decision Makers: Provable In-Context Reinforcement Learning via Supervised Pretraining. International Conference on Learning Representations (ICLR), 2024. arXiv

Licong Lin, Edgar Dobriban. What Causes the Test Error? Going Beyond Bias-Variance via ANOVA. The Journal of Machine Learning Research, 2021. journal arXiv

AI safety alignment

Chongyu Fan*, Jiancheng Liu*, Licong Lin*, Jinghan Jia, Ruiqi Zhang, Song Mei, Sijia Liu. Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning. Preprint, 2024+. arXiv

Ruiqi Zhang*, Licong Lin*, Yu Bai, Song Mei. Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning. The First Conference on Language Modeling (COLM), 2024. arXiv

Statistical inference for adaptive/high-dim. data

Licong Lin*, Koulik Khamaru*, Martin J. Wainwright. Semi-parametric inference based on adaptively collected data. The Annals of Statistics, 2025. journal arXiv

Licong Lin*, Fangzhou Su*, Wenlong Mou, Peng Ding, Martin J. Wainwright. When is it worthwhile to jackknife? Breaking the quadratic barrier for Z-estimators. Preprint, 2024+. arXiv

Michael Celentano*, Zhou Fan*, Licong Lin*, and Song Mei*. Mean-field variational inference with the TAP free energy: Geometric and statistical properties in linear models. Preprint, 2023+. arXiv

Licong Lin, Mufang Ying, Suvrojit Ghosh, Koulik Khamaru, and Cun-Hui Zhang. Statistical Limits of Adaptive Linear Models: Low-Dimensional Estimation and Inference. Advances in Neural Information Processing Systems (NeurIPS), 2023. arXiv

Statistical learning

Ruiqi Zhang, Jingfeng Wu, Licong Lin, Peter L. Bartlett. Minimax Optimal Convergence of Gradient Descent in Logistic Regression via Large and Adaptive Stepsizes. Preprint, 2025+. arXiv

Licong Lin, Tijana Zrnic. Plug-in Performative Optimization. International Conference on Machine Learning (ICML), 2024. arXiv