Haotian Lin

I am a Postdoctoral Scientist at Amazon.

My current work focuses on multimodal agentic post-training and workflows for conditional-based monitoring and prescription. Previously, my research centered on robust transfer learning, functional data learning, and differential privacy via kernel methods.

I received my Ph.D. in Statistics from Pennsylvania State University, working with Matthew Reimherr, and my B.S. in Statistics from USTC.

Google Scholar | Curriculum Vitae | Linkedin | htlin@amazon.com

Publications and Preprints

(* Equal Contribution)

ReSkill: Reconciling Skill Creation with Policy Optimization in Agentic RL
Zelin He*, Haotian Lin*, Boran Han, Wei Zhu, Haoyang Fang, Bernie Wang, Xuan Zhu, Runze Li, Matthew Reimherr
(Preprint)
[arXiv:2606.01619] [Project Website]

Discussion on ``INTACT: A method for integration of longitudinal physical activity data from multiple sources’’
Haotian Lin and Matthew Reimherr
Biometrics (Inivited discussion)
[Paper]

SenTSR-Bench: Thinking with Injected Knowledge for Time-Series Reasoning
Zelin He, Boran Han, Xiyuan Zhang, Shuai Zhang, Haotian Lin, Qi Zhu, Haoyang Fang, Danielle C. Maddix, Abdul Fatir Ansari, Akash Chandrayan, Abhinav Pradhan, Bernie Wang, Matthew Reimherr
International Conference on Artificial Intelligence and Statistics (AISTATS), 2026
[arXiv:2602.19455] [bibtex]

@article{he2026sentsr,
  title={SenTSR-Bench: Thinking with Injected Knowledge for Time-Series Reasoning},
  author={He, Zelin and Han, Boran and Zhang, Xiyuan and Zhang, Shuai and Lin, Haotian and Zhu, Qi and Fang, Haoyang and Maddix, Danielle C and Ansari, Abdul Fatir and Chandrayan, Akash and others},
  journal={arXiv preprint arXiv:2602.19455},
  year={2026}
}

Co-Regularization Enhances Knowledge Transfer in High Dimensions
Shuo-Shuo Liu*, Haotian Lin*, Matthew Reimherr, and Runze Li
Neural Information Processing Systems (NeurIPS), 2025
[Paper] [bibtex]

@article{liu2026co,
  title={Co-Regularization Enhances Knowledge Transfer in High Dimensions},
  author={Liu, Shuo Shuo and Lin, Haotian and Reimherr, Matthew and Li, Runze},
  journal={Advances in Neural Information Processing Systems},
  volume={38},
  pages={86176--86214},
  year={2026}
}

Fixed-Gaussian Spectral Algorithms: Minimax Optimal Rates for Misspecified Learning and Transfer
Haotian Lin and Matthew Reimherr
(Preprint)
[arXiv:2501.10870] [bibtex]

@article{lin2025model,
  title={Model-robust and adaptive-optimal transfer learning for tackling concept shifts in nonparametric regression},
  author={Lin, Haotian and Reimherr, Matthew},
  journal={arXiv preprint arXiv:2501.10870},
  year={2025}
}

Pure Differential Privacy for Functional Summaries with a Laplace-like Process
Haotian Lin and Matthew Reimherr
Journal of Machine Learning Research (JMLR), 2024
[Paper] [arXiv:2309.00125] [bibtex]

@article{lin2024pure,
  title={Pure differential privacy for functional summaries with a laplace-like process},
  author={Lin, Haotian and Reimherr, Matthew},
  journal={Journal of Machine Learning Research},
  volume={25},
  number={305},
  pages={1--50},
  year={2024}
}

Smoothness Adaptive Hypothesis Transfer Learning
Haotian Lin and Matthew Reimherr
International Conference on Machine Learning (ICML), 2024
[Paper] [arXiv:2402.14966] [bibtex]

@inproceedings{lin2024smoothness,
  title={Smoothness Adaptive Hypothesis Transfer Learning},
  author={Lin, Haotian and Reimherr, Matthew},
  booktitle={International Conference on Machine Learning},
  pages={30286--30316},
  year={2024},
  organization={PMLR}
}

On Hypothesis Transfer Learning in Functional Linear Models
Haotian Lin and Matthew Reimherr
International Conference on Machine Learning (ICML), 2024
[Paper] [arXiv:2206.04277] [Code] [bibtex]

@inproceedings{lin2024hypothesis,
  title={On Hypothesis Transfer Learning of Functional Linear Models},
  author={Lin, Haotian and Reimherr, Matthew},
  booktitle={International Conference on Machine Learning},
  pages={30252--30285},
  year={2024},
  organization={PMLR}
}