The directions below connect theory with systems that make decisions under uncertainty. Each block highlights a few papers; open its publication list for the full set in that area. Browse the complete bibliography by publication type.
Stochastic Optimization
Global structure and biased gradients
I study when hidden convexity, convex recursions, or other structure make global solutions possible in seemingly nonconvex decision problems. I also develop stochastic methods for bilevel, compositional, and multistage optimization when gradient estimates are biased or costly.
Structured nonconvexity
Global guarantees from hidden convexity and benign landscapes in policy and causal optimization.
Biased stochastic gradients
Methods for bilevel, compositional, and multistage problems with biased or costly gradient estimates.
View all 15 publications
Buse Şen, Yifan Hu, Daniel Kuhn
2026 · Under reviewTalk on multistage optimization at the SIAM Conference on Optimization, 2026.
Zhichun Yang, Ying Lin, Daniel Kuhn, Yifan Hu, Man-Chung Yue
2025Yifan Hu, Jie Wang, Xin Chen, Niao He
2024 · Under minor revision at Operations ResearchConference version: “On the Bias-Variance-Cost Tradeoff of Stochastic Optimization,” NeurIPS 2021.
Ilyas Fatkhullin, Niao He, Yifan Hu
SIAM Journal on Optimization · 35(4): 2544–2571 · 2025Presented at the 25th International Symposium on Mathematical Programming, 2024.
Xin Chen, Niao He, Yifan Hu, Zikun Ye
Operations Research · 73(2): 704–719 · 2025Selected for oral presentations at MSOM Supply Chain SIG 2022 and RMP Conference 2022. Further talks at MIT, the University of Sydney, Frankfurt School, and INFORMS.
Yifan Hu, Xin Chen, Niao He
SIAM Journal on Optimization · 2020Xin Chen, Yifan Hu, Minda Zhao
2024 · Under minor revision at Operations ResearchZhenyu Wang*, Yifan Hu*, Peter Bühlmann, Zijian Guo
2024 · Under reviewPresented at ICSDS (December 2025), the UT Dallas Statistics seminar (2025–26), and the Joint Conference on Statistics and Data Science (July 2025).
Siqi Zhang*, Yifan Hu*, Liang Zhang, Niao He
AISTATS · 2024Yifan Hu, Jie Wang, Yao Xie, Andreas Krause, Daniel Kuhn
NeurIPS · 2023Talks at Rutgers, Cornell ORIE, INFORMS, and SIAM Optimization, 2023–25.
Yifan Hu, Xin Chen, Niao He
NeurIPS · 2021Presented at NeurIPS, ETH Zurich, and INFORMS, 2021.
Yifan Hu*, Siqi Zhang*, Xin Chen, Niao He
NeurIPS · 2020Siqi Zhang, Yifan Hu
ICLR Blogposts · 2025Yifan Hu
Encyclopedia of Optimization, 3rd edition · 2024Xuxing Chen*, Abhishek Roy*, Yifan Hu, Krishna Balasubramanian
NeurIPS · 2024
Reinforcement Learning
Robust and aligned decisions
I develop methods for sequential decisions that remain reliable under model uncertainty. This includes policy optimization landscapes, robust Markov decision processes, and systems in which an agent’s incentives must align with a designer’s goal.
Policy structure
Understand when nonconvex policy methods reach globally good decisions.
Robustness and incentives
Design contextual bilevel methods, robust control, and preference learning under uncertainty.
View all 5 publications
Mengmeng Li, Yifan Hu, Daniel Kuhn, Yan Li
2026Xin Chen, Yifan Hu, Minda Zhao
2024 · Under minor revision at Operations ResearchVinzenz Thoma*, Barna Pasztor*, Andreas Krause, Giorgia Ramponi, Yifan Hu
NeurIPS · 2024Preliminary version accepted to the ICML Agentic Markets Workshop, 2024. Talks at the Isaac Newton Institute, University of Cambridge, and Rutgers ISE seminar.
Shyam Sundhar Ramesh, Yifan Hu, Iason Chaimalas, Viraj Mehta, Pier Giuseppe Sessa, Haitham Bou Ammar, Ilija Bogunovic
NeurIPS · 2024Shyam Sundhar Ramesh, Pier Giuseppe Sessa, Yifan Hu, Andreas Krause, Ilija Bogunovic
AISTATS · 2024
Causality & Statistical Learning
Inference with tractable structure
Causal questions and distribution shifts often lead to difficult estimation problems. I formulate them as optimization problems whose structure supports efficient computation, statistical guarantees, and fairer learning across groups.
Optimization for Causality
Model causal invariance learning for causal discovery and causal inference via instrumental variables as optimization problems.
Robust Inference and Fairness
Study group distributional robustness, reliable statistical inference, and fairness in federated learning.
View all 4 publications
Zijian Guo, Zhenyu Wang, Yifan Hu, Francis Bach
2025 · Under reviewZhenyu Wang*, Yifan Hu*, Peter Bühlmann, Zijian Guo
2024 · Under reviewPresented at ICSDS (December 2025), the UT Dallas Statistics seminar (2025–26), and the Joint Conference on Statistics and Data Science (July 2025).
Yves Rychener, Daniel Kuhn, Yifan Hu
AISTATS · 2025Xuxing Chen*, Abhishek Roy*, Yifan Hu, Krishna Balasubramanian
NeurIPS · 2024
Learning Systems
Reliable models at scale
I use optimization and statistical ideas to allocate test-time computation, improve distillation, and align learning systems to diverse preferences.
Test-time computation
Allocate rollout and voting budgets where an extra token brings the most value.
Alignment and distillation
Select useful supervision and combine diverse preference objectives.
View all 4 publications
Zhenyu Wang, Xiaozhi Zhu, Yifan Hu
2026 · NeurIPS FLLMPT oral presentationSelected for an oral presentation at the NeurIPS 2026 FLLMPT Workshop; also accepted to the NeurIPS 2026 MLxOR, TTCL, and RAAAI workshops.
Zhenyu Wang, Tianze Wang, Linjun Zhang, Yifan Hu
2026Tianze Wang, Dongnan Gui, Yifan Hu, Shuhang Lin, Linjun Zhang
ICML · 2025Shyam Sundhar Ramesh, Yifan Hu, Iason Chaimalas, Viraj Mehta, Pier Giuseppe Sessa, Haitham Bou Ammar, Ilija Bogunovic
NeurIPS · 2024