Tianyuan Jin
I am an assistant professor at the Data Science and Analytics Thrust (DSA) at the Hong Kong University of Science and Technology (Guangzhou). Prior to joining HKUST(GZ), I was a Research Fellow at NUS.
I obtained my Ph.D. in Computer Science from the National University of Singapore in 2024.
My research is centered around Machine Learning, with broad interests in the areas of Online learning, Bandits, Reinforcement Learning, Large Language Models, and their applications to real-world problems.
I am always excited to work with motivated, curious, and self-driven students
who share an interest in our research. Before reaching out, please read at
least one paper from our group, or a closely related paper, to get a sense of
the problems we study and the methods we use.
If you find our research aligns with your interests, please send me an email
including:
Due to the large number of inquiries I receive, I am unfortunately unable to
respond to every email. Priority will be given to students who demonstrate a
clear understanding of our work and a genuine, specific interest in the
research directions we pursue.
[Prospective Students/Postdoc/Research Assistant/Intern]
[CV] [Google Scholar] [Email]
Contact
Email: tianyuan1044 [at] gmail [dot] com
Recent papers
* denotes equal contribution. + denotes alphabetical author order.
- Avoiding exp(k*) Scaling for Thompson Sampling in Combinatorial Semi-Bandits: From Multiple Seeds to a Single Seed COLT 2026.
- Best Arm Identification with Minimal Regret JMLR 2026.
- Asymptotically and Minimax Optimal Regret Bounds for Multi-Armed Bandits with Abstention TMLR 2026.
- Breaking the Total Variance Barrier: Sharp Sample Complexity for Linear Heteroscedastic Bandits with Fixed Action Set ICLR 2026.
- SteerConf: Steering LLMs for Confidence Elicitation NeurIPS 2025.
- Breaking the log(1/Δ2) Barrier: Better Batched Best Arm Identification with Adaptive Grids ICLR 2025.
- Optimal Batched Best Arm Identification NeurIPS 2024.
- Sparsity-Agnostic Linear Bandits with Adaptive Adversaries NeurIPS 2024.
- Finite-Time Frequentist Regret Bounds of Multi-Agent Thompson Sampling on Sparse Hypergraphs AAAI 2024. Oral presentation.
Publications
2026
- Avoiding exp(k*) Scaling for Thompson Sampling in Combinatorial Semi-Bandits: From Multiple Seeds to a Single Seed Annual Conference on Learning Theory (COLT), 2026.
- Best Arm Identification with Minimal Regret Journal of Machine Learning Research, 2026+.
- Asymptotically and Minimax Optimal Regret Bounds for Multi-Armed Bandits with Abstention Transactions on Machine Learning Research, 2026.
- Breaking the Total Variance Barrier: Sharp Sample Complexity for Linear Heteroscedastic Bandits with Fixed Action Set International Conference on Learning Representations (ICLR), 2026.
2025
- SteerConf: Steering LLMs for Confidence Elicitation Advances in Neural Information Processing Systems (NeurIPS), 2025.
- Breaking the log(1/Δ2) Barrier: Better Batched Best Arm Identification with Adaptive Grids International Conference on Learning Representations (ICLR), 2025.
2024
- Optimal Batched Best Arm Identification Advances in Neural Information Processing Systems (NeurIPS), 2024.
- Sparsity-Agnostic Linear Bandits with Adaptive Adversaries Advances in Neural Information Processing Systems (NeurIPS), 2024.
- Finite-Time Frequentist Regret Bounds of Multi-Agent Thompson Sampling on Sparse Hypergraphs AAAI Conference on Artificial Intelligence (AAAI), 2024. Oral presentation.
- Optimal Batched Linear Bandits International Conference on Machine Learning (ICML), 2024.
2023
- Thompson Sampling with Less Exploration is Fast and Optimal International Conference on Machine Learning (ICML), 2023.
2022
- Finite-Time Regret of Thompson Sampling Algorithms for Exponential Family Multi-Armed Bandits Advances in Neural Information Processing Systems (NeurIPS), 2022.
2021
- MOTS: Minimax Optimal Thompson Sampling International Conference on Machine Learning (ICML), 2021.
- Optimal Streaming Algorithms for Multi-Armed Bandits International Conference on Machine Learning (ICML), 2021.
- Almost Optimal Anytime Algorithm for Batched Multi-Armed Bandits International Conference on Machine Learning (ICML), 2021.
- Double Explore-then-Commit: Asymptotic Optimality and Beyond Conference on Learning Theory (COLT), 2021.
- Unconstrained Submodular Maximization with Modular Costs: Tight Approximation and Application to Profit Maximization Proceedings of the VLDB Endowment (PVLDB), 2021.
2020 and earlier
- Realtime Index-Free Single Source SimRank Processing on Web-Scale Graphs Proceedings of the VLDB Endowment (PVLDB), 2020.
- Realtime Top-k Personalized PageRank over Large Graphs on GPUs Proceedings of the VLDB Endowment (PVLDB), 2020.
- Efficient Pure Exploration in Adaptive Round Model Advances in Neural Information Processing Systems (NeurIPS), 2019.
- Tracking Top-k Influential Users with Relative Errors ACM International Conference on Information and Knowledge Management (CIKM), 2019.
- Maximizing the Effect of Information Adoption: A General Framework SIAM International Conference on Data Mining (SDM), 2018.
Awards
- Dean's Graduate Research Excellence Award, 2024.
- Google PhD Fellowship, 2021.
Professional service