Yue Yu

Welcome to my personal website :wave:

prof_pic.jpg

Room 301, Swain East

729 E 3rd Street

Bloomington, IN, US, 47405

I’m Yue Yu, a PhD candidate in Statistical Science at Indiana University. I am fortunate to be jointly advised by Prof. Chunfeng Huang and Prof. David Crandall. I also work closely with Prof. Dongruo Zhou and Prof. Guanyu Hu.

My research broadly spans Machine Learning, Spatial Statistics, and Sports Analytics. My recent work in Machine Learning focuses on computationally efficient post-training and test-time methods for generative AI models.

Some of my past and current projects include:

  • Machine Learning: Continuous-Time and Multi-Task Reinforcement Learning, Test-Time Inference for LLMs, RL Post-Training for LLMs
  • Spatial Statistics: Deep Kriging on the Sphere
  • Sports Analytics: Causal Inference for basketball player injury and load management, Bayesian copula clustering of player types

Please feel free to reach out by email or LinkedIn message if you are interested in research collaborations or would like to connect for discussion! :fire:

News

Mar 27, 2026 One pre-print The Load Management Paradox: Correcting the Healthy-Worker Survivor Effect in NBA Injury Modeling has been posted on arXiv :basketball:
Dec 04, 2025 One pre-print On the Limits of Test-Time Compute: Sequential Reward Filtering for Better Inference has been posted on arXiv :wink:
Sep 25, 2025 I’m excited to share that I’ve been admitted to the M.S. in Computer Science program at Indiana University. One more degree to pick up along the Ph.D. journey! :feet: :stuck_out_tongue_winking_eye:
Aug 04, 2025 One pre-print Instance-Dependent Continuous-Time Reinforcement Learning via Maximum Likelihood Estimation has been posted on arXiv :relaxed:
May 07, 2025 Our paper Sample and Computationally Efficient Continuous-Time Reinforcement Learning with General Function Approximation has been accepted to the conference Uncertainty in Artificial Intelligence 2025. See you in July at Rio de Janeiro, Brazil :brazil: :smiley: