Xuanfei Ren

Xuanfei Ren

PhD student in Computer Science

Welcome to Xuanfei Ren’s Homepage!

I am Xuanfei Ren (任宣霏), a PhD student in Computer Science at the University of Wisconsin-Madison, advised by Prof. Tengyang Xie. I completed my Bachelor of Science in Mathematics (specialization in Probability and Statistics) from the University of Science and Technology of China (USTC) in 2024.

I believe that with dedicated research efforts, the five years of PhD study can be highly productive and meaningful.

📄 Download my CV (PDF)

Research Interests

Reinforcement learning (RL), from its theoretical foundations to its empirical scalability, lies at the core of my research. I am particularly interested in the following questions:

  • When is RL the right modeling framework? Determining whether RL is the right modeling framework is just as important as solving the resulting problem. I am interested in understanding when RL offers genuine advantages over simpler paradigms, especially in settings involving long-term credit assignment, sequential decision-making, and strategic exploration.

  • How can RL problems be solved efficiently? Once a problem is cast in the RL framework, I study algorithms that are both sample and computationally efficient, with an emphasis on rigorous theoretical guarantees alongside strong empirical performance.

  • How can RL be extended beyond standard MDPs? I am interested in extending RL beyond classical MDP formulations to richer, more open-ended settings. In collaboration with Ching-An Cheng (Google Research) and Allen Nie (Google DeepMind), I work on OpenTrace, an open-source framework for end-to-end training of AI agents. Our recent paper, POLCA, develops an efficient algorithm for generative optimization with LLMs. More broadly, I am excited about the role of LLMs in self-improving agentic systems, with applications ranging from kernel and code optimization to automated scientific discovery.

Recent News

  • [2026/03] We released POLCA, a search algorithm in OpenTrace for generative optimization with LLMs.

  • [2024/08] Started my PhD journey at University of Wisconsin-Madison, advised by Prof. Tengyang Xie.

  • [2024/05] Our paper “Optimal Batched Linear Bandits” gets accepted to ICML 2024. Joint work with Tianyuan Jin and Prof. Pan Xu.

Contact Me

If you are interested in discussing research or collaboration, or simply want to grab a coffee or chat, feel free to reach out. I always enjoy making new friends!