Zhiyang Xun 寻之扬

I'm a fourth-year Ph.D. student at the University of Texas at Austin, advised by Prof. Eric Price and Prof. David Zuckerman. My current research focuses on how to scalably supervise increasingly capable and complex AI.

My broader interests are in theoretical computer science and the foundations of machine learning, spanning diffusion models, randomized algorithms, and complexity theory. I enjoy using theory to understand the tractability and inherent limits of computational problems, especially those in modern AI.

Previously, I received my bachelor's degree from the ACM Honors Class at Shanghai Jiao Tong University. I'm grateful to have been mentored by Prof. Yijia Chen and Prof. Xi Chen during my undergraduate years.

I'm interning at Transluce in San Francisco this summer.

Email: zxun [at] cs.utexas.edu  ·  Google Scholar

Publications

Authors are ordered alphabetically unless they're not.

• Query Lower Bounds for Diffusion Sampling [arXiv]
  Zhiyang Xun and Eric Price
  ICLR 2026 Workshop on Deep Generative Models in Machine Learning. Outstanding Paper Award.
  
  
• Posterior Sampling by Combining Diffusion Models with Annealed Langevin Dynamics [arXiv]
  Zhiyang Xun, Shivam Gupta, and Eric Price
  Neural Information Processing Systems (NeurIPS 2025).
  
  
• Near-Optimal Averaging Samplers and Matrix Samplers [ECCC] [arXiv] [Oded's choice]
  Zhiyang Xun and David Zuckerman
  Computational Complexity Conference (CCC 2025).
  
  
• Query Complexity of Stochastic Minimum Vertex Cover [LIPIcs]
  Mahsa Derakhshan, Mohammad Saneian, and Zhiyang Xun
  Innovations in Theoretical Computer Science (ITCS 2025).
  
  
• Improved Sample Complexity Bounds for Diffusion Model Training [arXiv]
  Shivam Gupta, Aditya Parulekar, Eric Price, and Zhiyang Xun
  Neural Information Processing Systems (NeurIPS 2024).
  
  
• Spectral Guarantees for Adversarial Streaming PCA [arXiv]
  Eric Price and Zhiyang Xun
  Foundations of Computer Science (FOCS 2024).
  
  
• On Pigeonhole Principles and Ramsey in TFNP [ECCC] [arXiv]
  Siddhartha Jain, Jiawei Li, Robert Robere, and Zhiyang Xun
  Foundations of Computer Science (FOCS 2024).
  
  
• Diffusion Posterior Sampling is Computationally Intractable [arXiv]
  Shivam Gupta, Ajil Jalal, Aditya Parulekar, Eric Price, and Zhiyang Xun
  International Conference on Machine Learning (ICML 2024).
  
  
• On Algorithms Based on Finitely Many Homomorphism Counts [arXiv]
  Yijia Chen, Jörg Flum, Mingjun Liu, and Zhiyang Xun
  Information and Computation, 306 (2025): 105326.
  Conference version in Mathematical Foundations of Computer Science (MFCS 2022).
  
  

Teaching

• CS389r: Recursion and Induction, Spring 2025, TA, UT Austin
• CS1957: Algorithms, Spring 2022, TA, Shanghai Jiao Tong University
• CS301: Mathematical Logic, Fall 2020, TA, Shanghai Jiao Tong University
• CS151: C++ Programming (Honors), Fall 2019, TA, Shanghai Jiao Tong University