Zhengyan Wan

I am a first-year PhD student at the University of California, Los Angeles (UCLA), where I work under the supervision of Prof. Guang Cheng. Previously, I received my Master's degree from the School of Statistics at East China Normal University, where I was advised by Prof. Fang Fang, and my Bachelor's degree in Statistics from the School of Mathematics at Jilin University.

Email / Github / Google Scholar

Publications

(* : Equal Contribution, ^† : Corresponding Author)

Corrected Samplers for Discrete Flow Models [Code]
Zhengyan Wan, Yidong Ouyang, Liyan Xie, Hongyuan Zha^†, Fang Fang^†, Guang Cheng
ICML, 2026

We propose two corrected samplers: time-corrected sampler and location-corrected sampler. We establish a one-step lower bound for the Euler sampler and non-asymptotic upper bounds for the proposed corrected samplers, showing that the location-corrected sampler achieves lower iteration complexity than the Euler sampler.

Error Analysis of Discrete Flow with Generator Matching
Zhengyan Wan, Yidong Ouyang, Qiang Yao, Liyan Xie, Fang Fang^†, Hongyuan Zha^†, Guang Cheng
ICML, 2026

We establish non-asymptotic error bounds for distribution error in discrete flow models by taking approximation error into account. Both fast and slow convergence rates are derived with and without the lower-boundedness condition, respectively.

Discrete Guidance Matching: Exact Guidance for Discrete Flow Matching [Code]
Zhengyan Wan*, Yidong Ouyang*, Liyan Xie, Fang Fang^†, Hongyuan Zha^†, Guang Cheng
ICLR, 2026

We derive the exact transition rate for the desired distribution given a learned discrete flow matching model, leading to guidance that only requires a single forward pass in each sampling step, significantly improving sampling efficiency.

Preprints

Acceleration of Diffusion Language Model through Discrete Average Generator
Yidong Ouyang, Zhengyan Wan, Themis Haris, Tian Tan, Liqian Peng, Henry Li, Ziqian Lin, Jianhang Chen, Maryam Karimzadehgan, Alec Go, George Michailidis

Iterative Exact Discrete Guidance for Energy-Based Sampling [Code]
Yuwen Qian, Yidong Ouyang, Zhengyan Wan, Hongyuan Zha

dFlowGRPO: Rate-Aware Policy Optimization for Discrete Flow Models [Code]
Zhengyan Wan, Yidong Ouyang, Panwen Hu, Qiang Sun

Training-Free Self-Correction for Multimodal Masked Diffusion Models [Code]
Yidong Ouyang, Panwen Hu, Zhengyan Wan, Zhe Wang, Liyan Xie, Dmitriy Bespalov, Ying Nian Wu, Guang Cheng, Hongyuan Zha, Qiang Sun

High-Dimensional Model Averaging via Cross-Validation [Code]
Zhengyan Wan, Fang Fang^†, Binyan Jiang

Website credits go to Jon Barron!