Mingda WAN

Short Bio

I am an incoming Ph.D. at Westlake University, where I am fortunate to be advised by Prof. Tao LIN. My research interests lie at the intersection of agentic learning and generative modeling.

Currently, my research focuses on automated theorem proving, autoresearch, and the advancement of agentic rl&sft.

Email / Google Scholar / OpenReview

Selected Publications

* denotes alphabetical order.

Mind the Drift: Advantage Shaping with Future-Drift for Multi-Turn Agentic Reinforcement Learning
Bowen Ding, Mingda Wan, Yuhan Chen, Jiyao Yuan, Qi Zhu, Futing Wang, Heyuan Deng, Fei Mi, Lifeng Shang, Tao Lin
Under review at EMNLP 2026
paper

Unraveling the Smoothness Properties of Diffusion Models: A Gaussian Mixture Perspective
Yingyu Liang*, Zhizhou Sha*, Zhenmei Shi*, Zhao Song*, Mingda Wan*, and Yufa Zhou*
ICCV 2025
paper

SEHG: Bridging Interpretability and Prediction in Self-Explainable Heterogeneous Graph Neural Networks
Zhenhua Huang, Wenhao Zhou, YuFeng Li, Xiuyang Wu, Chengpei Xu, Junfeng Fang, Zhaohong Jia, Linyuan Lü, Feng Xia (Acknowledgements, initiative waived first author)
WWW 2025 Oral
paper

NRFlow: Towards Noise-Robust Generative Modeling via High-Order Flow Matching
Bo Chen*, Chengyue Gong*, Xiaoyu Li*, Yingyu Liang*, Zhizhou Sha*, Zhenmei Shi*, Zhao Song*, Mingda Wan*, and Xugang Ye*
UAI 2025
paper

Force Matching with Relativistic Constraints: A Physics-Inspired Approach to Stable and Efficient Generative Modeling
Yang Cao*, Bo Chen*, Xiaoyu Li*, Yingyu Liang*, Zhizhou Sha*, Zhenmei Shi*, Zhao Song* and Mingda Wan*
CIKM 2025
paper

Attention Scheme Inspired Softmax Regression
Zhihang Li*, Zhizhou Sha*, Zhao Song* and Mingda Wan*
ICLR 2025 Workshop
paper

An Improved Sample Complexity for Rank-1 Matrix Sensing
Zhihang Li*, Zhizhou Sha*, Zhao Song* and Mingda Wan*
ICLR 2025 Workshop
paper

High-Order Matching for One-Step Shortcut Diffusion Models
Bo Chen*, Chengyue Gong*, Xiaoyu Li*, Yingyu Liang*, Zhizhou Sha*, Zhenmei Shi*, Zhao Song* and Mingda Wan*
ICLR 2025 Workshop
paper

Theoretical Constraints on the Expressive Power of RoPE-based Tensor Attention Transformers
Xiaoyu Li*, Yingyu Liang*, Zhenmei Shi*, Zhao Song* and Mingda Wan*
arXiv preprint arXiv:2412.18040
paper

HOFAR: High-Order Augmentation of Flow Autoregressive Transformers
Yingyu Liang*, Zhizhou Sha*, Zhenmei Shi*, Zhao Song* and Mingda Wan*
arXiv preprint arXiv:2503.08032
paper

Academic Service

External Reviewer
Research Grants Council (Hong Kong)

Reviewer
UAI 2026
ICCV 2025
KDD 2024

Hobbies

During my spare time, you can always find me jogging, swimming, and fitness. I am also obsessed with classical novels, cinema, and progressive rock.

Design and source code from Jon Barron.