About

The person *NOT* wearing a blue cravat is me

I'm Yuanzhi Zhu and I'm a PhD student at École Polytechnique, supervised by Prof. Vicky Kalogeiton, Prof. Stéphane Lathuilière and Prof. Xi Wang. Previously I was a Master's student at ETH Zürich, where I have the privilege of being supervised by Prof. Kai Zhang. I also have the privilege of working as a RA with Prof. Qiang Liu and interned with Hanshu Yan during my master's.

firstname.lastname@polytechnique.edu

Viva la post-training! — pre-training fits the data; the rest is craft.

Selected Publications

Yuanzhi Zhu, Xi Wang, Stéphane Lathuilière, Vicky Kalogeiton.
Diffusion Reinforcement Learning via Centered Reward Distillation, Arxiv. arxiv: 2603.14128 [paper] [arxiv] [code] [blog]
Yuanzhi Zhu, Xi Wang, Stéphane Lathuilière, Vicky Kalogeiton.
Soft-Di[M]O: Improving One-Step Discrete Image Generation with Soft Embeddings, ICLR (2026). arxiv: 2509.22925 [paper] [arxiv] [code] [slides] [poster]
Yuanzhi Zhu, Xi Wang, Stéphane Lathuilière, Vicky Kalogeiton.
Di[M]O: Distilling Masked Diffusion Models into One-Step Generator, ICCV (2025). arxiv: 2503.15457 [paper] [arxiv] [code] [slides] [poster]
Yuanzhi Zhu*, Ruiqing Wang*, Shilin Lu, Junnan Li, Hanshu Yan, Kai Zhang.
OFTSR: One-Step Flow for Image Super-Resolution with Tunable Fidelity-Realism Trade-offs, ICLR (2026). arxiv: 2412.09465 [paper] [arxiv] [code] [slides] [poster]
Yuanzhi Zhu, Xingchao Liu, Qiang Liu.
SlimFlow: Training Smaller One-Step Diffusion Models with Rectified Flow, ECCV (2024). arxiv: 2407.12718 [paper] [arxiv] [code] [poster]
Yuanzhi Zhu, Kai Zhang, Jingyun Liang, Jiezhang Cao, Bihan Wen, Radu Timofte, Luc Van Gool.
Denoising Diffusion Models for Plug-and-Play Image Restoration, CVPRW (2023). arxiv: 2305.08995 [paper] [arxiv] [code] [slides] [poster]