Yunhao Shui税昀昊

AI Researcher & Engineer · Generative Models & Preference Alignment

I build and study image & video generative models, human preference alignment, and 3D reconstruction, turning research ideas into things that run. At Mizzen, I build voice agents for research interviews and evaluate how well they follow up on participants’ answers.

我研究图像与视频生成、人类偏好对齐和三维重建。在 Mizzen，我开发用于用户访谈的语音智能体，评估它们能否根据受访者的回答提出恰当的追问。这里收录了我的论文、开源项目和工作经历。

Trust the process.

Here are my links

GitHub Google Scholar 知乎小红书 Email CV

My research & skills

Image Diffusion ModelsVideo GenerationHuman Preference Alignment3D Gaussian SplattingNeural Radiance Fields3D ReconstructionPythonPyTorchCUDADiffusion ModelsComputer VisionDistributed TrainingLinuxGit

Here are my publications

Generated and real image examples ordered by HPSv3 preference score

HPSv3: Towards Wide-Spectrum Human Preference

ICCV 2025

Yuhang Ma*, Yunhao Shui*, Xiaoshi Wu, Keqiang Sun, Hongsheng Li

* Equal contribution

Code

Image generations compared without and with Negative Preference Optimization

Diffusion-NPO: Negative Preference Optimization for Better Preference Aligned Generation of Diffusion Models

ICLR 2025

Fu-Yun Wang, Yunhao Shui, Jingtan Piao, Keqiang Sun, Hongsheng Li

Paper

FlameGS pipeline from ray tracing initialization to 3D Gaussian rendering and reprojection loss

FlameGS: Reconstructing Flame Light Fields via Gaussian Splatting

SPNNA 2024

Yunhao Shui, Fuhao Zhang, Can Gao, Hao Xue, Zhiyin Ma, Gang Xun, Xuesong Li

arXiv

ECNet results generated from skeleton, facial landmark, and sketch conditions

ECNet: Effective Controllable Text-to-Image Diffusion Models

Computational Visual Media 2025

Sicheng Li, Zhixin Lai, Keqiang Sun, Xiaoshi Wu, Yunhao Shui, Feng Qiu, Haoran Xie, Kazunori Miyata

arXiv

Here are my projects

HPSv3

Wide-spectrum human preference score model for image generation.

MoviiGen 1.1

Towards cinematic-quality video generative models.

RaCig

RAG-based character-consistent story image generation.

My GitHub activity

@xilanhua12138 GitHub contribution graph

Here is my experience

Here is my education

Master's Degree

Shanghai Jiao Tong University

Generative models, preference alignment & 3D vision

2023.09 - 2026.03

Bachelor's Degree

Shanghai Jiao Tong University

2019.09 - 2023.09

My awards & honors

National Graduate Mathematical Modeling Competition: National First Prize (Top 1%), Huawei Second Prize (2023)
National College Student Mathematical Modeling Competition: National First Prize (2021)
CODALAB AutoCast Competition: 5th globally (2023)
SJTU University-level Scholarship, Class B (2020) & Class C (2021)
Junzheng Scholar (2021)

Where my visitors come from