Yukun Zhou

Yukun Zhou

GigaAI

I am now an algorithm researcher at an AI startup GigaAI, working under the supervision of Zheng Zhu and Xinze Chen. In June 2025, I obtained my Bachelor’s degree from Xidian University. During my undergraduate studies, I worked as a research intern under the guidance of Kai Wang in NUS-HPC-AI-Lab. I also interned at the DataGrand with Dr. Nuo Xu.

My research interests include world models, reinforcement learning and embodied intelligence. I hope to have the opportunity to explore the integration of world models for comprehensive environmental data and reinforcement models for intelligent decision-making, and to make modest contributions to the development of AI systems.

I am exploring potential PhD opportunities in my field of interest. Feel free to reach out at email!

📝 Selected Public Work

Publications

CVPR 2025
sym

HumanDreamer: Generating Controllable Human-Motion Videos via Decoupled Generation

Boyuan Wang, et al, Yukun Zhou, et al.

Project   Dataset   Paper

  • HumanDreamer is a decoupled human video generation framework that first generates diverse poses from text prompts and then leverages these poses to generate human-motion videos.
CVPR 2025
sym

A Closer Look at Time Steps is Worthy of Triple Speed-Up for Diffusion Model Training

Kai Wang, Mingjia Shi, Yukun Zhou, et al.

Project   Code   Paper

  • As a plug-and-play and architectureagnostic approach, SpeeD consistently achieves 3× acceleration across various diffusion architectures, datasets, and tasks.
ECCV 2024
sym

Dataset Growth

Ziheng Qin, Zhaopan Xu, Yukun Zhou, et al.

Code   Paper

  • InfoGrowth can improve data quality/efficiency on both single modal and multi-modal tasks, with an efficient and scalable design.

Preprints

Preprint
sym

GigaVideo-1: Advancing Video Generation via Automatic Feedback with 4 GPU-Hours Fine-Tuning

Xiaoyi Bao, et al, Yukun Zhou, et al.

Project   Paper

  • GigaVideo-1 consistently improves performance on almost all the dimensions with an average gain of ~4% using only 4 GPU-hours.
Preprint
sym

Neural Network Parameter Diffusion

Kai Wang, Zhaopan Xu, Yukun Zhou, et al.

Code   Paper

  • We demonstrate that diffusion models can also generate high-performing neural network parameters.

Projects

Open-Source Project
sym

Open-Sora: Democratizing Efficient Video Production for All

Project participant as intern of Luchen Technology.

Code   Paper   Model   Dataset

  • Open-Sora is an open-source video generation model designed to produce high-fidelity video content. We democratizes full access to all the training/inference/data preparation codes as well as model weights

📖 Educations

  • 2021.09 - 2025.06, School of Computer Science and Technology, Xidian University.

💻 Internships