---

I'm a Research Scientist at Luma AI. Previously I was a Research Scientist at Adobe Research. I obtained my Ph.D. from the Department of Computer Science and Engineering, University at Buffalo, under the supervision of Prof. Jinhui Xu and Prof. Changyou Chen. I received my B.E. degree from Zhejiang University.

I worked as a Research Intern with Chunyuan Li (Microsoft), Ruiyi Zhang (Adobe), Bingchen Liu (ByteDance).

If you're excited about building multimodal intelligence, feel free to reach out.

---

• Uni-1
• Ray3 Modify
• Two papers accepted by ICLR 2025.
• One paper accepted by AAAI 2025.
• One paper accepted by WACV 2025.
• One paper accepted by EMNLP 2024.
• One paper accepted by CoLM 2024.
• One paper accepted by ACL 2024.
• Two papers accepted by CVPR 2024.
• One paper accepted by ICLR 2024.

---

• SUGAR: Subject-Driven Video Customization in a Zero-Shot Manner

  Yufan Zhou, Ruiyi Zhang, Jiuxiang Gu, Nanxuan Zhao, Jing Shi, Tong Sun.

  A zero-shot method for video customization, which can generate creative videos for user-input subject image, with desired style, color, texture, background required by user-input text.

  


• Toffee: Efficient Million-Scale Dataset Construction for Subject-Driven Text-to-Image Generation

  Yufan Zhou, Ruiyi Zhang, Kaizhi Zheng, Nanxuan Zhao, Jiuxiang Gu, Zichao Wang, Xin Eric Wang, Tong Sun.

  Efficient method to construct dataset for subject-driven T2I generation, which can save at least tens of thousands of GPU hours.


• Customization Assistant for Text-to-Image Generation

  Yufan Zhou, Ruiyi Zhang, Jiuxiang Gu, Tong Sun.

  IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2024.

  An assistant which can generate creative images for specific user-input subject along with text explanation and elaboration in 2-5 seconds, without any fine-tuning.


• Enhancing Detail Preservation for Customized Text-to-Image Generation: A Regularization-Free Approach

  Yufan Zhou, Ruiyi Zhang, Tong Sun, Jinhui Xu.

  A novel framework for customized text-to-image generation without the use of regularization.
  We can efficiently customize a large-scale text-to-image generation model on single GPU, with only one image provided by the user.


• 

  Shifted Diffusion for Text-to-image Generation

  Yufan Zhou, Bingchen Liu, Yizhe Zhu, Xiao Yang, Changyou Chen, Jinhui Xu.

  IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023.

  We propose a method termed Corgi, which can better generate image embeddings from text inside multimodal embedding space.
  It benefits both standard and language-free text-to-image generation. And yes, I do have a Corgi.


• LAFITE: Towards Language-Free Training for Text-to-Image Generation

  Yufan Zhou, Ruiyi Zhang, Changyou Chen, Chunyuan Li, Chris Tensmeyer, Tong Yu, Jiuxiang Gu, Jinhui Xu, Tong Sun.

  IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022.

  Our proposed work, Lafite, is the first work which can successfully train text-to-image generation model with image-only dataset.


• TiGAN: Text-Based Interactive Image Generation and Manipulation

  Yufan Zhou, Ruiyi Zhang, Jiuxiang Gu, Chris Tensmeyer, Tong Yu, Changyou Chen, Jinhui Xu, Tong Sun.

  AAAI conference on Artificial Intelligence (AAAI), 2022.


• Meta-Learning with Neural Tangent Kernels

  Yufan Zhou*, Zhenyi Wang*, Jiayi Xian, Changyou Chen, Jinhui Xu.

  International Conference on Learning Representations (ICLR), 2021.


• Learning Manifold Implicitly via Explicit Heat Kernel Learning

  Yufan Zhou, Changyou Chen, Jinhui Xu.

  Conference on Neural Information Processing Systems (NeurIPS), 2020.

---

• Conferences Program Committee/Reviewer: NeurIPS; ICML; ICLR; CVPR; AISTATS; AAAI; IJCAI; EMNLP; ECCV; ACL;
• Journal Reviewer: IEEE Transactions on Pattern Analysis and Machine Intelligence; IEEE Transactions on Neural Networks and Learning Systems; IEEE Transactions on Circuits and Systems for Video Technology;