avatar

Yongsheng Yu

Ph.D. Candidate University of Rochester

Fourth-year Ph.D. student in the VIStA Lab at the University of Rochester, advised by Albert Arendt Hopeman Professor Jiebo Luo. My research focuses on visual generation models and agentic editing.

first-authored papers at venues
CVPR ICCV NeurIPS ECCV AAAI
On the job market. I am currently seeking a full-time industry position. If you are interested in my work, I would welcome the opportunity to connect and would be grateful for any inquiries or referrals. Please feel free to contact me.

News

09/2026
PixelDiT2 and Aurora are accepted to NeurIPS 2026!
09/2026
I will serve as an Area Chair for CVPR 2027.
06/2026
PixelDiT is selected as a CVPR 2026 Best Paper Finalist!
05/2026
Our agentic unified video editing framework Aurora is available.
02/2026
PixelDiT is accepted to CVPR 2026 Oral
12/2025
I will serve as an Area Chair for ECCV 2026.
11/2025
RealUHR is accepted to AAAI 2026.
09/2025
PixPerfect is accepted to NeurIPS 2025.
06/2025
OmniPaint is accepted to ICCV 2025.
04/2025
32×-compressed latent SR model ZipIR report is available! (Completed Nov 2024, delayed release due to patent application)
09/2024
PromptFix is accepted to NeurIPS 2024.
01/2024
MaGIC is accepted to ICLR 2024.

Show more

Experience

NVIDIA
Research Intern · Santa Clara, US
June 2026 - Present
Developing next-generation DLSS (beyond DLSS 5)
Research Intern · Santa Clara, US
May 2025 - May 2026
Research on the PixelDiT series (PixelDiT, PixelDiT2)
Adobe Research
May 2024 - Apr 2025
Research Intern · San Jose, US
Microsoft Research Asia
May 2023 - Aug 2023
Research Intern · Beijing, China

Selected Publications

PixelDiT2 architecture
NeurIPS 2026GitHub stars
PixelDiT2: Representation-Grounded Pixel Diffusion Transformers
Yongsheng Yu, Wei Xiong, Yichen Sheng, Shiqiu Liu, Jiebo Luo
Annual Conference on Neural Information Processing Systems (NeurIPS), 2026
Project PDF Code ImageNet Model
PixelDiT
CVPR 2026 Oral Best Paper Finalist GitHub stars
PixelDiT: Pixel Diffusion Transformers for Image Generation
Yongsheng Yu, Wei Xiong, Weili Nie, Yichen Sheng, Shiqiu Liu, Jiebo Luo
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026, Oral & Best Paper Finalist
Project PDF Code ImageNet Model T2I Model
Aurora
NeurIPS 2026 GitHub stars
Aurora: Unified Video Editing with a Tool-Using Agent
Yongsheng Yu, Ziyun Zeng, Zhiyuan Xiao, Zhenghong Zhou, Hang Hua, Wei Xiong, Jiebo Luo
Annual Conference on Neural Information Processing Systems (NeurIPS), 2026
Project PDF Code Model Dataset
OmniPaint
ICCV 2025 GitHub stars
OmniPaint: Mastering Object-Oriented Editing via Disentangled Insertion-Removal Inpainting
Yongsheng Yu*†, Ziyun Zeng*, Haitian Zheng, Jiebo Luo
† Project lead
IEEE/CVF International Conference on Computer Vision (ICCV), 2025
Project PDF Code Model Benchmark
PromptFix
NeurIPS 2024 GitHub stars
PromptFix: You Prompt and We Fix the Photo
Yongsheng Yu*†, Ziyun Zeng*, Hang Hua, Jianlong Fu, Jiebo Luo
† Project lead
Annual Conference on Neural Information Processing Systems (NeurIPS), 2024
Project PDF Code Dataset
Unbiased Multi-Modality
ECCV 2022
Unbiased Multi-Modality Guidance for Image Inpainting
Yongsheng Yu, Dawei Du, Libo Zhang, Tiejian Luo
European Conference on Computer Vision (ECCV), 2022
PDF Code
High-Fidelity Inpainting
ECCV 2022
High-Fidelity Image Inpainting with GAN Inversion
Yongsheng Yu, Libo Zhang, Heng Fan, Tiejian Luo
European Conference on Computer Vision (ECCV), 2022
PDF

Services

Area ChairECCV 2026, CVPR 2027
Conference Reviewer 2026: CVPR, AAAI, NeurIPS
2025: CVPR, ICCV, ICLR, NeurIPS, AAAI, MM
2024: CVPR, ECCV, ICML, ICLR, NeurIPS, ICME, MM, WACV
2023: CVPR, ICCV, NeurIPS, WACV
Journal ReviewerIEEE TIP, IJCV, IEEE TMM