Zongze Wu
Scholar

Zongze Wu

Google Scholar ID: V8FwQGkAAAAJ
Adobe Research
GenAImulti-modality generation/editing
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
2,291
 
H-index
7
 
i10-index
6
 
Publications
12
 
Co-authors
0
 
Contact
Emailzongze.wu@mail.huji.ac.ilCVOpen ↗TwitterOpen ↗GitHubOpen ↗
Publications
5 items
EditStream: A Unified Autoregressive Framework for Interactive Video Generation and Editing
2026
Cited
0
Improved Baselines with Representation Autoencoders
2026
Cited
0
End-to-End Training for Unified Tokenization and Latent Denoising
2026
Cited
0
Causality in Video Diffusers is Separable from Denoising
2026
Cited
0
Rethinking Global Text Conditioning in Diffusion Transformers
2026
Cited
0
Resume
Academic Achievements
  • Publications:
  • - Turboedit: Instant text-based image editing. (ECCV 2024)
  • - Lazy diffusion transformer for interactive image editing. (ECCV 2024)
  • - Third time’s the charm? image and video editing with stylegan3. (AIM ECCVW 2022)
  • - Stylealign: Analysis and applications of aligned stylegan models. (ICLR 2022, Oral Presentation)
  • - Styleclip: Text-driven manipulation of stylegan imagery. (ICCV 2021, Oral Presentation)
  • - Stylespace analysis: Disentangled controls for stylegan image generation. (CVPR 2021, Oral Presentation)
  • - Fine-grained foreground retrieval via teacher-student learning. (WACV 2021)
Research Experience
  • Research scientist/engineer at Adobe Research in San Francisco, working with the FireFly team on Structure Reference.
Education
  • Bachelor's degree from Tongji University in 2016; PhD from Hebrew University of Jerusalem in 2022, supervised by Prof. Dani Lischinski and Eli Shechtman from Adobe Research.
Background
  • Main interests: generative modelings (GenAI), including diffusion model, GAN, and autoregressive model. Work on multi-modality generation/editing tasks, including image, video, and text.