Zhenwei (Zavier) Wang
Research Interests:
- World Model
- 3D Reconstruction
- Video Generation
- 3D Asset Generation
About Me
I am a researcher at Tencent Hunyuan, working on the HY World project. I received my Ph.D. degree from the Department of Computer Science, City University of Hong Kong (CityU), under the supervision of Prof. Rynson W.H. Lau and Prof. Gerhard Hancke. I received my B.Eng. degree from Xiamen University.
During my Ph.D. studies, I was fortunate to work with Tengfei Wang and Ziwei Liu from Shanghai AI Lab and MMLab@NTU. I was also lucky to closely collaborate with Nanxuan Zhao from Adobe Research.
I'm interested in generative AI, including world models, video generation, 3D reconstruction, and 3D asset generation.
News
- [Aug. 2025] Three papers accepted to SIGGRAPH Asia 2025. See you in Hong Kong!
- [Aug. 2025] HunyuanWorld 1.0 is released. Check it out!
- [Feb. 2025] One paper accepted to CVPR 2025.
- [Feb. 2025] One paper accepted to ICLR 2025. See you in Singapore!
- [Apr. 2024] One paper accepted to SIGGRAPH 2024. See you in Denver!
- [Dec. 2023] One paper accepted to AAAI 2024.
- [Apr. 2023] One paper accepted to SIGGRAPH 2023 (journal track).
Experience
-
Researcher, Tencent HY World08.2025 - Present
-
Research Intern, Tencent HY World04.2025 - 07.2025
-
Visiting Scholar, Harvard University, VCG Group10.2024 - 02.2025
-
Research Intern, Shanghai AI Laboratory, 3D AIGC Team09.2023 - 10.2024
Open-Source Projects
|
Technical Report
|
HY-World 2.0 🔥Tencent HY World, April 2026 Next-gen multi-modal world model that reconstructs, generates, and simulates 3D worlds. |
|
Technical Report
|
HY-World 1.5Tencent HY World, December 2025 The first open-source world model with real-time latency, long-term memory, and interactive control. |
|
Technical Report
|
HY-World 1.0Tencent HY World, July 2025 Generate immersive, editable 3D scenes from text or a single image. |
Selected Publications
(*equal contribution, ^intern, †corresponding author)
|
Technical Report
|
HY-World 2.0: A Multi-Modal World Model for Reconstructing, Generating, and Simulating 3D WorldsTencent HY World, April 2026 Next-gen multi-modal world model that reconstructs, generates, and simulates 3D worlds. |
|
Preprint
|
WorldPlay: Towards Long-Term Geometric Consistency for Real-Time Interactive World Modelingpreprint, December 2025 Real-time interactive world model with long-term geometric memory. |
|
Preprint
|
WorldMirror: Universal 3D World Reconstruction with Any-Prior Promptingpreprint, October 2025 Universal feed-forward 3D reconstruction — any input, any output. |
|
Preprint
|
MoCA: Mixture-of-Components Attention for Scalable Compositional 3D Generationpreprint, December 2025 Compositional 3D scene and object generation via sparse mixture-of-components attention. |
|
SIGGRAPH Asia 2025
|
StyleSculptor: Zero-Shot Style-Controllable 3D Asset Generation with Texture-Geometry Dual GuidanceProc. ACM SIGGRAPH Asia, December 2025 Training-free 3D asset generation with joint geometry and texture style guidance. |
|
SIGGRAPH Asia 2025
|
Voyager: Long-Range and World-Consistent Video Diffusion for Explorable 3D Scene GenerationACM Trans. on Graphics (Proc. ACM SIGGRAPH 2025, Journal), December 2025 Long-range 3D world exploration powered by RGB-D video diffusion. |
|
SIGGRAPH Asia 2025
|
Shape-for-Motion: Precise and Consistent Video Editing with 3D ProxyProc. ACM SIGGRAPH Asia, December 2025 Precise video object manipulation via 3D proxies and diffusion rendering. |
|
CVPR 2025
|
MAGE: Single Image to Material-Aware 3D via the Multi-View G-Buffer Estimation ModelCVPR, June 2025 Single-image, material-aware 3D reconstruction via G-buffer estimation. |
|
ICLR 2025
|
Phidias: A Generative Model for Creating 3D Content from Text, Image, and 3D Conditions with Reference-Augmented DiffusionICLR, April 2025 Retrieval-augmented 3D diffusion — generate from text, image, or existing 3D assets. |
|
SIGGRAPH 2024
|
ThemeStation: Generating Theme-Aware 3D Assets from Few ExemplarsProc. ACM SIGGRAPH, August 2024 Generate theme-consistent 3D asset galleries from just a few exemplars. |
|
AAAI 2024
|
Recasting Regional Lighting for Shadow RemovalAAAI, February 2024 Shadow removal that restores degraded textures conditioned on recovered illumination. |
|
SIGGRAPH 2023
|
Language-based Photo Color Adjustment for Graphic DesignsACM Trans. on Graphics (Proc. ACM SIGGRAPH 2023, Journal), August 2023 Language-driven photo recoloring for graphic designs. |
Experience and Collaborations
Services
- Conference/Journal Reviewer: SIGGRAPH, 3DV, CVPR, ECCV, ICLR, ICML, TOG, TVCG, IJCV
- Teaching Assistant: Virtual Reality Technologies and Applications (CS4188/CS5188)
Selected Awards
- Outstanding Academic Performance Award, City University of Hong Kong, 2023, 2024
- Outstanding Graduate, Xiamen University, 2020
- Outstanding Student, Xiamen University, 2019, 2020
- Huang Xilie Scholarship, Xiamen University (Top 2%), 2019, 2020
- First-Class Scholarship (Top 3%), 2018
