Weijie Wang   |   王伟杰

I am a first-year Ph.D. student in Computer Science at College of Computer Science and Technology, Zhejiang University, advised by Prof. Bohan Zhuang. My research interests lie in the fields of efficient large 3D model/world model. I am currently a research intern at ByteDance Seed.

Previously, I was a research intern at Microsoft Research Asia. I obtained my B.E. degree from College of Computer Science and Technology and Chu Kochen Honors College (ACEE) at Zhejiang University in 2025. I was fortunate to work closely with Prof. Chunhua Shen, Dr. Donny Y. Chen, Dr. Jiawang Bian and Dr. Zheng Zhu.

If you are interested in working with us as an intern at Zhejiang University, please feel free to email me and CC Prof. Bohan Zhuang.

Email  /  Google Scholar  /  Github /  X /  LinkedIn /  CV

• [2026/05] 🎉 Two papers (World-R1, Flash-GRPO) have been accepted to ICML 2026!
• [2026/04] 🔥 World-R1 was selected as Hugging Face #1 Paper of the Day!
• [2026/04] 🎉 Our paper CoV has been accepted to ACL 2026 Findings!
• [2026/03] 🎉 Our paper DriveGen3D has been accepted to ICME 2026!
• [2026/01] 🎉 Two papers (MV-RoboBench, ReVisual-R1) have been accepted to ICLR 2026!
• [2025/11] 🎉 Our paper PM-Loss has been accepted to 3DV 2026!
• [2025/10] 🏅 I received the NeurIPS Scholar Award.
• [2025/09] 🎉 Our paper ZPressor has been accepted to NeurIPS 2025!
• [2025/06] 🎉 Our paper WonderTurbo has been accepted to ICCV 2025!
• [2025/06] 🏅 I received the Outstanding Graduates of Zhejiang University and Outstanding Thesis of Zhejiang University awards.
• [2025/01] 🎉 Our paper TransDiff has been accepted to ICRA 2025!
• [2024/07] 🎉 Our paper Street Gaussians has been accepted to ECCV 2024!
• [2023/11] 🏅 I received the Zhejiang Government Scholarship.
• [2023/01] 🏅 I was named a Meritorious Winner of The Interdisciplinary Contest in Modeling.

We present a problem-driven survey of feed-forward 3D scene modeling, covering representative methods, datasets, and downstream applications.

VolSplat improves multi-view consistency and geometric accuracy for feed-forward 3DGS with voxel-aligned prediction.

World-R1 aligns text-to-video generation with 3D constraints through reinforcement learning, improving geometric consistency while preserving visual quality and motion diversity.

We propose Chain-of-View (CoV) prompting, a training-free test-time reasoning framework that transforms VLMs into active viewpoint reasoners for spatial reasoning in 3D environments.

We present DriveGen3D, a novel framework for generating high-quality and highly controllable dynamic 3D driving scenes that addresses critical limitations in existing methodologies.

We introduce PM-Loss, a novel regularization loss based on a learned point map for feed-forward 3DGS, leading to smoother 3D geometry and better rendering.

ZPressor is an architecture-agnostic module that compresses multi-view inputs for scalable feed-forward 3DGS.

We introduce WonderTurbo, the first real-time interactive 3D scene generation framework capable of generating novel perspectives of 3D scenes within 0.72 seconds.

We introduce MV-RoboBench, a benchmark for evaluating multi-view spatial reasoning in robotic manipulation, revealing large gaps between state-of-the-art VLMs and human performance.

We introduce ReVisual-R1, achieving a new state-of-the-art among open-source 7B MLLMs on challenging benchmarks.

We propose a single-view RGB-D-based depth completion framework, TransDiff, that leverages the Denoising Diffusion Probabilistic Models(DDPM) to achieve material-agnostic object grasping in desktop.

This paper aims to tackle the problem of modeling dynamic urban street scenes from monocular videos. We introduce Street Gaussians, a new explicit scene representation that tackles some major limitations.

This template is a modification to Jon Barron's website.