Jiacheng Dong

Undergraduate, ZJU-UIUC Institute · Electrical Engineering

3D/4D scene perception & long-context world models

Hello! I am an undergraduate in Electrical Engineering at the ZJU-UIUC Joint Institute of Zhejiang University (B.S. 2024–2028). My research focuses on understanding and modeling the dynamic 3D world, with current interests in streaming 4D perception, long-context vision models, and 3D representation learning. Recently, I have become particularly interested in test-time training (TTT) as a form of memory. Since July 2025, I have been working with Prof. Gaoang Wang in the CVNext Lab.

Professional GPA: 4.0/4.0 · Cumulative GPA: 3.88/4.0

Portrait of Jiacheng Dong

News

2025.10
•Awarded the National Scholarship for academic excellence.
2025.07
•Joined CVNext Lab (RA), ZJU-UIUC Institute.
2025.07
•Personal academic homepage launched with blog.
2025.02
•Won Outstanding Award & SIAM Award in MCM 2025 (tourism sustainability modeling).
2024.11
•Started research at Prof. Jixing Cao's Lab (CV & point clouds).
2024.09
•Enrolled at ZJU-UIUC Institute, Zhejiang University.

Research

My research focuses on understanding and modeling the dynamic 3D world. Current interests include:

  • Streaming 4D perception and reconstruction of dynamic scenes.
  • Long-context vision models for learning from continuous visual experience.
  • 3D representation learning for geometry, motion, and generation.
Seeking Remote Intern/Assistant (Fall 2026): I am seeking remote research internship / assistant positions in 3D/4D scene perception and long-context vision modeling (remote during the semester; open to on-site visits in US labs during summer). If interested in collaboration, please contact me via Email.

Publications

RecurGS: Interactive Scene Modeling via Discrete-State Recurrent Gaussian Fusion
Conference Submission
2025 · under double-blind review

Authors: Wenhao Hu, Haonan Zhou, Zesheng Li, Liu Liu, Jiacheng Dong, Zhizhong Su, Gaoang Wang

Interactive scene modeling with discrete-state recurrent Gaussian fusion for dynamic 3D understanding.

TBR: A Training-Free Framework for Rapid Background Removal in Large-Scale Point Clouds
Pattern Recognition (under review)
2025

Authors: Jiacheng Dong, Yuxiao Zheng, Maolin Tian, Jun Liang, Jixing Cao

Led the complete rewrite of the Python implementation achieving 20× performance improvement. Designed multiple real-world 3D point cloud scenarios and led manuscript writing covering methodology, experimental results, and analysis.

Awards & Honors

National Scholarship
Ministry of Education, China · Oct 2025
Recognized for outstanding academic excellence.
Outstanding Award & SIAM Award
MCM 2025 - Mathematical Contest in Modeling
Developed a system dynamics model for tourism sustainability in Juneau, Alaska
Award rate: 0.11%

Education

B.S. in Electrical Engineering
ZJU-UIUC Institute, Zhejiang University
2024 - 2028 (Expected)

Professional GPA: 4.0/4.0 | Cumulative GPA: 3.88/4.0

Successfully transferred majors during my first to second year to focus on Electrical Engineering. Also planning to complete a Minor in Computer Science at University of Illinois Urbana-Champaign.

Projects

Learning-based 3D Vision cover
Learning-based 3D Vision
GitHub Repository

A list of works in Learning-based 3D Vision, aiming to provide a resource for scholars, practitioners, and enthusiasts interested in 3D vision and its potentially exciting roles in future embodied intelligence / world perception and other areas.

Visitors

Visitor locations (MapMyVisitors, approximate).