Yuanyuan Gao | 高源远

I am a Ph.D. student in the Department of Computer Science and Engineering at The Hong Kong University of Science and Technology (HKUST), advised by Prof. Dan Xu. I received my bachelor's and master's degrees from the School of Automation at Northwestern Polytechnical University (NWPU), where I was a member of BRAIN Lab, advised by Prof. Dingwen Zhang and Prof. Junwei Han. I have also had the pleasure of working with Zhihang Zhong and Xiao Sun.

My research aims to build scalable, cost-effective 3D/4D data pipelines for creating large-scale, realistic simulation environments. By bridging the real-to-sim gap, I seek to advance embodied intelligence and spatial understanding in vision-language models (VLMs).

Email / Google Scholar / GitHub

News

2026-05: ICML 2026 — Holi-Spatial accepted as Oral.
2026-02: CVPR 2026 — Proxy-GS accepted as Oral (full scores 6/6).
2025-09: ICCV 2025 — CityGS-X.
2024-07: ECCV 2024 — GGRT (equal contribution).

Selected Publications

*Equal contribution

Holi-Spatial: Evolving Video Streams into Holistic 3D Spatial Intelligence.
Yuanyuan Gao, Hao Li, Yifei Liu, Xinhao Ji, Yuning Gong, Dingwen Zhang, et al., Zhihang Zhong.
International Conference on Machine Learning (ICML), 2026 Oral(168/23918)
[Project Page] [Code]

Holistic 3D Spatial Intelligence from evolving video streams; benchmark and method for spatial reasoning.

Proxy-GS: Efficient 3D Gaussian Splatting via Proxy Mesh.
Yuanyuan Gao, Yuning Gong, Yifei Liu, Li Jingfeng, Zhihang Zhong, Dingwen Zhang, Yanci Zhang, Dan Xu, Xiao Sun.
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026 Oral (Award Candidate)
[Project Page] [Code]

Efficient 3DGS using a proxy mesh for faster rendering and training while keeping geometric quality.

CityGS-X: A Scalable Architecture for Efficient and Geometrically Accurate Large-Scale Scene Reconstruction.
Yuanyuan Gao, Hao Li, Jiaqi Chen, Zhengyu Zou, Zhihang Zhong, Dingwen Zhang, Xiao Sun, Junwei Han.
IEEE International Conference on Computer Vision (ICCV), 2025
[Project Page] [Code]

Scalable parallel architecture for city-scale 3D Gaussian reconstruction with strong geometry.

GGRT: Towards generalizable 3D gaussians without pose priors in real-time.
Hao Li*, Yuanyuan Gao*, Chenming Wu, Dingwen Zhang, Yalun Dai, Chen Zhao, Haocheng Feng, Errui Ding, Jingdong Wang, Junwei Han.
European Conference on Computer Vision (ECCV), 2024
[Project Page] [Code]

Generalizable 3D Gaussians in real time without pose priors.

Preprints & Reports

Visionary: The World Model Carrier Built on WebGPU-Powered Gaussian Splatting Platform.
Yuning Gong, Yifei Liu, Yifan Zhan, Muyao Niu, Xueying Li, Yuanjun Liao, Jiaming Chen, Yuanyuan Gao, Jiaqi Chen, Minming Chen, et al.
Technical Report, 2025
[Project Page] [Code]

An open, web-native WebGPU + ONNX Runtime platform for real-time 3DGS / 4DGS / Avatar rendering — a unified world-model carrier in the browser.

Education

Ph.D. student in Computer Science and Engineering, HKUST, Sep 2026 – Present. Advisor: Prof. Dan Xu.
M.Eng. in Control Science and Engineering, NWPU, Aug 2023 – Mar 2026. BRAIN Lab; advisors: Prof. Dingwen Zhang, Prof. Junwei Han.
B.Eng. in Automation, NWPU, Aug 2019 – Jun 2023.

Research Internships

Tencent · Qingyun Talent Program (青云人才计划)May 2026 – Aug 2026

Hand Motion Reconstruction & Data Pipelines
- Worked on hand motion reconstruction for Video2Motion, focusing on high-fidelity motion representations, temporal consistency, and data usability.
- Built a hand motion data pipeline for automatic mining, motion filtering, temporal quality checks, and standardized curation of large-scale human motion data. This work laid the foundation for a scalable 4D motion data pipeline, supporting Video2Motion reconstruction and providing training data for embodied AI.
Shanghai AI LaboratoryMar 2025 – Apr 2026

3D Spatial Intelligence & 3D Multimodal Large Language Models
- Led the development of CityGS-X's parallel architecture for efficient, scalable, and geometrically accurate city-scale reconstruction.
- Proposed Proxy-GS, introducing occlusion priors to accelerate structured 3D Gaussian splatting rendering; accepted to CVPR 2026 as an Oral and Award Candidate.
- Led Holi-Spatial's 4D data and benchmark construction, integrating automated 3D multimodal annotation, quality control, and evaluation to support spatial understanding in VLMs and VLAs; accepted to ICML 2026 as an Oral.
Xiaomi RoboticsDec 2024 – Feb 2025

Indoor 3D Gaussian Reconstruction
- Integrated StableNormal geometric priors into 3D Gaussian reconstruction to improve multiview consistency and mesh quality under challenging lighting and material variations.
- Developed robust indoor geometry and surface reconstruction to support spatial modeling and semantic alignment for robot navigation and manipulation.

Honors and Awards

National Scholarship (2025)
NWPU Outstanding Graduate Student (2025)
NPU College of Automation Sport Star (2023)
NPU College of Automation Academic Star — Top 20% (2022)