Xinhao Ji | 姬歆淏

I am an undergraduate majoring in Applied Physics at the School of EECS of Peking University, expected to graduate in June 2027. I am advised by Prof. Dingwen Zhang at Northwestern Polytechnical University (NWPU), and Prof. Zhihang Zhong at Shanghai Jiao Tong University (SJTU). From March to June 2026, I conduct a research internship at PsiBot, advised by Prof. Yaodong Yang. Starting in June 2026, I will join as an visiting student advised by Prof. Ziwei Liu at MMLab@NTU. My research interests lie in 3D vision & robotics and world models.

Email (PKU) / Email (Gmail) / Github

Research

* indicates equal contribution † indicates corresponding author

	Holi-Spatial: Evolving Video Streams into Holistic 3D Spatial Intelligence Yuanyuan Gao, Hao Li, Yifei Liu, Xinhao Ji*, Yuning Gong, Yuanjun Liao, Fangfu Liu, Manyuan Zhang, Yuchen Yang, Dan Xu, Xue Yang, Huaxi Huang, Hongjie Zhang, Ziwei Liu, Xiao Sun, Dingwen Zhang†, Zhihang Zhong† ICML 2026, Oral Paper / Project Page / Code / Dataset We evolve continuous egocentric video streams into holistic 3D spatial intelligence, with large-scale spatial QA benchmarks for unified scene understanding.
	LangSurf: Language-Embedded Surface Gaussians for 3D Scene Understanding Hao Li, Minghan Qin, Zhengyu Zou, Diqi He, Xinhao Ji, Bohan Li, Bingquan Dai, Dingwen Zhang†, Junwei Han IEEE TPAMI, Accepted* Paper / Project Page We propose LangSurf, a language-embedded surface field that aligns 3D language Gaussians with object surfaces for open-vocabulary 2D and 3D scene understanding.
	Zip-GS: Extreme Compression of 3D Gaussians with Diffusion Priors Jiaqi Chen, Xinhao Ji*, Yuanyuan Gao, Hao Li, Yuning Gong, Yifei Liu, Dingwen Zhang†, Dan Xu, Xiao Sun, Zhihang Zhong† IEEE TMM, Under Review Paper We present ZipGS, a feed-forward framework for extreme 3D Gaussian Splatting compression and restoration using diffusion priors, achieving up to 300× compression with high-fidelity reconstruction.
	EgoSteer: A Full-Stack System Towards Steerable Dexterous Manipulation from Egocentric Videos Yifan Zhong, Zhang Chen, Tianrui Guan, Fanlian Zeng, Yuyao Ye, Tianjia He, Ka Nam Lui, Jiayi Li, Tingrui Zhang, Ruilin Yan, Xinhao Ji, Guangyu Zhao, Wenjie Lou, Jiayuan Zhang, Yuanpei Chen†, Yaodong Yang† CoRL 2026, Under Review Project Page / Code We present EgoSteer, a full-stack system for steerable dexterous manipulation from egocentric videos, combining EgoSmith for scalable data curation with a world-model-enhanced VLA.
	Uni4DMo: Learning Universal 4D Probabilistic Spatial Prior from Heterogeneous Motion Data Diqi He, Xinhao Ji, Yan Li, Xuehao Gao, Junwei Han, Dingwen Zhang† IEEE TPAMI, Under Review** Paper (coming soon) We introduce PMF, a universal 4D Gaussian motion representation, and Uni4DMo, a diffusion model that learns a unified 4D spatial prior from heterogeneous motion data.

Experience

	Peking University (PKU) 2022.09 - 2027.06 Undergraduate Student School of EECS, Applied Physics
	Shanghai Artificial Intelligence Laboratory 2025.06 - 2026.02 Research Intern
	PsiBot 2026.03 - 2026.05 Research Intern
	MMLab@NTU, Ropedia 2026.06 - Present Visiting Student Research Advisor: Prof. Ziwei Liu

Selected Awards

2025: Second Prize, Spatial Intelligence Visual Perception Challenge, PRCV 2025

This homepage is designed based on Jon Barron's website.