Xuhang Chen

Beachy Head Lighthouse, England

I am a third-year PhD student at the University of Cambridge and a Cambridge Trust Scholar, supported by the Cambridge Trust. I started my PhD in 2023.

Before joining Cambridge, I received my MSc in Applied Machine Learning from Imperial College London. I completed my undergraduate studies at Beihang University from 2018 to 2022, where I received dual degrees in Instrumentation Science and Applied Mathematics.

My current research interests focus on Self Evolving Models, Agentic Learning Systems, and Multimodal Reasoning. I also have relevant experience in world models and video generation.

I am always open to research collaborations, technical discussions, and coffee chats. Feel free to reach out if you are interested in related topics.

news

May 01, 2026	One paper is accepted at ICML 2026.

selected publications

ICML

CamGeo: Sparse Camera-Conditioned Image-to-Video Generation with 3D Geometry Priors

Xuanyi Liu, Deyi Ji, Liqun Liu, Lanyun Zhu, Xuhang Chen, Qianxiong Xu, Peng Shu, Huan Yu, Jie Jiang, Feng Gao, and Siwei Ma

In International Conference on Machine Learning, 2026

Abs HTML Code

Sparse camera-conditioned image-to-video generation presents a pivotal challenge: synthesizing geometrically consistent 3D motion from minimal pose cues. Existing methods, which largely rely on dense supervision or naive interpolation, suffer from severe pose drift and motion discontinuities due to the lack of robust 3D priors. In this paper, we introduce \textbfCamGeo, a novel framework that distills rich 3D geometric knowledge from a pre-trained video-to-3D model (VGGT) directly into the diffusion backbone. To achieve this without incurring inference latency, we propose a training-only distillation strategy. Specifically, CamGeo incorporates: (1) keyframe trajectory distillation that enforces cycle-consistency with sparse input poses, (2) cross-frame consistency distillation with both camera trajectory and depth constraints to generate consistent structure across unsupervised frames, and (3) a three-stage coarse-to-fine curriculum learning, progressively scales geometric complexity, from global structure coherence to fine-grained refinement, achieving stable optimization. Extensive experiments demonstrate that CamGeo achieves consistent improvements under various sparsity ratios.