I’m Siyuan Cen (岑思远), a first year master at UMass Amherst, advised by Prof. Chuang Gan. I’m currently a research intern at CMU, co-advised by Prof. Deva Ramanan and Prof. Yilun Du. Previously, I earned my Bachelor’s degree from Nanjing University, where I had the privilege of working with Prof. Ying Tai.

My current focus is on video understanding, video editing and robotics.

📝 Publications

arXiv

Towards Understanding Camera Motions in Any Video

Zhiqiu Lin*, Siyuan Cen*, Daniel Jiang, Jay Karhade, Hewei Wang, Chancharik Mitra, Yu Tong Tiffany Ling, Yuhan Huang, Sifan Liu, Mingyu Chen, Rushikesh Zawar, Xue Bai, Yilun Du, Chuang Gan, Deva Ramanan

Website | arXiv |

We present CameraBench, a large-scale dataset and benchmark for evaluating and improving camera motion understanding in both SfM/SLAM systems and VLMs.

arXiv

RapVerse: Coherent Vocals and Whole-Body Motions Generations from Text

Jiaben Chen, Xin Yan, Yihang Chen, Siyuan Cen, Qinwei Ma, Haoyu Zhen, Kaizhi Qian, Lie Lu, and Chuang Gan

Website | arXiv

In this paper, we introduce a challenging task for simultaneously generating 3D holistic body motions and singing vocals directly from textual lyrics inputs. To facilitate this, we first collect the RapVerse dataset, a large dataset containing synchronous rapping vocals, lyrics, and high-quality 3D holistic body meshes.

Security and Communication Networks CameraBench teaser

Adversarial Attacks on Large Language Model-Based System and Mitigating Strategies: A Case Study on ChatGPT

Bowen Liu, Boao Xiao, Xutong Jiang, Siyuan Cen, Xin He, Wanchun Dou

Website | arXiv

The paper exposes how adversarial prompts can bypass ChatGPT's safeguards and proposes two defenses—a prefix prompt and a RoBERTa-based detector—that effectively block such attacks, with the latter achieving perfect accuracy.