Hi, I’m Hengyuan!

Hengyuan Xu Hi! I’m Hengyuan Xu, and this site is a documentation of myself. Contact me by email: doby-xu@outlook.com

⚠️
Notification: This site is currently under permanent construction, just as the person behind it.

ℹ️

News: WithAnyone is now released! WithAnyone is capable of generating high-quality, controllable, and ID consistent images. Demo is out, try it now!

Try WithAnyone!Try WithAnyone!
Leave a star if you like it!

About Me

I am a Master’s student in FVL at the School of Computer Science, Fudan University, advised by Xingjun Ma. I received my bachelor degree from SEIEE in Shanghai Jiao Tong University and I was a research intern at John Hopcroft Center for Computer Science, advised by Liyao Xiang. Here is my sincere gratitude to my mentors and collaborators. Currently I am interning at StepFun, exploring possibilities in the field of ID-consistency generation.

I am still exploring potential research directions I am interested in. My previous research explores the mathematical properties of Transformers and their potential applications in privacy-preserving deep learning and model intellectual property protections. Recently, I have been focusing on generative AI and its applications.

Publications

  • Hengyuan Xu, Wei Cheng, Peng Xing, Yixiao Fang, Shuhan Wu, Rui Wang, Xianfang Zeng, Daxin Jiang, Gang Yu, Xingjun Ma, Yu-Gang Jiang. “WithAnyone: Towards Controllable and ID Consistent Image Generation” arXiv preprint arXiv:2510.14975
Try WithAnyone!Try WithAnyone!
WithAnyone: Towards Controllable and ID Consistent Image Generation
  • Hengyuan Xu, Liyao Xiang, Xingjun Ma, Borui Yang, Baochun Li. “Hufu: A Modality-Agnostic Watermarking System for Pre-Trained Transformers via Permutation Equivariance” arXiv preprint arXiv:2403.05842
  • Xingjun Ma, et al. “Safety at scale: A comprehensive survey of large model safety” arXiv preprint arXiv:2502.05206

  • Cailin Zhuang, et al. “Vistorybench: Comprehensive benchmark suite for story visualization” arXiv preprint arXiv:2505.24862

  • Hengyuan Xu, Liyao Xiang, Hangyu Ye, Dixi Yao, Pengzhi Chu, Baochun Li. “Permutation Equivariance of Transformers and Its Applications” in Proceedings of The IEEE/CVF Conference on Computer Vision and Pattern Recognition 2024 (CVPR)

    My Work Accepted by CVPR'24My Work Accepted by CVPR'24
    Permutation Equivariance of Transformer

  • Dixi Yao, Liyao Xiang, Hengyuan Xu, Hangyu Ye, Yingqi Chen. “Privacy-Preserving Split Learning via Patch Shuffling over Transformers” in Proceedings of the 22nd IEEE International Conference on Data Mining (ICDM)

Intern Experience

  • Research Intern, StepFun, Shanghai, China (2024.11 - Present)

    • Supported several algorithm development for “Lipu” APP.
    • Developed and open-sourced WithAnyone.
      • curated million-level paired ID-centric datasets.
      • observed and formulated copy-paste patterns in current ID-consistency generation methods with newly developed benchmark
      • trained a 6B IPA model on FLUX from scratch and jumped out of trade-off between similarity and diversity.
    • Exploring MLLM reflection for TI2I models like editing and cref models.

  • Research Intern, John Hopcroft Center for Computer Science, Shanghai Jiao Tong University, Shanghai, China (2022.11 - 2024.06)

    • Worked on privacy-preserving deep learning and model intellectual property protections.
    • Explored the mathematical properties of Transformers.