Xianghong Fang

Ph.D. Student in Statistical Sciences, University of Toronto

Hi! I am a third-year Ph.D. student in Statistics at the University of Toronto. I am very fortunate to be advised by Prof. Dehan Kong and Tim G. J. Rudner. I am also affiliated with the Vector Institute. Prior to my Ph.D. study, I received my Master of Philosophy in Artificial Intelligence from The Hong Kong University of Science and Technology, where I was very fortunate to be advised by Dit-Yan Yeung. I received my bachelor's degree from the University of Electronic Science and Technology of China.

Research Interests

My current research interests center on tokenization, generative modeling, multimodal systems, and agentic AI:

Vector Quantization for Autoregressive Modeling: Studying optimal vector quantization algorithms for visual autoregressive models and other modalities.
Reconstruction and Generation Gap: Understanding the gap between reconstruction and generation in autoregressive generative models, diffusion models, and flow matching models.
Unified Multimodal Generation: Developing efficient tokenizers for unified multimodal generative models, mainly under the autoregressive paradigm.
Agentic AI and Safety: Exploring agent safety, agent harnesses, and recursive self-optimization for AI agents, which will be a major focus of my upcoming research.

I am open to collaboration on related topics and happy to mentor self-motivated students who are excited about research.

News

[10/2026] Two papers have been updated on arXiv.
[09/2026] Two papers have been updated on arXiv.
[06/2026] One paper has been updated on arXiv.
[01/2026] One first-author paper was accepted to ICLR 2026.

Publications (* equal contribution, † Project head)

Latent Prior GAN decoder refinement with inherited generative structure

Latent Prior GANs: Efficient Decoder Refinement via Inherited Generative Structure

Wenjie Shu*, Xianghong Fang*†, Dehan Kong, Tim G. J. Rudner

Under review, 2026

paper code project page

bibtex

@article{shu2026latent,
  title   = {Latent Prior GANs: Efficient Decoder Refinement via Inherited Generative Structure},
  author  = {Shu, Wenjie and Fang, Xianghong and Kong, Dehan and Rudner, Tim G. J.},
  journal = {arXiv},
  year    = {2026}
}

FDDA generator and decoder alignment comparison and results

Efficient Fréchet Distributional Decoder Alignment for Latent Generative Models

Xianghong Fang*, Jiajun Zhu*, Esma Aimeur, Dehan Kong, Tim G. J. Rudner

Under review, 2026

paper code project page

bibtex

@article{fang2026efficient,
  title   = {Efficient Fréchet Distributional Decoder Alignment for Latent Generative Models},
  author  = {Fang, Xianghong and Zhu, Jiajun and Aimeur, Esma and Kong, Dehan and Rudner, Tim G. J.},
  journal = {arXiv},
  year    = {2026}
}

AudioGAR reconstruction and generation pipeline for latent audio models

AudioGAR: Bridging Reconstruction and Generation in Latent Audio Generative Models

Xianghong Fang*, Geeyang Tay*, Wentao Ma, Tim G. J. Rudner, Dehan Kong

Under review, 2026

paper code project page

bibtex

@article{fang2026audiogar,
  title   = {AudioGAR: Bridging Reconstruction and Generation in Latent Audio Generative Models},
  author  = {Fang, Xianghong and Tay, Geeyang and Ma, Wentao and Rudner, Tim G. J. and Kong, Dehan},
  journal = {Arxiv},
  year    = {2026}
}

A Latent Distribution Perspective on Evaluating and Improving Latent Generative Models

Xianghong Fang*, Wenjie Shu*, Tongda Xu, Wenlong Mou, Dehan Kong, Tim G. J. Rudner

Under review, 2026

paper code project page

bibtex

@article{fang2026latent,
  title   = {A Latent Distribution Perspective on Evaluating and Improving Latent Generative Models},
  author  = {Fang, Xianghong and Shu, Wenjie and Xu, Tongda and Mou, Wenlong and Kong, Dehan and Rudner, Tim G. J.},
  journal = {arXiv},
  year    = {2026}
}

A Unified Rate–Distortion Perspective on Vector, Product, and Scalar Quantization

Xianghong Fang, Wenlong Mou, Yuan Yuan, Dehan Kong, Tim G. J. Rudner

Under review, 2026

paper code project page

bibtex

@article{Fang2026ratedistortion,
  title   = {A Unified Rate--Distortion Perspective on Vector,
             Product, and Scalar Quantization},
  author  = {Fang, Xianghong and Mou, Wenlong and Yuan, Yuan and
             Kong, Dehan and Rudner, Tim G. J.},
  journal = {arXiv},
  year    = {2026}
}

VQ-Transplant: Efficient VQ-Module Integration for Pre-trained Visual Tokenizers

Xianghong Fang, Yuan Yuan, Dehan Kong, Tim G. J. Rudner

ICLR, 2026

paper code project page

bibtex

@inproceedings{Fang2026VQTransplant,
  title={V{Q}-Transplant: Efficient Plug-and-Play VQ-Module Integration for Pre-trained Visual Tokenizers},
  author={Xianghong Fang and Yuan Yuan and Dehan Kong and Tim G. J. Rudner},
  booktitle={ICLR},
  year={2026},
}

Distributional Matching for Vector Quantization: A Unified Theoretical and Empirical Framework

Xianghong Fang, Litao Guo, Hengchao Chen, Yuxuan Zhang, Xiaofan Xia, Dingjie Song, Yexin Liu, Hao Wang, Harry Yang, Qiang Sun, Yuan Yuan

Under review, 2026

paper code project page

bibtex

@article{fang2026distributional,
  title   = {Distributional Matching for Vector Quantization:
             A Unified Theoretical and Empirical Framework},
  author  = {Fang, Xianghong and Guo, Litao and Chen, Hengchao and
             Zhang, Yuxuan and Xia, Xiaofan and Song, Dingjie and
             Liu, Yexin and Wang, Hao and Yang, Harry and Sun, Qiang and
             Yuan, Yuan},
  journal = {arXiv},
  year    = {2026}
}

Rethinking the uniformity metric in self-supervised learning

Xianghong Fang, Jian Li, Qiang Sun, Benyou Wang

ICLR, 2024

paper code

bibtex

@inproceedings{Fang2024RethinkingTU,
  title={Rethinking the Uniformity Metric in Self-Supervised Learning},
  author={Xianghong Fang and Jian Li and Qiang Sun and Benyou Wang},
  booktitle={ICLR},
  year={2024}
}

Controlled text generation using dictionary prior in variational autoencoders

Xianghong Fang, Jian Li, Lifeng Shang, Xin Jiang, Qun Liu, Dit-Yan Yeung

ACL Findings, 2022

paper

bibtex

@inproceedings{Fang2022ControlledTG,
  title={Controlled Text Generation Using Dictionary Prior in Variational Autoencoders},
  author={Xianghong Fang and Jian Li and Lifeng Shang and Xin Jiang and Qun Liu and Dit-Yan Yeung},
  booktitle={ACL},
  year={2022},
}

Discrete auto-regressive variational attention models for text modeling

Xianghong Fang*, Haoli Bai*, Jian Li, Zenglin Xu, Michael Lyu, Irwin King

IJCNN, 2021

paper code

bibtex

@article{Fang2021DiscreteAV,
  title={Discrete Auto-regressive Variational Attention Models for Text Modeling},
  author={Xianghong Fang and Haoli Bai and Zenglin Xu and Michael R. Lyu and Irwin King},
  journal={IJCNN},
  year={2021},
}

DART: Domain-adversarial residual-transfer networks for unsupervised cross-domain image classification

Xianghong Fang, Haoli Bai, Ziyi Guo, Bin Shen, Steven Hoi, Zenglin Xu

Neural Networks, 2020

paper

bibtex

@article{Fang2020DARTDR,
  title={DART: Domain-Adversarial Residual-Transfer Networks for Unsupervised Cross-Domain Image Classification},
  author={Xianghong Fang and Haoli Bai and Ziyi Guo and Bin Shen and Steven C. H. Hoi and Zenglin Xu},
  journal={Neural networks},
  year={2020},
}

Academic Activities

Reviewer

Conference: NeurIPS 2025, NeurIPS 2026, ICLR 2026

Journal: TPAMI

Mentorship

Wenjie Shu: Undergraduate student at the University of Electronic Science and Technology of China.
Jiajun Zhu: Master's student at the University of Montreal.
Jiaqi Huang: Undergraduate student at the University of Toronto.