Yirong Shen 沈毅荣

Welcome to my website :)

IMG_6657.jpg

Department of Electrical and Electronic Engineering

Imperial College London

London, United Kingdom

I am currently a PhD student at Imperial College London, where I am fortunate to be supervised by Prof. Cong Ling. I study information theory and learning theory, focusing on Shannon-theoretic foundations of machine learning and privacy. My research explores how information-theoretic principles can be used to better understand and improve large language models and modern generative models.

I received my Master’s degree from the University of Sydney, where I was fortunate to be supervised by Prof. Yonghui Li.

Before my PhD, I largely squandered my odyssey years on CS:GO — perhaps not the most wisely spent years. ;( Check out my FACEIT profile.

news

Apr 28, 2026 Our paper “Rényi diffusion models” has been accepted by ICML 2026!
Mar 28, 2026 Our paper “Smoothing Linear Codes by Rényi Divergence and Applications to Security Reduction” has been accepted by IEEE ISIT 2026!
Sep 28, 2025 Finished a rewarding two-month visiting stay at Shanghai Jiao Tong University, hosted by Prof. Jia Wang. I learned a great deal and enjoyed many stimulating discussions on generative modeling and information theory.
Sep 24, 2025 Our paper “Information Theoretic Learning for Diffusion Models with Warm Start” has been accepted by NeurIPS 2025!
Apr 10, 2025 Our paper “Generalized Score Matching: Bridging $f$-Divergence and Statistical Estimation Under Correlated Noise” has been accepted by IEEE ISIT 2025!

selected publications

  1. brownian-motion.gif
    Information Theoretic Learning for Diffusion Models with Warm Start
    Yirong Shen, Lu Gan, and Cong Ling
    In The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS), 2025
  2. triangle.png
    Generalized Score Matching: Bridging f-Divergence and Statistical Estimation Under Correlated Noise
    Yirong Shen, Lu Gan, and Cong Ling
    In 2025 IEEE International Symposium on Information Theory (ISIT), 2025