Wenda Chu 储闻达

Wenda Chu

储闻达

PhD Student

Caltech

     

Biography

I am currently a third year PhD student at CMS department, Caltech, where I am very fortunate to be advised by Yisong Yue and Yang Song. My research focuses on generative modeling, in particular inference-time posterior sampling algorithms with pretrained diffusion priors, with application to reward guided generation and solving scientific inverse problems. More recently, I have been working on developing more generation-friendly visual tokenizers for latent generative models.

I like playing piano and I especially enjoy playing the songs from animes. I am also a fan of chess and photography.

Download my curriculum vitae.

Interests
  • Generative Models
  • Inverse Problems
Education

  • California Institute of Technology, 2023-Present

    Computing + Mathematical Science

  • Tsinghua University, 2019-2023

    Yao Class, Institute for Interdisciplinary Information Sciences

  • Shenzhen Middle School, 2013-2019

News
Sep, 2025 Two papers accepted to NeurIPS 2025! Meet me at San Diego :)
April, 2025 Our paper DAPS received an oral presentation at CVPR 2025!
Jan, 2025 Our paper InverseBench received a spotlight presentation at ICLR 2025!
Sep, 2023 I joined Caltech CMS as a PhD student!
June, 2023 I graduated from Tsinghua University!

Experience

 
 
 
 
 
Secure Learning Lab, University of Illinois at Urbana-Champaign
Research Intern
ByteDance Seed
June 2025 - Sep 2025 San Jose
 
 
 
 
 
Secure Learning Lab, University of Illinois at Urbana-Champaign
Research Intern
Secure Learning Lab, University of Illinois at Urbana-Champaign
Nov 2021 - June 2023 Illinois

I was fortunate to be advised by Prof. Bo Li. Projects include:

  • Certified robustness for point cloud models
  • Certified robustness for multi-sensor fusion systems
  • Fair federated learning by EM clustering
 
 
 
 
 
Tsinghua University
Research Intern
Tsinghua University
Sep 2021 - Feb 2023 Beijing China
  • I was fortunate to be advised by Prof. Xiaolin Hu and researched on launching physically realizable adversarial attacks on object detection models.

Selected Projects

Distributed Robust Principal Component Analysis
The first distributed robust principal component analysis algorithm with some convergence guarantees.
PDF Code arxiv
Distributed Robust Principal Component Analysis
Traffic at Peak Hours - A Game Theory View
Analyze an illustrative traffic game using game theory and queuing theory. We show how excessive competition to limited resources could lead to a dramatic decrease of social welfare.
PDF Code
Traffic at Peak Hours - A Game Theory View
Ray Tracing Renderer
A simple C++ implementation of ray tracing rendering, based on Monte Carlo path tracing.
Ray Tracing Renderer

Contact