Zilu Li

Hi! I'm a second-year CS PhD student at UC San Diego, co-advised by Prof. Ravi Ramamoorthi and Prof. Tzu-Mao Li. My research interest lies in computer graphics and computer vision, such as physics simulation, geometry processing, etc. Previously, I was an undergrad at Cornell where I was fortunate to work with Prof. Bharath Hariharan, Prof. Steve Marschner, and Prof. Gordon Wetzstein.

Email: zil124 [at] ucsd [dot] edu

Email / Google Scholar / Twitter / Github

Research

: * equal contribution, Highlighted Papers

	Differentiable Light Transport with Gaussian Surfels via Adapted Radiosity for Efficient Relighting and Geometry Reconstruction Kaiwen Jiang, Jia-Mu Sun, Zilu Li, Dan Wang, Tzu-Mao Li, Ravi Ramamoorthi ACM Transactions on Graphics (SIGGRAPH Asia), 2025 [Paper] / [Page] We adopt Gaussian surfels as the primitives and build an efficient framework for differentiable light transport, inspired by classic radiosity theory, enabling efficient relighting and geometry reconstruction.
	Neural Control Variates with Automatic Integration Zilu Li, Guandao Yang, Qingqing Zhao, Xi Deng, Leonidas Guibas, Bharath Hariharan, Gordon Wetzstein SIGGRAPH, 2024 We present a method that uses arbitrary neural network architectures as control variates with automatic differentiation to create unbiased, low-variance, and numerically stable Monte Carlo estimators for various problem setups.
	Neural Cache for Monte Carlo Partial Differential Equation Solver Zilu Li, Guandao Yang, Xi Deng, Christopher De Sa, Bharath Hariharan, Steve Marschner SIGGRAPH Asia, 2023 [Page] Using neural field caches to reduce variance in Monte Carlo PDE solvers. Our method excels when computing is limited and enables effective bias and variance trade-off.
	BT^2: Backward-compatible Training with Basis Transformation Yifei Zhou, Zilu Li, Abhinav Shrivastava, Hengshuang Zhao, Antonio Torralba, Taipeng Tian, Ser-Nam Lim ICCV, 2023 [Paper] Introducing a Basis Transformation Method to avoid backfilling in learning new representations for modern retrieval systems.

Industry Experience

Research Scientist Intern, High-Fidelity Physics Research Team
Summer 2025 - Present
Mentors: Rohan Sawhney, David I. W. Levin, Ken Museth

Miscellaneous

Outside of research, I like rock climbing, making ceramics (both hand-building and wheel throwing), and playing guitar. I used to sing in a band called Cornell Interlude.

Teaching

I served as a teaching assistant for the following courses.

Spring 2024: CS 4670 Computer Vision

Spring 2023, Fall 2023: CS 4780 Machine Learning for Intelligent Systems

Website Template: Jon Barron's website.