Photo of Zhehao Li


Zhehao Li 李喆昊

         

I am a rising third-year CS Ph.D. student at Stanford University, advised by Prof. Doug James. I work at the intersection of simulation, graphics, and AI.

My current research focuses on physics-based simulation and learning. More broadly, I am excited about building intelligent machines at scale that can understand, reason about, and act in the physical world — machines with reasoning grounded in physical knowledge and a thinking process that is traceable and verifiable, and that can ultimately assist people in the real world.

I received both my B.E. in CS and my M.S. from the University of Science and Technology of China (USTC), where I was very fortunate to be advised by Prof. Ligang Liu in the Graphics & Geometric Computing Lab. I have also worked with Prof. Bo Ren at Nankai University and Prof. Tao Du at the Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University.


News


Selected Publications

BubbleGym
BubbleGym: A Practical Shape-to-Frequency Model for Acoustic Bubbles

Zhehao Li, Kui Wu, Wei Li, Doug L. James
SIGGRAPH Asia 2026 Conference Papers

TL;DR: Bubbles make most of the sound of water, and their pitch depends partly on shape; we predict it from shape ~1400× faster than the exact solver, within 1% error.

Webpage Paper Code Publisher
neural preconditioner
Learning Sparse Approximate Inverse Preconditioners for Conjugate Gradient Solvers on GPUs

Zherui Yang, Zhehao Li, Kangbo Lyu, Yixuan Li, Tao Du, Ligang Liu.
NeurIPS 2025

TL;DR: Solving large linear systems is the bottleneck in simulation and optimization; a graph neural network learns a GPU-friendly preconditioner that cuts solve time by 40–53%.

Webpage Paper Code
WindDancer
WindDancer: Understanding Acoustic Sensing under Ambient Airflow

Kuang Yuan, Dong Li, Hao Zhou, Zhehao Li, Lili Qiu, Swarun Kumar, Jie Xiong
Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), 2025

TL;DR: Moving air changes the speed of sound, quietly degrading microphone-based sensing such as motion tracking; we measure and explain the effect.

Publisher
DiffFR
DiffFR: Differentiable SPH-based Fluid-Rigid Coupling for Rigid Body Control

Zhehao Li, Qingyu Xu, Xiaohan Ye, Bo Ren, Ligang Liu
ACM Transactions on Graphics (Proceedings of SIGGRAPH Asia 2023)

TL;DR: A particle-based fluid-rigid coupling simulator you can get gradients from, optimizing a rigid body's motion in water up to 10× faster than gradient-free baselines.

Webpage Paper(15MB) Low-res Paper(2MB) Video Slides(59MB) Code Publisher
neural coarsen
Numerical Coarsening with Neural Shape Functions

Ning Ni, Qingyu Xu, Zhehao Li, Xiao-Ming Fu, Ligang Liu
Computer Graphics Forum, 2023

TL;DR: Simulate deformable objects on coarse meshes while keeping fine-scale behavior, by learning the coarse-to-fine mapping with geometric priors.

Webpage Paper

Service


Teaching