Undergraduate Student
Xingjian College
Tsinghua University
My research interests center on ML for inverse problem and sparse reconstruction, ML-driven methods for numerical computation and physical field simulation, and multiphysics. I am also interested in embodied intelligence and robotics.
I continue to collaborate with the CAB Lab at Princeton University following my exchange there in spring 2026.
I welcome research opportunities for summer 2027 and plan to apply to PhD programs for Fall 2028. Let's connect.
Preprint Submitted to ICLR 2027
NeTMY combines neural fields with differentiable physics for sparse reconstruction, revealing how forward-model fidelity and neural parameterization shape inverse optimization.
Preprint Submitted to ICLR 2027
LCG-SFNO uses low-rank Clebsch–Gordan tensor products to preserve SO(3) equivariance and improve weather forecasting accuracy. The SSHCG library makes these interactions computationally practical.
Accepted Embodied Spatial Reasoning Workshop at NeurIPS 2026 (Paper Track)
OrbitRIR uses frame averaging to enforce receiver-yaw invariance in few-shot room impulse response synthesis and supports 3D sound source localization through analysis-by-synthesis.
Project PDF will be released soon.
Co-first author
ASME IDETC-CIE 2026
Houston, TX, USA –
–Present
Advisor: Prof. Christine Allen-Blanchette Mentor: Tao Zhong
–
Advisor: Prof. Xianghua Xu (徐向华)
I developed a CNN surrogate to infer cell-wise correction coefficients from temperatures on two opposing surfaces, addressing non-orthogonal mesh errors in finite-volume heat-transfer simulations. I led the implementation, including ANSYS mesh parsing, and benchmarked the model against least-squares-based methods on simple 3D geometries.
Dual Bachelor's Degree Program in Mechanics and Advanced Clean Energy Program (Under CCE, Department of Energy and Power Engineering), Xingjian College
– (expected)
Department of Mechanical and Aerospace Engineering (MAE)
Exchange Student
–