About me
I'm a first year PhD student at UCLA CS department, advised by Prof. Baharan Mirzasoleiman. I'm mostly interested in efficient and trustworthy machine learning algorithms especially in the context of reasoning LLMs and video diffusion models. Topics I've worked on include supervised learning with weak labels, adversarial attack, algorithmic fairness and distribution shift.
Before coming to UCLA, I did my undergrad at Yao Class, Tsinghua University, Beijing, China. I also spent a wonderful 6 months in 2024 as a research intern at CMU, advised by Prof. Steven Wu.
Page modified from codewithsadee.
Resume
Education
-
Computer Science Department, UCLA
Sep. 2025 - PresentPhD Student, advised by Prof. Baharan Mirzasoleiman
-
Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University, Beijing, China
Sep. 2021 - Jul. 2025AKA "Yao Class", an honors undergraduate program in CS. My curriculum mainly focuses on AI/ML.
GPA: 3.97/4.0, Rank: 2/89.
Experience
-
Research Intern @ Tencent
Jul. - Aug., 2025I was recruited through Tencent's Qingyun Program (青云计划). I spent two months at Tencent, Shanghai working on step distillation for video diffusion models.
-
Research Intern @ CMU
Feb. - Aug., 2024I was working on algorithmic fairness, more specifically, multicalibration, under the supervision of Prof. Steven Wu.
Projects
-
Accepted by NeurIPS 2025Discretization-free Multicalibration through Loss Minimization over Tree Ensembles
We propose a novel multicalibration method that directly optimizes over tree ensembles without discretization, achieving provable multicalibration guarantees.
-
Accepted by AAAI 2026
POSE: Phased One-Step Adversarial Equilibrium for Video Diffusion Models
A distillation framework that enables high-quality video generation from large-scale diffusion models in a single step through phased adversarial training.
-
Game Solution Searcher Based on Simulated Annealing (SA)
I devised a two-level sampling method and perform SA to search for a suboptimal solution.