Zixun Huang
Research Study Assistant at 北大
Haidian District, Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolZixun Huang is a fourth-year Statistics undergraduate at Peking University and a visiting student at UC Berkeley, specializing in theoretical machine learning and reinforcement learning theory. As a Research Study Assistant, he developed a Functional Scaling Law that models SGD intrinsic time to accurately predict loss trajectories across learning-rate schedules, work that was accepted as a NeurIPS 2025 Spotlight. His research validated efficiency gains on 0.1B–1B-scale LLMs and highlighted the practical benefits of warmup–stable–decay schedules. With seven years of hands-on experience in quantitative research, he bridges rigorous theory with empirical validation and collaborates closely with leading ML theorists. Quietly pragmatic, he pairs mathematical depth from his mathematics training with an experimental sensibility shaped by large-model optimization studies.
6 years of coding experience
Visiting Student, Visiting Student at University of California, BerkeleyBachelor of Science - BS Mathematics, Bachelor of Science - BS Mathematics at Peking University
Github Skills (8)
tower8
computer-vision8
valine8
lts7
javascript6
vue6
mathematics4
large-language-models2
Programming languages (3)
Github contributions (5)
CHNZYX/Auto_Simulated_Universe
May 2023 - Mar 2025
崩坏：星穹铁道 模拟宇宙自动化 （Honkai Star Rail - Auto Simulated Universe）
Contributions:41 releases, 43 PRs, 707 pushes in 1 year 10 months
CHNZYX/asu_version_latest
Aug 2023 - Mar 2024
Contributions:1 release, 20 pushes, 1 branch in 7 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.