Qiuyang Mang (忙秋阳)

Qiuyang Mang
email: qmang AT berkeley.edu

Hi, I’m Qiuyang Mang, a second-year CS PhD student in the Sky Computing Lab at UC Berkeley, advised by Prof. Alvin Cheung. I lead FrontierCS and FrontierSmith, a benchmark and data synthesis system for LLM-driven algorithm evolution on open-ended coding tasks. My research interests center on two themes:

  • Long-Horizon LLM Agents: Understanding, developing, and improving LLM agents for complex, multi-step optimization across data synthesis, test-time scaling, post-training algorithms, and domain-specific applications. Elo-per-token, FrontierCS, FrontierSmith, Argus, Combee, SkyDiscover

  • Machine Learning Systems: Efficient algorithms for ML workloads, from data-processing to inference scheduling. SVG-EAR, Continuum, PLOP

Prior to joining Berkeley, I received my B.E. from The Chinese University of Hong Kong, Shenzhen, where I was advised by Prof. Pinjia He. I also spent an unforgettable year as a research assistant at the National University of Singapore with Prof. Manuel Rigger. I was in the 46th ICPC World Finalist 🎈 and served as problem setters for regionals.

Mentees: Runyuan He, Kaiyuan Liu, Xuanyi Zhou

Selected Publications

* Equal contributionMentee first author

FrontierSmith pipeline overview

FrontierSmith: Synthesizing Open-Ended Coding Problems at Scale

Runyuan He*, Qiuyang Mang*, Shang Zhou, Kaiyuan Liu, Hanchen Li, Huanzhi Mao, Qizheng Zhang, Zerui Li, Bo Peng, Lufeng Cheng, Tianfu Fu, Yichuan Wang, Wenhao Chai, Jingbo Shang, Alex Dimakis, Joseph E. Gonzalez, Alvin Cheung

NeurIPS 2026Spotlight · 3.7%

A system for synthesizing open-ended coding problems at scale, connecting data generation, validation, and agent evaluation.

paper code slides talk blog