Harper Hua

My Chinese name is 花硕.

prof_pic.jpg

I am a M.S. student in Computer Science at Stanford University, where I am fortunate to work with Prof. Jure Leskovec and Stefano Ermon. Previously, I earned my B.S. from Tsinghua University. My primary research interests include post-training for LLM agents, reinforcement learning, and adapting/customizing agents for novel tasks and complex systems. I’m also broadly interested in long-context LLMs, privacy-preserving AI, and multimodal diffusion models.

In summer 2025, I was an Applied Scientist Intern at Amazon Web Service Bedrock, where I worked on multi-turn RL post-training for code generation.

I’m always happy to chat or collaborate. Feel free to reach out!

Email: shuohua [at] stanford.edu

LinkedIn/Google Scholar/Github

selected publications

  1. ACL
    SQL-Trail: Multi-Turn Reinforcement Learning with Interleaved Feedback for Text-to-SQL
    Harper Hua, Zhen Han, Zhengyuan Shen, and 9 more authors
    In ACL Main Conference, 2026
    HTML
  2. ICML
    DSGym: A Holistic Framework for Evaluating and Training Data Science Agents
    Fan Nie*, Junlin Wang*, Harper Hua*, and 6 more authors
    In ICML, 2026
    HTML
  3. NeurIPS
    ResearchCodeBench: Benchmarking LLMs on Implementing Novel Machine Learning Research Code
    Tianyu Hua, Harper Hua, Violet Xiang, and 5 more authors
    In NeurIPS Dataset & Benchmark Track (Spotlight), 2025
    HTML
  4. ICLR
    TabDiff: a Mixed-type Diffusion Model for Tabular Data Generation
    Juntong Shi*, Minkai Xu*, Harper Hua*, and 3 more authors
    In ICLR, 2025
    HTML

teaching

Winter 2026 - Stanford University, CS246: Mining Massive Data Sets
Fall 2025, Fall 2024 - Stanford University, CS224W: Machine Learning with Graphs