About Me
I am an incoming second-year undergraduate at the Paris Elite Institute of Technology (SPEIT), Shanghai Jiao Tong University, following a French engineering curriculum with foundations in mathematics, physics, and computer science.
我是上海交通大学巴黎卓越工程师学院即将进入二年级的本科生，接受以数学、物理与计算机为基础的法语工程师教育。
My prior research experience has been in LLM and software-agent evaluation — in particular, studying how LLM agents behave and fail. Three preprints from this line of work are available on Research Square (see Publications). I am now actively expanding toward embodied intelligence and robotics, while continuing to explore AI evaluation. Current work in progress includes:
- Building counterfactual benchmarks to study how LLM agents recover from ambiguous tool outcomes
- Applying mutation testing and state-oracle calibration to evaluate the reliability of software-agent benchmarks
- Investigating how JSON Schema serialization order affects black-box LLM generation
I am particularly interested in bringing prior AI and computational research experience into embodied intelligence and robotics（具身智能与机器人学）. I also enjoy learning about mechanical systems（机械）。
Education | 教育背景
- B.Eng. — Shanghai Jiao Tong University（上海交通大学）
- School: Paris Elite Institute of Technology (SPEIT)（巴黎卓越工程师学院）
- 2025 – Present, Shanghai, China
Areas of Interest | 兴趣领域
- Large language models（大语言模型）
- Embodied intelligence and robotics（具身智能与机器人学）
- Mechanical systems（机械）
Beyond Research | 研究之外
Outside of academics, I enjoy vocaloid music production (so-vits-svc) and studying historical orders and decorations (phaleristics). 业余时间喜欢 vocaloid 音乐制作与历史勋赏研究。
I also build things outside of research:
- VEX Robotics — served as the team’s lead programmer and driver for VEX V5 robot control code; the team won 2nd Prize at the 2025 SJTU campus competition
- Snowbound School Mystery（雪闭校园） — a Ren’Py suspense visual novel set in a snowbound school, with 3 branching routes; play online