Shaoyang Guo
Undergraduate, Peking University School of Physics (Class of 2023)
Research Interests: Embodied Intelligence (VLA), Large Language Model Safety
Focused on embodied intelligence and large language model safety, with interdisciplinary experience in physics and computer science. Designed the PHYBench evaluation framework and proposed the "expression tree edit distance" metric to quantify the similarity and accuracy of physical reasoning answers.
Research Areas
- Embodied Intelligence (Vision-Language-Action, VLA)
- Large Language Model Safety (LLM Safety)
Research Experience
-
Mar 2025 – Present, PKU-PsiRobot Lab | Embodied Intelligence Research, Supervisor: Asst. Prof. Yaodong Yang
- Conduct research on Vision-Language-Action (VLA) models, exploring physical reasoning and practical applications in the field of embodied intelligence
-
May 2023 – May 2024, PKU LHCb Experiment Group | High-Energy Physics Data Analysis, Supervisor: Asst. Prof. Yanxi Zhang
- Participated in data analysis for the LHCb experiment in high-energy physics, gaining extensive experience in data processing and statistical analysis
Publications
PHYBench: Holistic Evaluation of Physical Perception and Reasoning in Large Language Models
arXiv:2504.16074, 2025
Co-authors: Qiu Shi et al.
Contributions: Designed data collection workflows, proposed the "expression tree edit distance" metric, and led experimental validation
Awards & Honors
- National Scholarship, Peking University (2023–2024)
- Team Captain, Peking University CUPT (2024)
- Gold Medal, Chinese Physics Olympiad (CPhO 2022)
- First Prize, National Olympiad in Informatics (NOIP 2020)
Skills & Tools
| Programming Languages | Python, C/C++, MATLAB, Wolfram Alpha |
| Deep Learning Frameworks | PyTorch, TensorFlow, vLLM, DeepSpeed |
| AI for Science | ROOT |
Contact
School of Physics, Peking University
5 Yiheyuan Road, Haidian District, Beijing
📧 guoshaoyang@stu.pku.edu.cn
🔗 GitHub Profile