Hang Yu

About Me

I am a first-year master’s student at Tongji University supervised by Prof.Junqiao Zhao. I am interested in reinforcement learning, LLM reasoning and embodied intelligence, with a focus on improving agent generalization in dynamic environments. Ultimately, I aim to build generalizable action models to enable intelligent and versatile robot behaviors.

Research Interests

• Reinforcement Learning: model-based RL, offline RL
• Large Language Model(LLM): reasoning, RLHF
• Emboddied AI: Visual Prompt, DAgger

Experience

• Spirit AI Research Intern 「Aug. 2025 – Present」
  • Supervised by Prof. Yang Gao and Junyuan Xie
  • Involved in algorithm design and optimization for Vision-Language-Action (VLA) models, with deployment on real-world robotic systems.
  • Currently working on Visual Prompt and Dagger research.
• Graz University of Technology Exchange Researcher (Austria) 「March. 2025 – July. 2024」
  • funded by Erasmus+, work with Prof. Eduardo Veas on RL application
• TAL MathGPT Group research intern 「March. 2023 – July. 2023」
  • Contributed to the RLHF pipeline design, including reward modeling, and PPO/DPO alignment for TAL’s MathGPT, a math-focused large model used in education scenarios.
• Master’s Degree 「Sep. 2024 – Present」
  • Dept. of Computer Science and Technology, Tongji University
• Bachelor’s Degree 「Sep. 2020 – Jul. 2024 」
  • Dept. of Computer Science and Technology, Tongji University
  • Overall Ranking College No.1/113
  • Outstanding Graduate of Shanghai 2024
  • First Prize Tongji University Excellent Student Scholarship, 2020 & 2021 & 2022 & 2023
  • First Prize in the National Finals of the 17th Undergraduate Intelligent Vehicle Competition

Publications