CV

name: Zhiyuan Wei(魏知原)

Undergraduate, 3rd year

School of Electronics Engineering and Computer Science(EECS), Peking University

Research Interests

• Computer Vision, Embodied AI, World Models, Robotics, Machine Learning, Reinforcement Learning etc.

Education

• Peking University (PKU): 2023.8$\rightarrow$2027.6 (not graduated yet)

Research Experience

Research Intern, PKU-GALBOT Embodied AI Joint Laboratory

Project: FoldNet++ (Supervised by PhD Candidate Yuxing Chen, Prof. He Wang’s Group)

Focusing on Generalizable Deformable Object Manipulation using VLA models and synthetic data.

• VLA For Deformable Objects: Fine-tuned foundation models for long-horizon laundry folding capabilities, validating the generalization potential of VLA architectures in complex deformable manipulation tasks.
• Generative Data Pipeline: Developed a Diffusion & LLM-driven data synthesis pipeline to generate scalable, high-diversity training environments. Demonstrated that broad synthetic data distribution significantly enhances policy robustness.
• World Model Integration: Proposed a World-Model based synthetic data generation framework, exploring the use of video generative models as a third data source beyond real-world demos and physics simulations for Embodied AI training.

Technical Skills

Programming: Python, C/C++, CUDA ML Frameworks: PyTorch Simulators: Mujoco, Style3D