Jiajun Cao 曹家骏

About Me

Hi! I’m Jiajun Cao (曹家骏), a Ph.D. student at the School of Computer Science, Peking University, advised by Prof. Shanghang Zhang. My research focuses on multimodal large models, world-action models, and embodied intelligence.

I received my B.Eng. in Artificial Intelligence from Xi’an Jiaotong University. I have worked on world models at the Beijing Innovation Center of Humanoid Robotics and on multimodal perception and autonomous driving at XPeng and NIO. Feel free to get in touch!

Research Interests

How can multimodal models perceive efficiently and act reliably in the physical world?

• Embodied intelligence: connecting language reasoning, world modeling, and robotic action.
• Efficient multimodal learning: visual knowledge distillation, token pruning, and expert specialization.
• Vision-language-action models: perception and planning for autonomous driving and manipulation.

News

• [Oct 2026] LAW is released, unifying language reasoning, action generation, and world modeling.
• [2026] SafeDojo is accepted to CoRL 2026.
• [Jun 2026] WAM4D is released.
• [Mar 2026] EvoDriveVLA is released.
• [2026] FastDriveVLA is accepted to AAAI 2026.

* Equal contribution. My name is highlighted.

Research Experience

Education

Honors & Awards

• Baosteel Excellent Student Scholarship
• National Silver Award, China International College Students’ “Internet+” Innovation and Entrepreneurship Competition
• Megvii Comprehensive Development Scholarship
• Megvii Distinctive Development Scholarship

Professional Service

Conference reviewer: NeurIPS, ICML, ICLR, AAAI, ECCV

Journal reviewer: IEEE TPAMI, SCIENCE CHINA Information Sciences