I am Boyao Han, a senior undergraduate at the College of Computer Science and Electronic Engineering, Hunan University. I am currently a research assistant at the School of Data Science, The Chinese University of Hong Kong, Shenzhen, advised by Prof. Li Jiang. I will be joining the CUHK-SZ as an MPhil student in Computer Science in Fall 2026.

My research interests lie in Embodied AI and Computer Vision.

I am actively looking for research internship opportunities. Please feel free to reach out if our interests align.

Research Interests

Embodied AI

My research focuses on embodied artificial intelligence, with an emphasis on improving the generalization of Vision-Language-Action (VLA) models in real-world environments.

I am particularly interested in the role of scaling laws in embodied agents — how large-scale real-world data can be leveraged to enhance the robustness and generalization of policies across diverse tasks and environments. I also believe that VLA models relying solely on RGB input struggle to handle the complexity of real-world interaction; incorporating physical signals such as depth and tactile feedback is essential for modeling environmental structure and enabling fine-grained manipulation.

Computer Vision

In computer vision, I focus on the application of 3D Computer Vision to embodied intelligence, including how to build structured 3D representations of the environment, improve spatial understanding, and combine multimodal information for scene modeling and decision making.

I am also interested in world models in the context of embodied AI. By learning the dynamics and structure of the environment, world models can provide embodied agents with stronger predictive capabilities, serving as a foundation for higher-level spatial intelligence.

News

• 2026.04 — GeoPredict: Leveraging Predictive Kinematics and 3D Gaussian Geometry for Precise VLA Manipulation accepted as Highlight at CVPR 2026.
• 2025.11 — PointSLAM++: Robust Dense Neural Gaussian Point Cloud-based SLAM accepted as Poster at AAAI 2026.
• 2025.10 — Memory Forcing: Spatio-temporal Memory for Consistent Scene Generation on Minecraft is available on arXiv.
• 2025.07 — Joined the School of Data Science, CUHK-Shenzhen as a research assistant, advised by Prof. Li Jiang.

Selected Publications

• GeoPredict: Leveraging Predictive Kinematics and 3D Gaussian Geometry for Precise VLA Manipulation. CVPR 2026 (Highlight).
• Memory Forcing: Spatio-temporal Memory for Consistent Scene Generation on Minecraft. arXiv preprint, 2025.
• PointSLAM++: Robust Dense Neural Gaussian Point Cloud-based SLAM. AAAI 2026 (Poster).

A complete list is available on the Publications page.

Contact

Feel free to contact me for discussion or collaboration! You can reach me at boyaohan22@gmail.com.