Junbo ZHAO 赵浚博
Currently pursuing a Master's degree in Computer Science at Artificial Intelligence School, Beijing Normal University
Currently pursuing a Master's degree in Computer Science at Artificial Intelligence School, Beijing Normal University
I am a Computer Science Master's student at Artificial Intelligence School, Beijing Normal University, under the supervision of Professor Hua Huang. I graduated with a Bachelor's degree from Minzu University of China, where I was honored as an Outstanding Graduate of Beijing.
My academic interests include Educational AI, Multimodal LLM, and Computer Vision. I'm passionate about developing AI systems that can enhance learning experiences and solve complex problems through multimodal understanding.
Outside of academia, I enjoy exploring new places, capturing moments through photography, and staying active. I believe that maintaining a balance between intellectual pursuits and personal interests is essential for creativity and well-being.
Master of Computer Science
Advisor: Professor Hua Huang (Dean, Distinguished Young Scholar)
Research Areas: Educational LMMs, Multimodal LLMs, Computer Vision
Major Courses: Computer Graphics, Computer Vision, etc.
Student Work: Teaching Assistant for Linear Algebra class (Mar 2024 – Jul 2024)
Bachelor of Computer Science
Honors: Outstanding Graduate of Beijing
Major Courses: Algorithm Analysis, Databases, Data Structures, Software Engineering, Data Mining, etc.
Student Work: President, Minzu University of China Computer Association (Sep 2021 – Jul 2022)
Authors:
Institutions: 1Beijing Normal University, 2TAL
Paper Link: https://arxiv.org/abs/2503.05543
Overview: We proposed a novel framework for geometry problem solving that leverages diagrammatic information to resolve textual ambiguities. Our approach combines multi-modal understanding with symbolic reasoning to achieve state-of-the-art performance.
Key Components:
Results: Pi-GPS achieves state-of-the-art results, outperforming previous neural-symbolic methods by nearly 10% on standard benchmarks, including Geometry3K (77.8% on choice tasks) and PGPS9K (69.8% on choice tasks).
MLLM Algorithm Intern
Paper Link: https://arxiv.org/abs/2509.17336
Participated in building UI-Agent tasks and visual multimodal planning models, contributing to model training, dataset design, and case analysis.
Main Responsibilities:
Achievements: Improved model accuracy and successfully facilitated the model's online deployment.
Beijing Normal University
15th Chinese Collegiate Computer Game Competition
11th Lanqiao Cup C/C++ Programming Competition, Beijing Division
Minzu University of China