PhD student @ HKU CDS
Undergrad student @ HUST EICHi! I am a first year PhD student at The University of Hong Kong (HKU), School of Computing and Data Science, supervised by Prof. Hengshuang Zhao. Previously, I received my B.Eng. from the School of Electronic Information and Communications at Huazhong University of Science and Technology (HUST) with a GPA of 3.98/4.0 (Rank: 1/155). During my undergraduate studies, I was fortunate to be supervised by Prof. Xiang Bai. My current research interests primarily lie in Multimodal Large Language Models and Reinforcement Learning.
Warning
Problem: The current name of your GitHub Pages repository ("Solution: Please consider renaming the repository to "
http://".
However, if the current repository name is intended, you can ignore this message by removing "{% include widgets/debug_repo_name.html %}" in index.html.
Action required
Problem: The current root path of this site is "baseurl ("_config.yml.
Solution: Please set the
baseurl in _config.yml to "Education
-
The University of Hong KongSchool of Computing and Data Science
Ph.D. StudentSept. 2025 - Present -
Huazhong University of Science and TechnologyB.S. in Electronic Information and CommunicationsSept. 2021 - Jul. 2025
Experience
-
TiktokResearch InternNov. 2024 - Present
Honors & Awards
-
National Scholarship2024
-
National Scholarship2023
-
National Scholarship2022
-
Outstanding Undergraduate2023
-
Meritorious Winner in Mathematical Contest In Modeling (MCM)2023
News
Selected Publications (view all )
Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search
Xin Lai*, Junyi Li*, Wei Li, Tao Liu, Tianjian Li, Hengshuang Zhao# (* equal contribution, # corresponding author)
Under review. 2025
A full training recipe to reproduce OpenAI o3-style thinking-with-images capability.
Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search
Xin Lai*, Junyi Li*, Wei Li, Tao Liu, Tianjian Li, Hengshuang Zhao# (* equal contribution, # corresponding author)
Under review. 2025
A full training recipe to reproduce OpenAI o3-style thinking-with-images capability.
VisionThink: Smart and Efficient Vision Language Model via Reinforcement Learning
Senqiao Yang*, Junyi Li*, Xin Lai*, Bei Yu, Hengshuang Zhao, Jiaya Jia (* equal contribution)
Advances in Neural Information Processing Systems (NeurIPS) 2025 Poster
A new paradigm of efficient VLMs with token compression.
VisionThink: Smart and Efficient Vision Language Model via Reinforcement Learning
Senqiao Yang*, Junyi Li*, Xin Lai*, Bei Yu, Hengshuang Zhao, Jiaya Jia (* equal contribution)
Advances in Neural Information Processing Systems (NeurIPS) 2025 Poster
A new paradigm of efficient VLMs with token compression.
TokBench: Evaluating Your Visual Tokenizer before Visual Generation
Junfeng Wu*, Dongliang Luo*, Weizhi Zhao, Zhihao Xie, Yuanhao Wang, Junyi Li, Xudong Xie, Yuliang Liu, Xiang Bai# (* equal contribution, # corresponding author)
Under review. 2025
A simple Benchmark for evaluating your visual tokenizer.
TokBench: Evaluating Your Visual Tokenizer before Visual Generation
Junfeng Wu*, Dongliang Luo*, Weizhi Zhao, Zhihao Xie, Yuanhao Wang, Junyi Li, Xudong Xie, Yuliang Liu, Xiang Bai# (* equal contribution, # corresponding author)
Under review. 2025
A simple Benchmark for evaluating your visual tokenizer.
PartGLEE: A Foundation Model for Recognizing and Parsing Any Objects
Junyi Li*, Junfeng Wu*, Weizhi Zhao, Song Bai, Xiang Bai# (* equal contribution, # corresponding author)
Eurepean Conference on Computer Vision (ECCV) 2024 Poster
The first part-level foundation model for locating and identifying both objects and parts in images through a hierarchical framework.
PartGLEE: A Foundation Model for Recognizing and Parsing Any Objects
Junyi Li*, Junfeng Wu*, Weizhi Zhao, Song Bai, Xiang Bai# (* equal contribution, # corresponding author)
Eurepean Conference on Computer Vision (ECCV) 2024 Poster
The first part-level foundation model for locating and identifying both objects and parts in images through a hierarchical framework.