About Me
I'm now a First-year M.S. student in Computer Science, UIUC. I obtained my Bachelor's Degree from ZJU-UIUC Institute. My research interests mainly lie in multimodal learning, VLM reasoning, and visual grounding.
Research Projects
TextRegion: Text-Aligned Region Tokens from Frozen Image-Text Models
UIUC · Sep 2024 – Apr 2025 · Accepted at TMLR (J2C, top 10%)
Proposed TextRegion, a training-free method to extract text-aligned region-level representations from frozen image-text models (CLIP, SigLIP) that natively only produce image-level features. Core idea: inject SAM2 segmentation masks as attention constraints into the final transformer layer, forcing region tokens to aggregate only region-relevant patches. Designed multi-resolution patch encoding, implemented SAM2 mask merging, and built the inference and evaluation pipeline across open-world segmentation, referring expression comprehension, and grounding.
Personal Projects
LeetCode Agent: AI-Powered Terminal Problem-Solving CLI
Python, DeepSeek API, SQLite, prompt_toolkit, Rich · Mar 2026 · GitHub
Built a terminal CLI integrating LeetCode GraphQL API and CodeTop API, supporting natural-language problem selection, contextual hints, code review, and company-targeted high-frequency problem recommendations. Designed a dual-layer interaction architecture: 12 slash commands handle deterministic operations while only intent-requiring interactions are routed to the AI Agent. Implemented a spaced-repetition scheduler driven by struggle score and hint count for personalized review plans.
Internship Experience
Tencent — MLE Intern
Guangzhou, China · Oct 2025 – Dec 2025
Huawei — AI Engineer Intern
Shenzhen, China · Jun 2024 – Aug 2024