- Paper 'HoliTom: Holistic Token Merging for Fast Video Large Language Models' accepted by NeurIPS 2025
- Published a systematic review of multimodal long-context token compression methods
- Received Zhejiang University Outstanding Graduates
- Introduced HoliTom, a training-free holistic token merge method for fast video LLMs
- Paper 'LI-GS: Gaussian Splatting with LiDAR Incorporated for Accurate Large-Scale Reconstruction' accepted by RA-L'25
Education
Incoming Ph.D. student at ENCODE Lab, Westlake University, advised by Prof. Huan Wang in a joint program with Zhejiang University. Currently, in the final year of undergraduate studies at the College of Control Science and Engineering, Zhejiang University, while also enrolled in the Advanced Honor Class of Engineering Education at Chu Kochen Honors College.
Background
Research interests: Efficient AI and Machine Learning Systems (MLSys), focusing on model compression and acceleration. Research spans computer vision, large language models (LLMs), and embodied AI.