Scholar
Kunjun Li
Google Scholar ID: z8lpsMsAAAAJ
National University of Singapore
Citations & Impact
Citations
32
H-index
3
i10-index
1
Publications
4
Co-authors
6
list available
Publications
Resume (English only)
Academic Achievements
- Publications:
- - 'Memory-Efficient Visual Autoregressive Modeling with Scale-Aware KV Cache' accepted to NeurIPS 2025.
- - 'TinyFusion: Diffusion Transformers Learned Shallow' selected as CVPR 2025 Highlighted Paper (3%).
- Awards:
- - Second Place in 2025 SkiTB Visual Tracking Challenge.
- - ACM/IEEE IPSN 2024 Best Demonstration Runner-Up award.
Research Experience
- Research intern at Princeton University under the supervision of Prof. Zhuang Liu.
Education
- Undergraduate student at National University of Singapore (NUS); Research intern at Princeton University, supervised by Prof. Zhuang Liu; Closely collaborates with Prof. Jenq-Neng Hwang from UW and Prof. Xinchao Wang from NUS.
Background
- Research Interests: Efficient Deep Learning, particularly optimizing the training and inference of LLMs, diffusion models, and multimodal models. Focuses on sparse attention, network pruning, and efficient architectures. Aims to achieve computational breakthroughs, making deep learning affordable and accessible to everyone, everywhere.
Miscellany
- Contact: kunjun@u.nus.edu
- LinkedIn, Github, Google Scholar
