Scholar
Song Han
Google Scholar ID: E0iCaa4AAAAJ
Massachusetts Institute of Technology
Citations & Impact
Citations
69,193
H-index
75
i10-index
153
Publications
20
Co-authors
9
list available
Publications
Resume (English only)
Academic Achievements
- The 'Deep Compression' paper is among the top-5 most cited in the 50-year history of ISCA (1953–2023)
- Best paper awards at ICLR'16, FPGA'17, and MLSys'24
- MLSys'24 award-winning work AWQ enables 4-bit quantization with over 19 million downloads on HuggingFace
- Recipient of NSF CAREER Award and Sloan Research Fellowship
- Named to MIT Technology Review’s '35 Innovators Under 35'
- Selected as one of IEEE’s 'AI’s 10 to Watch'
Research Experience
- Led a series of works on LLM quantization and acceleration, including SmoothQuant, AWQ, and StreamingLLM
- Proposed TinyML and Once-for-All Network (hardware-aware neural architecture search)
- Heads the HAN Lab, researching efficient generative AI, visual generation, and vision-language models
- Developed efficient visual generation models such as HART, SANA series, and DC-VideoGen
- Designed multiple sparse attention mechanisms including SpAtten, Radial Attention, and XAttention
- Built the VILA family of efficient vision-language models: VILA, VILA-U, and LongVILA
