معرفی
Greg Dongyoon Han serves as a Senior Research Scientist at NAVER AI Lab since January 2018 and an Adjunct Professor at KAIST's Graduate School of AI since September 2021. His work bridges industrial AI research and academic instruction, focusing on foundational advancements in large-scale machine learning systems without maintaining a dedicated KAIST laboratory.
His research spans machine learning, deep learning, and multi-modal AI with emphasis on large language models, vision-language systems, and transformer architectures. Key themes include model efficiency through token compression, robustness against adversarial attacks, mathematical reasoning enhancement, and novel approaches to model merging and unlearning. His work consistently addresses scalability challenges in vision and language domains while exploring theoretical aspects of neural network dynamics.
Recent publications (2024-2025) reveal three dominant trends: 1) Token-level innovations for model compression and efficiency, 2) Robust adaptation techniques for vision-language systems, and 3) Mathematical reasoning augmentation for large language models. These contributions appear in top venues including NeurIPS, ICML, CVPR, and ICLR, often featuring novel architectural modifications and training paradigms.
Award highlights include:
- 4th place in ImageNet ILSVRC 2017 object localization
- Outstanding Reviewer at NeurIPS 2018 and CVPR 2021
- Outstanding Paper Award at ICACT 2014
- NAVER AI Lab's Best 2022 Paper for "Learning Features with Parameter-Free Layers"
Though not currently advising KAIST students directly, Han actively co-mentors research projects as evidenced by multiple "co-mentored project" publications. His academic service is extensive: Area Chair for ICLR 2026 and NeurIPS 2025, plus continuous reviewing for NeurIPS (2018-2024), ICLR (2019-2025), and CVPR (2018-2026). He maintains strong industry-academia collaboration through NAVER internships rather than traditional university lab structures.
Based at NAVER AI Lab, Han operates within South Korea's premier industrial research environment while contributing to KAIST's AI graduate education. His work demonstrates the growing synergy between corporate AI labs and academic institutions in advancing foundational machine learning research.
