Mingu Kang
M.S./Ph.D. Student, Seoul National University
gms5560@snu.ac.edu
Bio
I am an M.S./Ph.D. student at the VIP Lab, Seoul National University, advised by Prof. Joonseok Lee.
My research focuses on video understanding, with an emphasis on modeling temporal dynamics and high-level semantics in long-form videos. I am especially interested in building reliable multimodal representations for downstream tasks such as video summarization, video retrieval, and motion-related understanding.
News
- 2026.02 Our paper TripleSumm was accepted to ICLR 2026.
- 2025.06 Our paper DisCoRD was accepted to ICCV 2025 as a Highlight.
Publications
Research outputs across venues. * = equal contribution.
From Matching to Reasoning: Query-Aware Long Video Summarization
Preprint, 2026.
Do Not Stop at Frame Scoring: Enabling End-to-end Training for Video Summarization
Preprint, 2026.
From Matching to Reasoning: Query-Aware Long Video Summarization
Preprint, 2026.
DRUM: Dense Reliability Utilization Module for Robust Autonomous Driving Perception
Preprint, 2026.
Do Not Stop at Frame Scoring: Enabling End-to-end Training for Video Summarization
Preprint, 2026.
Research Topic
Click a topic to filter publications above.
Research Experience
Seoul National University Seoul, South Korea
M.S./Ph.D. Student Present
Advisor: Joonseok Lee
KAIST Graduate School of AI Seoul, South Korea
Research Intern Previously
Advisor: Beomjoon Kim
Sungkyunkwan University Suwon, South Korea
Research Intern Previously
Advisor: Yusung Kim
Awards
Best Research Award
Top-rated research selected by industry and government-funded institute evaluators.
Minister of Employment and Labor Award
Developed an AI model to predict the likelihood of pregnancy success in infertile patients.
Best Research Award
Top-rated research selected by industry and government-funded institute evaluators.
President's Award, Institute for Information & Communications Technology Planning & Evaluation (IITP)
Developed AI models for paragraph-level classification to predict whether text is human-written or AI-generated, leveraging contextual clues within the same document for reliable detection of generative AI content.
Intelligent Multimedia Research Center Director's Award
Developed prompts and retrieval-augmented generation strategies for fair and neutral responses based on biased context and user intent.
Minister of Employment and Labor Award
Developed an AI model to predict the likelihood of pregnancy success in infertile patients.
President's List, High Student Success Performance
Selected as one of the Top 20 students for outstanding achievement across Sungkyunkwan University.
Korea Water Resources Corporation President Award
Developed an AI model that can accurately detect abnormal timing and leakage sections of the water supply network.
Sungkyunkwan University brAIn Research Program Award
Conducted research on the image inpainting model, CoReaP: Collaborative Reconstruction with Assistive Priors.
Korea Agro-Fisheries Food Trade Corporation Award
Developed an AI model for agriculture price forecasting.
Korea Computer Education President Award
Developed a text-to-3D motion generation model and web production for rehabilitation exercise.
Hankuk University of Foreign Studies Presidential Award
Developed an AI model that classifies article categories from news title and text keywords.
SKKU Univ AI Director Award
Developed an AI model to detect illegal gambling advertisements on web pages.
National Institute of Korean Language Award
Developed an AI model to distinguish conversation context in Korean LLMs.
Dean's List Award, High Scholastic Achievement
Top 1% for the fall semester 2023.
Scholarship
Presidential Science Scholarship
Selected as one of the Top 60 nationwide for fostering world-class scientists in engineering.
CV
Full CV in PDF.