Scholar
Kangsan Kim
Google Scholar ID: 9awek3YAAAAJ
Ph.D. student, KAIST
Citations & Impact
Citations
30
H-index
4
i10-index
1
Publications
4
Co-authors
8
list available
Publications
Resume (English only)
Academic Achievements
- Paper 'VideoICL: Confidence-based Iterative In-context Learning for Out-of-Distribution Video Understanding' accepted to CVPR 2025 (equal contribution)
- Paper 'VideoRAG: Retrieval-Augmented Generation over Video Corpus' accepted to ACL Findings 2025 (equal contribution)
- Preprint 'UniversalRAG: Retrieval-Augmented Generation over Corpora of Diverse Modalities and Granularities' released on arXiv in 2025 (equal contribution)
- Preprint 'HoliSafe: Holistic Safety Benchmarking and Modeling with Safety Meta Token for Vision-Language Model' released on arXiv in 2025
- Recipient of the Qualcomm-KAIST Innovation Award in 2023
- Dean’s List, College of Engineering, KAIST in 2020
- Invited talks on VideoRAG at Multimodal Weekly (hosted by TwelveLabs) and NYU Global AI Frontier Lab in 2025
Background
- Ph.D. student at the Graduate School of AI, KAIST, member of the MLAI lab
- Advised by Prof. Sung Ju Hwang
- Research focuses on developing multimodal large language models (MLLMs) that understand the world and interact with humans through visual data
- Previously worked on video understanding and multimodal Retrieval-Augmented Generation (RAG)
- Interested in embodied AI models operating on egocentric video requiring spatial reasoning
