About Me
Hi, my name is Jaeik (Jake) Kim, and I am a student researcher specializing in ML and AI. I am an incoming integrated M.S. & Ph.D. student at the AIDAS (AI, Big Data, and System) Lab at Seoul National University (SNU), where I will be advised by Prof. Jaeyoung Do.
I am particularly interested in efficient and scalable approaches to vision and large language models (LLMs), especially in areas such as model merging, test-time adaptation, and continual learning.
Papers
[2025.01] Exploring and Leveraging Class Vectors for Classifier Editing
Under review
[2025.01] SECOND: Mitigating Perceptual Hallucination in Vision-Language Models via Selective and Contrastive Decoding
Under review
[2024.11] Refining Task Vectors for Enhanced Model Merging with Dynamic Denoising
Under review
[2024.08] Locality Guidance with Dual-Path Tokens for Semantic Segmentation
Under review
Experience
Research on the Optimization of Model Merging and representation Learning.
Advised by Prof. Jaeyoung Do
Research on Computer Vision (Semantic Segmentation), PEFT (Parameter-Efficient Fine-Tuning) and Continual Learning.
Advised by Prof. Sungeun Hong
Education
Seoul National University
Integrated M.S & Ph.D. in Interdisciplinary Program in Artificial Intelligence (IPAI)
2025 -
Sungkyunkwan University
B.S. in Mathematics & Applied Artifical intelligence.
2021 - 2025 (Expected)