combines image, text, audio, touch and etc. to accommodate various learning preferences, enhancing overall comprehension
adapts an AI model from one domain to another, enhancing its performance on related but different data distributionsย
optimizes an AI model using a small subset of its parameters while minimizing computational costs and memory usageย
aims to interpret a video scene by recognizing objects, their temporal relationships, and the surrounding events to enable meaningful understanding of the videoย
Join Our Lab ๐
Recruiting Undergraduate Interns,
Graduate Students & Postdoctoral Researchers
enables robots to understand visual data from their surroundings using cameras and multiple sensors, enhancing their autonomy and versatilityย
[Paper] 2026.05 ์ ์ ์ฒ, ์ด์นํ ํ์์ "SyMerge: From Non-Interference to Synergistic Merging via Single-Layer Adaptation" ๋ ผ๋ฌธ์ด ICML 2026์ accept๋์์ต๋๋ค. ์ถํํฉ๋๋ค!
[Project] 2026.03 AIM Lab์ "๋นํ์ค ์ผ์ ํ๊ฒฝ์ ๊ณ ๋ คํ ์-์ด๊ฐ ์ธ๊ณ๋ชจ๋ธ ๊ธฐ๋ฐ ์ ๋ฐ ๋ก๋ด ์กฐ์ ๊ธฐ์ "์ด ํ๊ตญ์ฐ๊ตฌ์ฌ๋จ(NRF) ์ฐ์์ฐ๊ตฌ-์ ์ง์ฐ๊ตฌ์ฌ์ ์ ์ ์ ๋์์ต๋๋ค.
[Paper] 2026.02 ์ฅ์ ์ง์, ๋ฅํ์ด๋ง, ์ ์ ์ฒ ํ์์ "ZOO-Prune: Training-Free Token Pruning via Zeroth-Order Gradient Estimation in Vision-Language Models" ๋
ผ๋ฌธ์ด CVPR 2026์ accept๋์์ต๋๋ค. ์ถํํฉ๋๋ค!
[Member] 2026.02 ๋ผํ์ฌ, ์ ์ ์ฒ, ์ด์นํ ํ์์ด IPIU 2026 ์ฐ์ ํฌ์คํฐ ๋ฐํ์์ ์์ํ์์ต๋๋ค. ์ถํํฉ๋๋ค.
[Paper] 2026.01 ํ์ฑ์ ๊ต์๋์ "Hyper-STTN: Hypergraph Augmented Spatial-Temporal Transformer Network for Trajectory Prediction" ๋
ผ๋ฌธ์ด IEEE International Conference on Robotics & Automation (ICRA) 2026 ํํ์ accept๋์์ต๋๋ค.
[Member] 2025.11 ์ต์์ค ํ์์ด ์์ธ์ฅํ์ฌ๋จ AI์์ธํ ํฌ ๋ํ์ ์ฅํ์์ผ๋ก ์ ์ ๋์์ต๋๋ค.
AI & Media Lab (AIM Lab)
Office : 90313, International Hall, Sungkyunkwan University,
25-2 Sungkyunkwan-ro, Jongno-gu, Seoul 03063, Korea
Email : csehong@skku.eduย ย Web : aim.skku.edu ย Tel : 02-740-1809