Our demo paper, “Looking at Your Photo, What Comes to Mind? Personalized Memory Internalization for Dementia Reminiscence,” has been accepted to the IJCAI 2026 Demo Track. This system represents a demonstration outcome of our five-year Human-Centered AI project, developed since last year to support memory recall for elderly individuals with dementia. Many thanks to everyone for their hard work, especially to Shunjie for leading the writing, and congratulations to all!

Our consortium, led by Polaris Office, has been selected as one of only two funded consortia in an IITP R&D program on lightweight AI training and inference. The project, titled “Development of Ultra-Efficient Lightweight AI Model Technologies Specialized for Document Collaboration toward Digital Sovereignty,” will receive KRW 7.5 billion over four years. Our lab will contribute to the development of efficient AI inference technologies for document collaboration, with a particular focus on task/context-aware reasoning optimization, token pruning, and other lightweight inference techniques.

Our paper, "PGB: One-Shot Pruning for BERT via Weight Grouping and Permutation ," (1st author: Hyemin Lim, MS alumna) has been accepted for publication in Journal of Artificial Intelligence Research, which is a prestigious SCIE journal (IF 4.6) in artificial intelligence. Well deserved, Hyemin & Jaeyeon!

Our paper, "STARK: Structure-Aware and Adaptive Representation Learning for Continual Knowledge Graph Embedding ," (1st author: Kyung-Hwan Lee, MS Student) has been accepted for the WWW 2026 (ACM The Web) conference, which is a top-tier conference in the fields of web-based data mining and AI (BK 4).  Congratulations, Kyung-Hwan!

Appeared in [서울신문] [기호일보] [경기신문]  

Our paper, "Balanced Online Class-Incremental Learning via Dual Classifiers ," (1st author: Shunjie Wen, PhD Student) has been accepted for the ACM SAC 2026 conference, which is a reputable international venue in applied computing research (BK 1).  Congratulations, Shunjie!