딥러닝학습추론지능 연구실 (LRNING)

I'm a researcher, an assistant professor of Computer Science and Artificial Intelligence at Hanyang University, the principal investigator of LRNING, and an associate member of KIAS. My research goal is to understand fundamental mechanisms of computational learning and reasoning, and to make AI safe and reliable. I hope my/our research contributes to solving important problems and developing technologies that benefit people and society. We show that the implicit bias of large-step Gradient Descent toward flat minima overrides the architectural bias of multi-pathway deep linear networks, reversing the "winner-takes-all" symmetry breaking predicted by Gradient Flow. Instead of collapsing onto a single dominant pathway, the network is driven into a re-balancing phase at the Edge of Stability that restores shared representations across parallel pathways. We propose a novel method for localizing memorization in diffusion models using coordinate-wise curvature differences, offering a new geometric interpretation of score-difference-based metrics. We propose Inconsistency-Aware Minimization (IAM), which recovers SAM's flat-minima bias from unlabeled data alone by regularizing local inconsistency, an output-sensitivity measure tied to the largest eigenvalue of the Fisher Information Matrix. IAM matches SAM in supervised learning while uniquely leveraging unlabeled data to improve semi- and self-supervised methods such as FixMatch and SimCLR. We investigate the emergence and dynamics of the two modes of in-context learning during pretraining: task recognition and task learning. We show that the model first learns the task learning and the task recognition abilities together in the beginning, but it (a) gradually forgets the task recognition ability to recall the priorly learned