Kimin Lee

Kimin Lee is a Faculty member at the Graduate School of AI at KAIST. They are affiliated with the Computer Science department and can be contacted via email at kiminlee@kaist.ac.kr. Their research focuses on the Kim Jaechul Graduate School of AI, with an emphasis on developing safe and capable decision-making agents. Their work includes topics such as Vision-Language Action Models, reinforcement learning from human feedback, AI safety evaluation, and human-centric AI. Lee has contributed to various conferences and publications, including ICML, CVPR, and ICLR, and has been involved in multiple research projects related to AI safety and alignment. They are also the Co-founder of Config.

Kimin Lee is a researcher specializing in Artificial Intelligence, with a focus on Reinforcement Learning and Deep Learning. Their work explores methods to enhance model robustness, detect out-of-distribution samples, and improve uncertainty estimation. Lee has contributed to frameworks for ensemble learning, offline-to-online reinforcement learning, and techniques for generalization in deep reinforcement learning. They have also investigated the use of pre-training to improve model performance and developed approaches for aligning text-to-image models using human feedback. Their research emphasizes the integration of self-knowledge distillation, randomization, and unsupervised learning to address challenges in reinforcement learning and model robustness.