UNIST Professor Han Seung-yeol's Team Has 3 Papers Accepted by Top AI Conference ICLR

Three research papers from Professor Han Seung-yeol's team at UNIST's Graduate School of Artificial Intelligence have been accepted by the International Conference on Learning Representations (ICLR), one of the world's top three artificial intelligence (AI) conferences. ICLR is recognized as one of the top three AI conferences globally, alongside the Conference on Neural Information Processing Systems (NeurIPS) and the International Conference on Machine Learning (ICML). This year, approximately 19,000 papers were submitted to ICLR, with only about 27% passing the review process. The simultaneous acceptance of three papers from a single lab is a rare occurrence.

These achievements stem from advancements in reinforcement learning, a crucial technology for realizing Physical AI. The research team developed the 'Self-Imitation Skill Learning' (SISL) method, which effectively trains AI using offline data collected from industrial sites. They also developed the 'Strict Subgoal Execution' (SSE) learning technique, which enhances success rates in complex, long-horizon tasks, and the 'Sequential Sub-value Q-Learning' (S2Q) technique, which resolves optimization problems in scenarios involving multiple AI agents collaborating.

Researchers Lee Sang-hyun, Hwang Jae-bak, and Cho Yong-hyun participated as the first authors for each respective study. These research efforts were supported by the Institute for Information & Communications Technology Planning & Evaluation (IITP) and the National Research Foundation of Korea, funded by the Ministry of Science and ICT. Professor Han Seung-yeol stated that this research demonstrates the potential for stable application of reinforcement learning even with limited data and in uncertain environments, anticipating its expansion into various industrial sectors such as autonomous driving, robotics, and smart manufacturing.