Jongwook Choi

Papers

Total Citations

H-Index

About

Jongwook Choi is a leading researcher in reinforcement learning and unsupervised skill discovery, whose work has significantly advanced how agents learn diverse, reusable behaviors without external rewards. His most-cited paper, "Lipschitz-constrained Unsupervised Skill Discovery" (2022, 8 citations), tackles a critical limitation in mutual information-based skill discovery methods. Choi identified that these approaches often collapse to learning only a few skills, and introduced a novel Lipschitz constraint on the skill-conditioned policy to enforce smoothness and diversity. This contribution provides a principled solution to skill collapse, enabling more robust exploration and hierarchical learning in complex environments. His research sits at the intersection of unsupervised learning, control, and representation learning, with implications for robotics and autonomous systems. Beyond this work, Choi has contributed to deep reinforcement learning algorithms and policy optimization, earning recognition for his ability to blend theoretical rigor with practical algorithm design. His insights into skill discovery continue to influence researchers seeking to build agents that can autonomously acquire rich behavioral repertoires, making his work essential reading for those advancing unsupervised RL and lifelong learning.