Sunwoo Bae

I am an undergraduate student at KAIST, currently interning at the COCO Lab under the supervision of Hyunwoo Kim.

My research focuses on understanding how intelligent systems learn from experience and use accumulated information for reasoning, planning, and decision-making. My current work investigates agent memory and long-horizon reasoning, while I am also interested in reinforcement learning and post-training methods for improving the reasoning and adaptive capabilities of language models.

These interests naturally lead me to world modeling: how an intelligent system can form predictive representations of its environment and use them to reason about future outcomes. I am particularly interested in social world modeling—how models can infer and represent human beliefs, intentions, social contexts, and interaction dynamics—and, more broadly, in how social and physical world models can support robust decision-making in complex environments.