Junhyun Ha

Hi! I am a master’s student at KAIST AI, advised by Prof. Juho Lee. I am a member of the SIML lab. Before joining KAIST, I received my B.S. in Statistics from Seoul National University in 2025.

My research focuses on reinforcement learning and policy improvement. I am interested in how agents can improve their behavior through data and interaction, and how exploration and learned representations can support this process.

A question that motivates my research is how agents can decide what to explore next and make the most of the experience they collect. Ultimately, I aim to develop learning algorithms that turn limited interaction into reliable improvements in behavior.

latest posts

Oct 01, 2026	PReFlow: Learning to Refine Diffusion Policies

publications

Diffusion Policy Improvement with Proposal-Conditioned Refinement Flows

Junhyun Ha, Juho Lee, and Byoungwoo Park

arXiv preprint arXiv:2609.36812, Sep 2026

arXiv Bib PDF Blog

@article{ha2026preflow,
  title = {Diffusion Policy Improvement with Proposal-Conditioned Refinement Flows},
  author = {Ha, Junhyun and Lee, Juho and Park, Byoungwoo},
  journal = {arXiv preprint arXiv:2609.36812},
  year = {2026},
  month = sep,
  url = {https://arxiv.org/abs/2609.36812},
}