Jaemin Kim

About Me

I am a Ph.D. student at Graduate School of AI, KAIST, where I also completed my master’s degree under the supervision of Prof. Jong Chul Ye.

My research focuses on controllable generative AI, exploring ways to steer the outputs of advanced models such as diffusion models and large language models (LLMs) efficiently. I work at the intersection of generative modeling, trajectory-level optimization, and black-box guidance, with the goal of making generative AI more controllable and interpretable.

News

• [2026.01] Our paper (Training-free Reward Guiding) is accecpted to ICLR 2026.
• [2026.01] Our paper (UniR) received a Silver prize at 32nd Samsung Humantech Paper Award.
• [2025.06] Our paper (Free²Guide) is accecpted to ICCV 2025.
• [2025.03] Two papers (MotionPrompt, FreeMCG) are accepted to CVPR 2025.
• [2025.03] I started Ph.D course in BIPSL, KAIST!

My work trajectory

• How can Diffusion Models (DMs) connect two arbitrary distributions? ([P1])
• Can Consistency Trajectory Models (CTMs) be generalized to connect arbitrary distributions? ([C1])
• How can Video DMs generate more natural motion through guidance? ([C3])
• Do the samples from DMs provide meaningful directions, as in XAI? ([C2])
• Can the samples from DMs serve as guidance? ([C4])
• Do LLMs serve as guidance in RL approaches? ([P2])
• Can DMs edit images to maximize reward in a training-free manner? ([C5])
• Can train LLMs with GRPO in prediction tasks? ([P3])

Publications

Services

Conference Reviewers

• ICLR 2026

Powered by Jekyll and Minimal Light theme.