Minsu Kim

I am a Ph.D. student in AI at KAIST, advised by Se-Young Yun. My research focuses on reinforcement learning, reasoning, and the epistemic foundations of large language models, with a particular interest in making language models more reliable, verifiable, and capable of rigorous reasoning.

Previously, I received my M.S. in AI from KAIST, where I worked with James Thorne on the epistemology of language models and evidence-based belief formation.

• Reinforcement Learning for LLMs — reward shaping, self-improving systems, and reasoning-oriented training
• Theorem Proving — Lean-based formal reasoning, proof generation and Autoformalization
• Epistemology of LLMs —philosophical views of belief, evidence, and knowledge in language models