Minsu Kim

I’ll be joining Microsoft’s Copilot Tuning Research team as a Senior Researcher in 2026. There, I will work on advancing Copilot agents through research on LLM post-training, with a focus on reinforcement learning, agentic behavior, and reliable decision-making.

I am currently a postdoctoral fellow working with Prof. Yoshua Bengio at Mila, and Prof. Sungjin Ahn and Prof. Sungsoo Ahn at KAIST.

My research focuses on reinforcement learning, particularly on sample efficiency, credit assignment, and exploration, with applications to frontier LLMs, AI safety, and AI4Science.

Background

I received my Ph.D. from KAIST under the guidance of Prof. Jinkyoo Park, where I worked on bridging reinforcement learning and combinatorial optimization. My dissertation received the Presidential Best Ph.D. Thesis Award.

During my Ph.D., I collaborated with Prof. Sungsoo Ahn and conducted extended research visits at Mila, where I worked with Prof. Yoshua Bengio and his group on GFlowNet-related research.

I also received my master’s degree from KAIST under the supervision of Prof. Joungho Kim, where I worked on signal and power integrity in 2.5D/3D semiconductor systems, including HBM, using deep learning-based optimization methods.

Before that, I earned my B.S. in Mathematics and Computer Science from KAIST.

Academic Service

• Area Chair: NeurIPS (Position Paper Track, 2026)
• Reviewer (Conferences): NeurIPS (2022–2025), ICML (2023–2026), ICLR (2024–2026)
• Reviewer (Journals): TNNLS (2025–2026), TPAMI (2025), TMLR (2025)