Hi! I’m Kihyun, a 4th-year PhD student in MIT EECS. My research focuses on intelligent decision-making systems under uncertainty, with an emphasis on reinforcement learning (e.g., offline RL, inverse RL, imitation learning) and control theory (e.g., stochastic optimal control, robust control).
Recently, I have been exploring preference learning (e.g., RLHF, DPO) to align decision-making systems, particularly LLMs, with diverse human behaviors and preferences.
I am affiliated with MIT LIDS and am fortunate to be co-advised by Prof. Asuman Ozdaglar and Prof. Pablo Parrilo. Before joining MIT, I worked with Prof. Insoon Yang at Seoul National University, where I had the wonderful opportunity to deepen my theoretical background.
My research interests span the theoretical foundations of machine learning, optimization, and AI alignment. Feel free to reach out if you share similar interests!
Recent News
- Oct. 2024: Presented my work on offline inverse RL and RLHF at an invited session at the 2024 INFORMS Annual Meeting.
- Sep. 2024: Began TA for the graduate-level RL course 6.7920 Reinforcement Learning: Foundations and Methods at MIT.
- Jun. 2024: Started a ML research internship with the Advanced ML Lab at LG AI Research.
- May. 2024: Completed my MS degree (Thesis: “Offline Reward Learning from Human Demonstrations and Feedback: A Linear Programming Approach”)