About

Subin Park.

Undergraduate at Yonsei University, double majoring in Computer Science and Mathematics, with research interests in NLP, reinforcement learning, and foundation models.

Now

I am pursuing a B.S. at Yonsei University, double majoring in Computer Science and Mathematics. My coursework spans software engineering, computer systems, deep neural networks, mathematics for deep learning, probability and statistics, text mining, and cognitive neuroscience.

I am an Undergraduate Researcher at the Decision Making Intelligence & Learning Lab, advised by Prof. Jongmin Lee. I am leading research on Offline Direct Multi-Preference LLM Alignment via RLHF and participating in research on RL foundation models for world modeling.

You can reach me anytime at enlightkorean@gmail.com.

Interests

AI & Machine Learning

NLP, reinforcement learning, and foundation models, viewed through computer science, mathematics, and cognitive neuroscience.

Natural Language Processing

LLM alignment, transformer-alternative architectures such as Mamba and Titans, and LLM reasoning.

Reinforcement Learning

Offline RL, sequential decision-making problems, data-efficient RL, and RLHF for preference alignment.

Research Foundations

Mathematics, linguistics, and cognitive neuroscience as tools for studying intelligence.

Timeline

Apr 2025 - now
Undergraduate Researcher
Decision Making Intelligence & Learning Lab, Yonsei University
Research on Offline Direct Multi-Preference LLM Alignment via RLHF and RL foundation models for world modeling.
Mar 2022 - now
B.S. Computer Science & Mathematics
Yonsei University, Seoul
Double major with coursework in computer systems, deep learning, probability and statistics, text mining, and cognitive neuroscience.
Jul 2023 - Jan 2025
Military Intelligence Specialist, Sergeant
Republic of Korea Army
Gathered and analyzed tactical intelligence, led a squad, and completed CBRN threat analysis and mission planning training.
Apr 2023 - Jun 2023
Polyglot Project Member
EleutherAI
Contributed to Korean Chain-of-Thought dataset construction and multilingual LLM preprocessing and evaluation.

Goals

Research. Study the fundamental question of whether machines can think by grounding intelligence in mathematics, linguistics, and cognitive neuroscience.

Alignment and reasoning. Build better language-model alignment methods, reasoning systems, and foundation models for sequential decision making.

Implementation. Keep research tied to working systems through PyTorch, HuggingFace, reinforcement learning, and careful evaluation.

Contact