Dohyeon Lee

I am an undergraduate student in the Department of IT Convergence Engineering at POSTECH in South Korea.
For robotic models, unlike LLMs, I believe the ability to learn data sequentially without forgetting is more important than the ability to fully train on a massive dataset.
To address this continual learning problem, I aim to answer the following questions:
When fine-tuning a robot model for a new task, • What information should the model focus on learning from the new data?
(ex: language-motion relationships, ..)
• What information is shared across different tasks?
(ex: physics knowledge, reusable skills, ..)
I am trying to address this problem from various perspectives, including RL, Control, VLA, and hardware development.

Email  /  Scholar  /  Youtube  /  CV