Sangkyu Lee
M.S. student, Yonsei University
![prof_pic.jpg](/assets/img/prof_pic.jpg?f2f5d331c4bab8c495906511e3892ad1)
I am an M.S. student in the Multimodal Intelligence Research (MIR) Lab at Yonsei University, advised by Youngjae Yu. I was a research intern at Naver Cloud and participated in the alignment learning of HyperCLOVA X, a large language model specialized for Korean. Also, I was a technical lead for WoRV (World Model for Robotics and Vehicle Control) at maumAI. Before the M.S. degree course, I received a B.A. in Applied Statistics and a B.S. in Computer Science (double major) from Yonsei University.
I am interested in developing end-to-end embodied AI models that can interact with the physical world and predict the changes in the world that come from their actions. I believe that successful embodied AI can be achieved by combining experiences from the physical world and experiences from the models’ own imagination. To this end, I am researching topics about models for both perceiving and generating various modalities, how to align them to interact with the world, and how to pre-train them efficiently.
Currently, I’m looking for Ph.D. programs! (2025 Fall) 🙌
News
May 16, 2024 | Our paper is accepted as a main conference paper at ACL 2024! |
---|---|
Apr 02, 2024 | Technical report of HyperCLOVA X is finally released at ArXiv! |
Feb 17, 2024 | Check out our new preprint, Aligning Large Language Models by On-Policy Self-Judgment. |
Jan 15, 2024 | I am serving as a technical lead at maumAI to lead the project for developing WoRV. |
Jul 17, 2023 | I am starting a research internship at the Foundation Model team in Naver Cloud for HyperCLOVA X. |
Publications
- SEAL: Entangled White-box Watermarks on Low-Rank AdaptationIn submission, 2024
- MASS: Overcoming Language Bias in Image-Text MatchingIn submission, 2024