Papers
8
Total Citations
97
H-Index
6
About
Younggyo Seo is a machine learning researcher whose work spans model-based reinforcement learning, reward learning, and vision-based robotic manipulation. His research addresses some of the most pressing challenges in making reinforcement learning more practical, data-efficient, and generalizable across real-world settings. Seo's most-cited work, "Context-aware Dynamics Model for Generalization in Model-Based Reinforcement Learning" (2020, 28 citations), tackles the difficult problem of training dynamics models that transfer across varying environments — a critical bottleneck in deploying RL agents at scale. His subsequent work on offline-to-online RL (2021, 18 citations) advanced the field's understanding of how to safely and effectively fine-tune agents pretrained on offline datasets, bridging a key gap between static training and live deployment. Beyond dynamics modeling, Seo has made notable contributions to feedback-efficient preference-based RL through semi-supervised reward learning (SURF, 2022) and explored language-conditioned reward modulation for sparse-reward tasks. More recently, his research has pushed into multimodal robot learning, combining vision and touch for dexterous manipulation and developing masked world models for visual control — work that reflects a growing ambition to build robots that perceive and act more like humans.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5Masked World Models for Visual Control10 citations · 2022
- 6
- 7Multi-View Masked World Models for Visual Robotic Manipulation6 citations · 2023
- 8Language Reward Modulation for Pretraining Reinforcement Learning4 citations · 2023