Papers
2
Total Citations
21
H-Index
2
About
Yubin Kim is a rising researcher at the forefront of robotic manipulation, specializing in the intersection of generative modeling and geometric deep learning. Her primary research focuses on developing equivariant diffusion models for visual robotic manipulation, a field that seeks to enable robots to learn complex, dexterous tasks from human demonstrations. Kim’s major contribution is the introduction of Diffusion-EDFs, a novel framework that applies bi-equivariant denoising generative modeling on the SE(3) group—the space of rotations and translations in 3D space. This work, detailed in her most-cited paper (2024, 18 citations), demonstrates how incorporating SE(3) equivariance into diffusion models allows robots to generalize manipulation skills across different orientations and positions, significantly improving sample efficiency and robustness. By grounding generative AI in the physical symmetries of the world, Kim’s approach addresses a critical challenge in robotics: learning from stochastic, varied human demonstrations while maintaining spatial consistency. Her research, which has already garnered over 21 citations in a short span, represents a promising step toward more adaptable and intelligent robotic systems, positioning her as a key contributor to the next generation of embodied AI.
Research Focus
Key Achievements
Top Papers
- 1
- 2