Seungryong Kim

ENS achieves up to 10x better accuracy than traditional methods in turbulent flow scenarios by directly utilizing the PDE residual as input for iterative error correction.

Motion-aware correspondences can drastically enhance the geometric consistency of novel-view video generation, outperforming traditional methods.

GAM revolutionizes robot policy learning by seamlessly integrating 3D geometric reasoning, outperforming traditional models in accuracy and efficiency.

SpatialClaw enables agents to dynamically compose and adapt their reasoning strategies, achieving a remarkable 11.2-point accuracy boost over traditional spatial agents.