Subramanian Ramamoorthy

Human decision-making is governed by a bounded trade-off screening process that reveals how context shapes our preferences in multi-attribute choices.

Modeling human actions as executable programs not only improves action recognition, but also unlocks more data-efficient and interpretable models compared to standard monolithic approaches.

Robots can now unfold a variety of cloth types in the real world, thanks to a world model enhanced with surface normals and improved data augmentation.