UNITAS is a 3D-native world action model that unifies observations, actions, and scene dynamics in a shared metric 3D frame. Unlike image-based models, it uses a common representation for both robot embodiments and human hands.
HOW THIS AFFECTS YOU
●
builderYou can use this model to achieve more physically accurate robotic manipulation in 3D space.
●
researcherThis shifts the paradigm from view-dependent visual latents to metric-based spatial reasoning.