Continuous-Action World Models: Learning Cross-Embodiment Spatiotemporal Dynamics Priors
CoAWM (Continuous-Action World Models) learns a shared latent function space for robot actions across embodiments and execution frequencies. It uses two stages: (A) pretraining an action-function encoder/decoder with action reconstruction and consistency across temporal sampling views, and (B) aligning the learned function space to a world-model backbone conditioned on observations, language, and robot state, then decoding actions via time queries at desired execution frequencies.
