Formal Framework for Causal World Models (CWMs)

Causal World Models must move beyond generative density estimation to capture entity properties and structural interactions. This framework connects causal representation learning and structural causal models to model-based decision-making to enable agents that explain environment dynamics.