Researchers have proposed a new type of world model for continuous control tasks that can generalize across different physical systems. The Graph-Operator World Models (GraphOp-WM) approach represents the body and its kinematic relations as an attributed graph and factorizes each transition into two parts: a morphology-independent local dynamics basis and a morphology-conditioned structured operator. This allows the model to adapt to changes in parameters such as link lengths, masses, and actuation without retraining from scratch. The authors demonstrate their approach on three MuJoCo environments: Hopper, Walker2d, and HalfCheetah. They use controlled parameter splits to test generalization across interpolation, extrapolation, and held-out compositions of physical parameters. --- Why it matters: This work matters because it addresses a common challenge in continuous control tasks: adapting to changes in the physical system without retraining from scratch. The proposed Graph-Operator World Models approach provides a structured way to factorize transitions into reusable and morphology-dependent components, which can improve the robustness of world models in real-world applications. Source: https://arxiv.org/abs/2608.20936