NeuIDO: Neural Intrinsic Dynamics Operator for Physics-Informed 4D World Models

This paper introduces NeuIDO, a neural intrinsic dynamics operator designed for physics-informed 4D world models. The approach addresses the limitation that existing physics-guided 4D generation relies on manually imposed dynamical assumptions rather than internalized world dynamics. NeuIDO integrates physical constraints directly into the neural operator architecture, enabling more accurate prediction of 3D object interactions and future trajectories. This work contributes to the advancement of embodied intelligence by providing a more principled approach to modeling environmental dynamics.