Physically Grounded Multimodal World Models for Embodied Robot Interaction via 3D Representation Alignment. (2026). Journal of Advanced Artificial Intelligence Research, 1(1). https://www.jaair.org/index.php/home/article/view/170