Abstract: Multimodal models that fuse diverse data sources, such as images and text, are pivotal for advancing toward artificial general intelligence. However, their deployment on resource-constrained ...