Cross-modal Alignment

The ability of a model to establish semantic correspondences between inputs of different modalities, such as audio events and visual content. It is a core technical challenge in omni-modal video understanding.