Training a model to understand and process multiple types of input data (like text and images) together rather than separately.
Quality of vision, audio, and image understanding (distinct from modality support)