Connect text, images, documents, and video frames in shared model architectures. Build visual question-answering and document-understanding tasks while addressing modality alignment and visual hallucination.
