InternLM-XComposer-2.5: A Versatile Large Vision Language Model Supporting Long-Contextual Input and Output | Arxiv Papers | Podwise