Multimodal LLMs Without Compromise: Researchers from UCLA, UW–Madison, and Adobe Introduce X-Fusion to Add Vision to Frozen Language Models Without Losing Language Capabilities