LLaVA is a large multimodal model that combines a vision encoder and Vicuna for general-purpose visual and language understanding, achieving impressive chat capabilities mimicking GPT-4 and setting a new state-of-the-art accuracy on Science QA
#multimodal