Vision-Language Models — The ViT-MLP-LLM Pattern