Description

Adept AI introduced Fuyu-8B, a smaller version of their multimodal model that powers their product. This base model, with a decoder-only multi-modal transformer architecture, has a simpler design and faster response time while providing satisfactory performance on standard image understanding benchmarks like visual question-answering and natural-image-captioning. The model supports arbitrary image resolutions and can answer questions about graphs, diagrams, UI-based queries, and screen images with high precision. It is designed for digital agents but needs fine-tuning to cater to specific use cases such as verbose captioning or multimodal chat.