Most recently, a group of researchers have announced MiniGPT-4-an open-sourced model performing complex vision-language tasks like GPT-4.