[144] Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyondmultilingual alibaba 2023Q3 MLLM qwen