The benchmark is designed to evaluate whether Multimodal Large Language Models (MLLMs) can process multi-UAV collaborative visual data for question answering, covering perception, reasoning, and ...