The ability to query visual data effectively is crucial for intelligence operations, yet current Vision-Language Models (VLMs) struggle with complex queries such as identifying ‘potential ambush ...