A standardized test that evaluates how well models understand and reason about images combined with text.