Believing without Seeing: Quality Scores for Contextualizing Vision-Language Model Explanations - Databubble