+ + Dataset: Vision-Language Model Sensitivity to Semantic and Lexical Alterations