Negation Blindness in Large Language Models: Unveiling the NO Syndrome in Image Generation