F. Sylviana

Semantic Specificity Degradation in Zero-Shot Generative Vision-Language Models

By distinguishing general semantic correctness from exact class identification, this evaluation provides a more detailed view of how VLMs preserve different levels of semantic information.