Abstract: The pretrained vision-language models (VLMs) have achieved outstanding performance in various visual tasks, primarily due to the knowledge they have acquired from massive image-text pairs.

Abstract: Classification of incomplete data remains a challenging task since the distribution of training and test sets may be inconsistently caused by missing values. To address such a problem, this ...