24
6月
Teng Wang et al. Abstract: Existing vision-language pre-training (VLP) methods primarily rely on pa…