Abstract: Contrastive Language-Image Pre-training (CLIP) models have shown significant potential, particularly in zero-shot classification across diverse distribution shifts. Building on existing ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results