<ul><li>Llip, Latent Language Image Pretraining, is introduced to model the diversity of captions that could match an image.</li><li>Llip's vision encoder outputs a set of visual features that are mixed into a final representation by conditioning on information derived from the text.</li><li>Llip outperforms non-contextualized baselines like CLIP and SigLIP on various tasks, including zero-shot classification and retrieval.</li><li>Llip achieves a zero-shot top-1 accuracy of 83.5% on ImageNet, outperforming similarly sized CLIP by 1.4%.</li></ul>

Modeling Caption Diversity in Contrastive Vision-Language Pretraining

Discover more