English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Contrastively trained language-image models such as CLIP, ALIGN, and BASIChave demonstrated unprecedented robustness to multiple challenging naturaldistribution shifts. Since these language-image models differ from previoustraining approaches in several ways, an important question is what causes thelarge robustness gains. We answer this question via a systematic experimentalinvestigation. Concretely, we study five different possible causes for therobustness gains: (i) the training set size, (ii) the training distribution,(iii) language supervision at training time, (iv) language supervision at testtime, and (v) the contrastive loss function. Our experiments show that the morediverse training distribution is the main cause for the robustness gains, withthe other factors contributing little to no robustness. Beyond our experimentalresults, we also introduce ImageNet-Captions, a version of ImageNet withoriginal text annotations from Flickr, to enable further controlled experimentsof language-image training.

Data Determines Distributional Robustness in Contrastive Language Image Pre-training (CLIP)