Data Determines Distributional Robustness in Contrastive Language Image Pre-training (CLIP)
This work addresses the problem of understanding robustness in AI models for researchers and practitioners, but it is incremental as it builds on existing CLIP research.
The study investigated the causes of robustness gains in contrastive language-image models like CLIP, finding that a more diverse training distribution is the primary factor, with other elements contributing little to no robustness.
Contrastively trained language-image models such as CLIP, ALIGN, and BASIC have demonstrated unprecedented robustness to multiple challenging natural distribution shifts. Since these language-image models differ from previous training approaches in several ways, an important question is what causes the large robustness gains. We answer this question via a systematic experimental investigation. Concretely, we study five different possible causes for the robustness gains: (i) the training set size, (ii) the training distribution, (iii) language supervision at training time, (iv) language supervision at test time, and (v) the contrastive loss function. Our experiments show that the more diverse training distribution is the main cause for the robustness gains, with the other factors contributing little to no robustness. Beyond our experimental results, we also introduce ImageNet-Captions, a version of ImageNet with original text annotations from Flickr, to enable further controlled experiments of language-image training.