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

Deep neural networks may be susceptible to learning spurious correlationsthat hold on average but not in atypical test samples. As with the recentemergence of vision transformer (ViT) models, it remains underexplored howspurious correlations are manifested in such architectures. In this paper, wesystematically investigate the robustness of vision transformers to spuriouscorrelations on three challenging benchmark datasets and compare theirperformance with popular CNNs. Our study reveals that when pre-trained on asufficiently large dataset, ViT models are more robust to spurious correlationsthan CNNs. Key to their success is the ability to generalize better from theexamples where spurious correlations do not hold. Further, we perform extensiveablations and experiments to understand the role of the self-attentionmechanism in providing robustness under spuriously correlated environments. Wehope that our work will inspire future research on further understanding therobustness of ViT models.

Are Vision Transformers Robust to Spurious Correlations?