AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Are Vision Transformers Robust to Spurious Correlations?

Soumya Suvra Ghosal, Yixuan Li

IJCV 2023 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 35 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use ViT-B/16 instead of BiT-M-R50x3 (both ImageNet-21k)ViT-B/16 · Waterbirdsworst-group accuracy (%): 80.51 -> 89.30Improved
use ViT-B/16 instead of BiT-M-R50x3 (both ImageNet-21k)ViT-B/16 · CelebAworst-group accuracy (%): 89.80 -> 94.10Improved
pre-train ViT-B on ImageNet-21k instead of ImageNet-1k (DeiT-B) (ablation)ViT-B/16 · Waterbirdsworst-group accuracy (%): 44.6 -> 89.3Improved
pre-train BiT-R50x3 on ImageNet-21k instead of ImageNet-1k (ablation)BiT-R50x3 · Waterbirdsworst-group accuracy (%): 60.3 -> 80.5Improved
pre-train ViT-B on ImageNet-21k instead of ImageNet-1k (ablation)ViT-B/16 · CelebAworst-group accuracy (%): 88.0 -> 94.0Improved
pre-train BiT-M-R50x3 on ImageNet-21k instead of ImageNet-1k (ablation)BiT-R50x3 · CelebAworst-group accuracy (%): 88.3 -> 89.8Improved
use DeiT-B (ImageNet-1k transformer) instead of BiT-S-R50x3 CNNDeiT-B/16 · Waterbirdsworst-group accuracy (%): 60.3 -> 44.6Hurt
scale model from ViT-Ti/16 to ViT-B/16 (ablation)ViT · Waterbirdsworst-group accuracy (%): 71.65 -> 89.30Improved
use ViT-B/16 instead of BiT-M-R50x3 for spurious OOD detectionViT-B/16 · Waterbirds OOD (energy score)AUROC: 59.0 -> 91.0Improved
restrict self-attention to local neighbourhood at inference (masked attention) (ablation)ViT-B/16 · Waterbirds and CelebAworst-group accuracy (%)Hurt
remove minority samples from fine-tuning data (data imbalance)ViT-S/16 and BiT-M-R50x1 · Waterbirdsworst-group accuracy (%)Hurt
fine-tune ViT longer (ablation)ViT-S/16 · Waterbirdsminority accuracyImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy