AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Revisiting adapters with adversarial training

Sylvestre-Alvise Rebuffi, Francesco Croce, Sven Gowal

ICLR 2023 · Adversarial robustness and security › Certified and empirical robustness · approach: Robust overfitting and regularization · 0 citations (OpenAlex)

Read the paper

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 clean and adversarial classification-token adapters for co-trainingViT-B16 on ImageNet · ImageNetclean accuracy: 82.64% -> 83.56%Improved
Use clean and adversarial classification-token adapters for co-trainingViT-B16 on ImageNet · ImageNetrobust accuracy: 0% -> 49.87%Improved
Add clean/adversarial-specific trainable scale and offset parameters while sharing batch statistics (ablation)Co-trained normalization layers · ImageNetrobust accuracyImproved
Replace dual BatchNorm statistics with shared statistics but retain domain-specific affine parameters (ablation)Co-trained BatchNorm · ImageNetrobust accuracyNo clear effect
Use only two classification tokens rather than dual LayerNorm scales and offsets (ablation)ViT-B16 co-training · ImageNetrobust accuracy: 49.87% vs 50.02%No clear effect
Replace PGD2 with one-step Fast-AT (ablation)ViT-B16 with classification-token adapter · ImageNetrobust accuracy: 49.87% -> 0.00%Hurt
Replace PGD2 with one-step N-FGSM (ablation)ViT-B16 with classification-token adapter · ImageNetrobust accuracy: 49.87% -> 0.00%Hurt
Increase training attack steps from 2 to 5 (ablation)ViT-B16 with classification-token adapter · ImageNetrobust accuracy: 49.87% -> 50.27%Improved
Set co-training clean-loss weight α=0.4 instead of α=0.5 (ablation)ViT-B16 with classification-token adapter · ImageNetclean accuracy: 83.56% -> 83.76%Improved
Set co-training clean-loss weight α=0.4 instead of α=0.5 (ablation)ViT-B16 with classification-token adapter · ImageNetrobust accuracy: 49.87% -> 52.19%Improved
Interpolate the co-trained clean and adversarial tokens to form a model soupViT-B16 · ImageNet variants, mean accuracymean accuracy: 57.23% -> 61.23%Improved
Form a weight soup from independently nominally and adversarially trained models (ablation)Independent ViT-B16 models · ImageNetclean accuracy: 82.64% -> 28.72% (β=0.1)Hurt

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