AI Paper Atlas

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

Revisiting Adversarial Robustness Distillation from the Perspective of Robust Fairness

Xinli Yue, Ningping Mou, Qian Wang, Lingchen Zhao

NeurIPS 2023 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training with feature consistency · 4 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
add class-wise re-weighting by LPS difficulty (Fair-ARD)ARD, ResNet-18 · CIFAR-10AA worst-class robustness: 16.90 -> 29.20Improved
add class-wise re-weighting (Fair-IAD)IAD, ResNet-18 · CIFAR-10AA worst-class robustness: 18.60 -> 25.60Improved
add class-wise re-weighting (Fair-RSLAD)RSLAD, ResNet-18 · CIFAR-10AA worst-class robustness: 20.90 -> 25.10Improved
add class-wise re-weighting (Fair-MTARD)MTARD, ResNet-18 · CIFAR-10AA worst-class robustness: 16.10 -> 22.60Improved
add class-wise re-weighting (Fair-MTARD)MTARD, ResNet-18 · CIFAR-10clean average accuracy: 87.14 -> 81.98Hurt
add class-wise re-weighting (Fair-ARD)ARD, ResNet-18 · CIFAR-10AA average robustness: 49.05 -> 49.13No clear effect
add class-wise re-weighting (Fair-ARD)ARD, MobileNetV2 · CIFAR-10AA worst-class robustness: 18.60 -> 24.00Improved
add class-wise re-weighting (Fair-RSLAD)RSLAD, MobileNetV2 · CIFAR-10AA worst-class robustness: 19.60 -> 26.30Improved
add example-wise GAIRAT re-weighting instead of class-wise (GAIR-ARD)ARD, ResNet-18 · CIFAR-10AA average robustness: 49.05 -> 40.88Hurt
add FRL re-weighting instead of LPS (FRL-ARD)ARD, ResNet-18 · CIFAR-10AA worst-class robustness: 16.90 -> 23.90Improved
add FAT fairness training instead of LPS (FAT-ARD)ARD, ResNet-18 · CIFAR-10AA worst-class robustness: 16.90 -> 25.90Improved
add CFA re-weighting instead of LPS (CFA-ARD)ARD, ResNet-18 · CIFAR-10AA worst-class robustness: 16.90 -> 20.00Improved

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