AI Paper Atlas

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

S$^{2}$ 2O: Enhancing Adversarial Training With Second-Order Statistics of Weights

Gaojie Jin, Xinping Yi, Wei Huang, Sven Schewe, Xiaowei Huang

TPAMI 2025 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training refinements · 5 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 S2O to adversarial trainingAT PreAct ResNet-18 · CIFAR-10 l_infPGD-20 accuracy: 52.77 -> 55.11Improved
add S2O to adversarial trainingAT PreAct ResNet-18 · CIFAR-100 l_infPGD-20 accuracy: 28.02 -> 30.58Improved
add S2O to adversarial trainingAT PreAct ResNet-18 · SVHN l_infPGD-20 accuracy: 60.91 -> 64.83Improved
add S2O to adversarial training under l2AT PreAct ResNet-18 · SVHN l_2PGD accuracy: 66.76 -> 76.19Improved
add S2O to adversarial trainingAT PreAct ResNet-18 · CIFAR-10training time per epoch: 309s -> 368sHurt
compute correlation matrices from adversarial data only (ablation)AT+S2O · CIFAR-10 l_inf ResNet-18clean accuracy: 83.65 -> 82.89Hurt
compute correlation matrices from adversarial data only (ablation)AT+S2O · CIFAR-10 l_inf ResNet-18training time per epoch: 368s -> 340sImproved
add S2O to TRADESTRADES · CIFAR-10 ResNet-18AutoAttack accuracy: 48.6 -> 49.5Improved
add S2O to TRADES+AWPTRADES+AWP · CIFAR-10 ResNet-18AutoAttack accuracy: 51.7 -> 52.4Improved
add S2O to TRADES+AWPTRADES+AWP · CIFAR-10 WRNAutoAttack accuracy: 56.2 -> 55.9Hurt
add S2O to TRADESTRADES · CIFAR-100 WRNclean accuracy: 60.38 -> 63.40Improved
add S2O to DDPM-data adversarial trainingPang et al. (2022) · CIFAR-10 WRN-28-10AutoAttack accuracy: 61.04 -> 61.45Improved

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