AI Paper Atlas

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

Improving Adversarial Robustness by Mitigating Instability through Relearning

Yi Zeng, Ling Zhou, Ruilong Yu, Qihe Liu, Shijie Zhou

ECCV 2026 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training refinements · 0 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 relearning (RE) to PGD-AT in place of TEPGD-AT · CIFAR-10 ResNet-18 PGD-20 finalrobust accuracy: 53.87 -> 55.40Improved
add relearning to PGD-ATPGD-AT · CIFAR-10 ResNet-18 Auto Attack finalrobust accuracy: 40.74 -> 50.83Improved
add relearning to PGD-ATPGD-AT · CIFAR-10 ResNet-18RSR: -88.61 -> -22.80Improved
add relearning to PGD-AT on ViTPGD-AT · CIFAR-10 ViT PGD-20 finalrobust accuracy: 38.78 -> 44.55Improved
add relearning to MARTMART · CIFAR-10 ResNet-18 Auto Attackrobust accuracy: 43.61 -> 49.30Improved
add relearning to TRADESTRADES · CIFAR-10 ResNet-18 Auto Attackrobust accuracy: 47.29 -> 49.76Improved
add relearning to AT-AWPAT-AWP · CIFAR-100 ResNet-18 Auto Attackrobust accuracy: 25.30 -> 25.87Improved
add relearning to PGD-ATPGD-AT · CIFAR-100 ResNet-18 Auto Attackrobust accuracy: 19.69 -> 26.00Improved
use relearning loss in adversarial example generation only (ablation)PGD-AT · CIFAR-10 ResNet-18 PGD-20 finalrobust accuracy: 41.40 -> 51.58Improved
use relearning loss in training only (ablation)PGD-AT · CIFAR-10 ResNet-18 PGD-20 finalrobust accuracy: 41.40 -> 49.08Improved
use relearning loss in both generation and training instead of one phase (ablation)PGD-AT+RE · CIFAR-10 ResNet-18 PGD-20 finalrobust accuracy: 51.58 -> 55.40Improved
increase relearning weight omega beyond 30 (ablation)PGD-AT+RE · CIFAR-10 ResNet-18robust accuracyHurt

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