AI Paper Atlas

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

Provable Unrestricted Adversarial Training Without Compromise With Generalizability

Lilin Zhang, Ning Yang, Yanchao Sun, Philip S. Yu

TPAMI 2024 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training refinements

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 UAE adversarial training (lambda from 0 up to 10) (ablation)PUAT · CIFAR10natural and robust accuracyImproved
raise UAE weight lambda above 10 (ablation)PUAT · CIFAR10natural and robust accuracyHurt
add unlabeled data during training (ablation)PUAT · CIFAR10natural and robust accuracyImproved
train with UAEs to reduce overfitting on natural samples (ablation)PUAT · CIFAR10test natural accuracy curveImproved
use unlabeled data and generated data (PUAT vs DMAT)PUAT · CIFAR100robust accuracyImproved
PUAT vs RST under GPGD-0.1 and USong on CIFAR100PUAT · CIFAR100robust accuracy: 64.91 vs 64.99Hurt
use PUAT instead of RST on SVHNPUAT · SVHNnatural accuracy: 85.18 -> 92.27Improved
use PUAT instead of TRADES on Tiny ImageNetPUAT · Tiny ImageNetPGD-8/255 robust accuracy: 31.47 -> 55.60Improved

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