AI Paper Atlas

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

Adversarially Robust Neural Architectures

Minjing Dong, Yanxi Li, Yunhe Wang, Chang Xu

TPAMI 2025 · Adversarial robustness and security › Certified and empirical robustness · approach: Robustness-accuracy trade-off analysis · 66 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 confidence learning of architecture parameters (ablation)RACL · CIFAR-10 PGD-20adversarial accuracy: 54.58 -> 55.68Improved
add confidence learning of architecture parameters (ablation)RACL · CIFAR-10 AutoAttackadversarial accuracy: 48.89 -> 50.07Improved
add confidence learning of architecture parameters (ablation)RACL · CIFAR-10 cleannatural accuracy: 84.35 -> 84.04Hurt
add gradient norm lower-bound constraint (ablation)RACL · CIFAR-10 PGD-20adversarial accuracy: 53.87 -> 55.68Improved
search with RACL instead of random search (ablation)RACL · CIFAR-10 PGD-20adversarial accuracy: 51.36 -> 55.68Improved
use rho=0.001 instead of rho=0.01 (eta=0.9) (ablation)RACL · CIFAR-10 PGD-20adversarial accuracy: 55.66 -> 55.68Improved
use eta=0.9 instead of eta=0.7 (rho=0.001) (ablation)RACL · CIFAR-10 PGD-20adversarial accuracy: 55.00 -> 55.68Improved
replace PC-DARTS cells with RACL cellsPC-DARTS · CIFAR-10 PGD-100adversarial accuracy: 52.36 -> 55.32Improved
replace RobNet-free with RACLRobNet-free · CIFAR-10 PGD-100adversarial accuracy: 52.57 -> 55.32Improved
replace DARTS with RACL on CIFAR-100DARTS · CIFAR-100 AutoAttackadversarial accuracy: 24.30 -> 25.55Improved
replace NASNet with RACL on Tiny-ImageNetNASNet · Tiny-ImageNet CWadversarial accuracy: 41.93 -> 42.99Improved
replace NASNet with RACL under transfer black-box attackNASNet · CIFAR-10 transfer black-box PGD-20adversarial accuracy: 83.21 -> 83.45Improved

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