AI Paper Atlas

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

A3: A Platform of Searching for Robust Neural Architectures and Efficient Adversarial Attacks

Jialiang Sun, Wen Yao, Tingsong Jiang, Chao Li, Xiaoqian Chen

TPAMI 2025 · Adversarial robustness and security › Certified and empirical robustness · approach: Robustness-accuracy trade-off analysis · 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
use DE instead of random search to search the attack (ablation)attack search on Regularization ResNet-18 · CIFAR10 l_infrobust accuracy (%): 53.6 -> 5.00Improved
search attack with DE instead of using manual PGD-7PGDAT ResNet-50 · CIFAR10 l_infrobust accuracy (%): 54.6 -> 50.8Improved
use multi-objective NSGA-II instead of DE for attack search (ablation)PGDAT ResNet-50 · CIFAR10 l_inftime cost (s): 458.9 -> 145.88 time, 50.8 -> 52.0 accuracyMixed
use NSGA-II instead of random search for attack search (ablation)attack search · CIFAR10 l_inf and l_2robust accuracy (%)Improved
use DE instead of random search for NAS (FGSM evaluation) (ablation)NAS for robust architectures · CIFAR10 PGD-LinfAttackrobust accuracy (%): 43.6 -> 45.2Improved
use DE instead of random search for NAS (ablation)NAS search cost · FGSM evaluationsearch time (GPU days): 1.15 -> 12Hurt
use Jacobian instead of FGSM as NAS robustness metric (ablation)DARTS · CIFAR10 PGD-LinfAttackrobust accuracy (%): 38.8 -> 51.2Improved
use Jacobian instead of FGSM as NAS robustness metric (ablation)SmoothDARTS · CIFAR10 PGD-LinfAttackrobust accuracy (%): 53.0 -> 42.2Hurt
use searched attack AAA instead of FGSM as NAS robustness evaluation (ablation)Random search NAS · CIFAR10 AAArobust accuracy (%): 41.6 -> 49.4Improved
use more robust source model (FairDARTS_Jacobian) for attack search (ablation)random search attack on SmoothDARTS_FGSM · CIFAR10robust accuracy (%): 54.0 -> 52.8Improved
use natural-noise evaluation in NAS (DARTS_Natural) vs manual ResNet-18 (ablation)DARTS · CIFAR10-C snowaccuracy (%): 76.56 -> 79.17Improved
use SmoothDARTS with Natural evaluation instead of DE (ablation)NAS · CIFAR10 cleanaccuracy (%): 92.61 -> 89.42Hurt

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