AI Paper Atlas

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

SAFARI: Versatile and Efficient Evaluations for Robustness of Interpretability

Wei Huang, Xingyu Zhao, Gaojie Jin, Xiaowei Huang

ICCV 2023 · Adversarial robustness and security › Certified and empirical robustness · approach: Decision-boundary geometry of robustness · 40 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 Genetic Algorithm instead of simple Monte Carlo to find worst-case interpretation discrepancyworst-case evaluation · CIFAR10 (100 seeds)worst-case MSE: 0.589 -> 42.436Improved
use Subset Simulation instead of simple Monte Carlo for rare misinterpretation probabilityprobabilistic evaluation · MNIST F-hatsamples required: 1.13x10^7 -> 15000Improved
use Subset Simulation on a common eventprobabilistic evaluation · CIFAR10 F-hat (ln P=-0.79)samples required: 2500 vs 2500No clear effect
increase MH steps in Subset Simulation (ablation)Subset Simulation · MNIST seedestimation accuracyImproved
increase GA population size beyond 500 (ablation)GA · CelebA seedoptimisation objectiveNo clear effect
add input Hessian norm regularisation to training (ablation)DL model · MNISTworst-case PCC under F-hat: 0.06 -> 0.17Improved
add input gradient norm regularisation to training (ablation)DL model · CIFAR10worst-case classification J: 42.58 -> 42.34No clear effect
add PGD adversarial training (ablation)DL model · CelebAworst-case PCC under F-hat: 0.08 -> 0.55Improved
add PGD adversarial training (ablation)DL model · CIFAR10test accuracy: 0.878 -> 0.703Hurt
compare XAI methods by robustness to F-hatDeepLift and Integrated Gradients · MNISTinterpretation robustnessImproved

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