Research map / Adversarial robustness and security / Certified and empirical robustness
Robust Explanation for Free or At the Cost of Faithfulness
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| robust training of the model (instead of non-robust) | Gradient, GI, IG, SmoothGrad, LIME, SHAP explanations · CIFAR10, ResNet18 and other backbones | local Lipschitz of explanation (lower = more robust) | Helped |
| increase SmoothGrad noise level sigma (smoothing the model with Gaussian kernel) (ablation) | SmoothGrad on non-robust models · CIFAR10, four models | local Lipschitz of explanation (lower = more robust) | Helped |
| increase SmoothGrad noise level sigma (ablation) | SmoothGrad on non-robust models · CIFAR10, five models | faithfulness (similarity to marginal contributions) | Mixed |
| across explanation methods, more robust methods (e.g., LIME, SHAP) tend to be less faithful | six explanation methods on five non-robust models · CIFAR10 | faithfulness vs local Lipschitz | Hurt |
Closest papers by meaning
- Evaluations and Methods for Explanation through Robustness Analysis (ICLR 2021 · similarity 0.91)
- Promoting Counterfactual Robustness through Diversity (AAAI 2024 · similarity 0.88)
- Trust Regions for Explanations via Black-Box Probabilistic Certification (ICML 2024 · similarity 0.86)
- Improving Perturbation-based Explanations by Understanding the Role of Uncertainty Calibration (NeurIPS 2025 · similarity 0.86)
- Training for Stable Explanation for Free (NeurIPS 2024 · similarity 0.86)
- ExpProof : Operationalizing Explanations for Confidential Models with ZKPs (ICML 2025 · similarity 0.85)
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.