Research map / Label-efficient and robust learning / Recognition and long-tail
Robust Models Are More Interpretable Because Attributions Look Normal
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 |
|---|---|---|---|
| Use BIG instead of BSM | ImageNet standard ResNet50 attributions · ImageNet (standard ResNet50) | Loc.: 0.33 -> 0.38 | Helped |
| Use BIG instead of Smoothed Gradient | ImageNet standard ResNet50 attributions · ImageNet (standard ResNet50) | EG: 0.55 -> 0.54 | Hurt |
| Use BIG instead of AGI | ImageNet standard ResNet50 attributions · ImageNet (standard ResNet50) | PP: 0.58 -> 0.87 | Helped |
| Use BIG instead of AGI | ImageNet standard ResNet50 attributions · ImageNet (standard ResNet50) | Con.: 4.01 -> 4.35 | Helped |
| Use BIG instead of AGI | ImageNet robust ResNet50 attributions · ImageNet (robust ResNet50, ℓ₂|3.0) | Loc.: 0.39 -> 0.39 | No clear effect |
| Use BIG instead of IG | ImageNet robust ResNet50 attributions · ImageNet (robust ResNet50, ℓ₂|3.0) | EG: 0.65 -> 0.74 | Helped |
| Use BIG instead of AGI | ImageNet robust ResNet50 attributions · ImageNet (robust ResNet50, ℓ₂|3.0) | PP: 0.88 -> 0.92 | Helped |
| Use BIG instead of IG | ImageNet robust ResNet50 attributions · ImageNet (robust ResNet50, ℓ₂|3.0) | Con.: 4.37 -> 5.03 | Helped |
| Use BIG instead of IG with a global baseline | Standard ResNet50 explanations · ImageNet | baseline sensitivity | Helped |
| Increase robustness through adversarial training | ImageNet ResNet50 attributions · ImageNet | alignment with nearby boundary vectors | Helped |
Closest papers by meaning
- Towards More Robust Interpretation via Local Gradient Alignment (AAAI 2023 · similarity 0.89)
- Hold me tight! Influence of discriminative features on deep network boundaries (NeurIPS 2020 · similarity 0.89)
- Pixel-level Certified Explanations via Randomized Smoothing (ICML 2025 · similarity 0.89)
- Rethinking Robustness of Model Attributions (AAAI 2024 · similarity 0.86)
- One Wave To Explain Them All: A Unifying Perspective On Feature Attribution (ICML 2025 · similarity 0.86)
- Harnessing Edge Information for Improved Robustness in Vision Transformers (AAAI 2024 · similarity 0.86)
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.