Research map / Learning theory and algorithms / Prediction and decision losses
Robust Attribution Regularization
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 |
|---|---|---|---|
| Train with IG-NORM or IG-SUM-NORM instead of natural training | Image classifiers · MNIST, Fashion-MNIST, GTSRB, Flower | TopK Inter.: 46.61% -> 71.36%; 39.01% -> 59.22%; 54.16% -> 74.81%; 8.12% -> 66.33% | Helped |
| Train with IG-NORM or IG-SUM-NORM instead of natural training | Image classifiers · MNIST, Fashion-MNIST, GTSRB, Flower | Rank Corr.: 0.1758 -> 0.2841; 0.4610 -> 0.6747; 0.6790 -> 0.7684; 0.4978 -> 0.7974 | Helped |
| Use IG-SUM-NORM rather than IG-NORM | Robust attribution regularization · MNIST, Fashion-MNIST, GTSRB, Flower | TopK Inter.: 71.36% -> 72.45%; 59.22% -> 72.08%; 74.81% -> 74.04%; 64.68% -> 66.33% | Mixed |
| Use IG-SUM-NORM rather than IG-NORM | Robust attribution regularization · MNIST, Fashion-MNIST, GTSRB, Flower | Rank Corr.: 0.2841 -> 0.3111; 0.6171 -> 0.6747; 0.7555 -> 0.7684; 0.7591 -> 0.7974 | Mixed |
| Use IG-SUM-NORM rather than Madry et al.'s robust prediction training | Robustly trained image classifiers · MNIST, Fashion-MNIST, GTSRB, Flower | Adv Acc.: 92.47% -> 88.17%; 73.01% -> 70.26%; 83.24% -> 77.12%; 41.91% -> 47.06% | Mixed |
Closest papers by meaning
- Enhanced Regularizers for Attributional Robustness (AAAI 2021 · similarity 0.90)
- Improving Adversarial Robustness of Attribution via Implicit Regularization (ICML 2026 · similarity 0.90)
- Universal generalization guarantees for Wasserstein distributionally robust models (ICLR 2025 · similarity 0.86)
- The Price of Implicit Bias in Adversarially Robust Generalization (NeurIPS 2024 · similarity 0.86)
- Attack-free Evaluating and Enhancing Adversarial Robustness on Categorical Data (ICML 2024 · similarity 0.85)
- Integrated Decision Gradients: Compute Your Attributions Where the Model Makes Its Decision (AAAI 2024 · 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.