Research map / Learning theory and algorithms / Prediction and decision losses
Towards Better Understanding Attribution Methods
Check 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 |
|---|---|---|---|
| Smooth IntGrad attributions at the input layer with a Gaussian kernel (ablation) | IntGrad · GridPG, VGG11 | Spearman rank correlation with Grad-CAM at the final layer: 0.34 -> 0.80 | Helped |
| Smooth IntGrad attributions with a larger Gaussian kernel (ablation) | IntGrad · GridPG | localization | Helped |
| Smooth IxG attributions with a larger Gaussian kernel (ablation) | IxG · GridPG, VGG11 | localization | Helped |
| Smooth IxG attributions with a Gaussian kernel (ablation) | IxG on Resnet18 · GridPG | localization | No clear effect |
| Use Gaussian smoothing rather than SmoothGrad (ablation) | IntGrad and IxG · GridPG | localization | Helped |
| Apply Gaussian smoothing, which can spread attribution into neighboring grid cells (ablation) | IntGrad and IxG · DiFull | localization | Hurt |
| Evaluate attribution methods at the final layer rather than the input layer | Attribution methods · GridPG | correlation between methods | Helped |
| Evaluate on 3×3 grids rather than 2×2 grids | Attribution methods · CIFAR10 GridPG | localization | Hurt |
Closest papers by meaning
- Benchmarking the Attribution Quality of Vision Models (NeurIPS 2024 · similarity 0.92)
- How to Probe: Simple Yet Effective Techniques for Improving Post-hoc Explanations (ICLR 2025 · similarity 0.89)
- Pixel-level Certified Explanations via Randomized Smoothing (ICML 2025 · similarity 0.88)
- One Wave To Explain Them All: A Unifying Perspective On Feature Attribution (ICML 2025 · similarity 0.87)
- LibraGrad: Balancing Gradient Flow for Universally Better Vision Transformer Attributions (CVPR 2025 · similarity 0.87)
- Iterative Search Attribution for Deep Neural Networks (ICML 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.