Research map / Learning theory and algorithms / Prediction and decision losses
Shortcomings of Top-Down Randomization-Based Sanity Checks for Evaluations of Deep Neural Network Explanations
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 Guided Backpropagation or LRP instead of Gradient, Gradient × Input, or Integrated Gradients | Attribution methods evaluated with blurred-patch iterative occlusion · ImageNet validation | occlusion faithfulness (softmax score) | Helped |
| Use noisy or statistically uncorrelated attribution maps | Top-down model-randomization sanity checks · Attribution maps before and after model randomization | SSIM similarity | Helped |
| Randomize upper model layers while keeping the feature extractor fixed | Top-down randomization of ReLU networks · ImageNet examples | Preservation of high-activation contributions | No clear effect |
| Retain skip connections and lower-layer invariances during top-down randomization | ResNet-34 explanations · ImageNet examples | Explanation similarity after randomization | No clear effect |
| Rank explanation methods by top-down randomization scores rather than assessing only whether randomization affects explanations | Attribution methods · Model-randomization sanity checks | Agreement with explanation faithfulness | Hurt |
Closest papers by meaning
- Towards Modeling Uncertainties of Self-Explaining Neural Networks via Conformal Prediction (AAAI 2024 · similarity 0.88)
- How to Probe: Simple Yet Effective Techniques for Improving Post-hoc Explanations (ICLR 2025 · similarity 0.87)
- Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanations (ICML 2025 · similarity 0.87)
- Evaluating Neuron Explanations: A Unified Framework with Sanity Checks (ICML 2025 · similarity 0.86)
- Redundancy Undermines the Trustworthiness of Self-Interpretable GNNs (ICML 2025 · similarity 0.86)
- FaCT: Faithful Concept Traces for Explaining Neural Network Decisions (NeurIPS 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.