mlsanityData to 7 Oct 2026

Rethinking Stability for Attribution-based Explanations

arXiv 2203.06877 ↗First posted 14 Mar 2022cs.LGcited by 52

Not on the map. Why?

Abstract (arXiv)

As attribution-based explanation methods are increasingly used to establish model trustworthiness in high-stakes situations, it is critical to ensure that these explanations are stable, e.g., robust to infinitesimal perturbations to an input. However, previous works have shown that state-of-the-art explanation methods generate unstable explanations. Here, we introduce metrics to quantify the stability of an explanation and show that several popular explanation methods are unstable. In particular, we propose new Relative Stability metrics that measure the change in output explanation with respect to change in input, model representation, or output of the underlying predictor. Finally, our experimental evaluation with three real-world datasets demonstrates interesting insights for seven explanation methods and different stability metrics.

Key facts. Cited by 52 papers in the map (data through 7 Oct 2026). Builds on 15 of its 22 references that are in the map. First posted on arXiv 14 Mar 2022.

Builds on

15 of 22 references.

References · 15

  1. 2017A Unified Approach to Interpreting Model Predictionscited by 6,488
  2. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
  3. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  4. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
  5. 2017Learning Important Features Through Propagating Activation Differencescited by 1,454
  6. 2018Sanity Checks for Saliency Mapscited by 1,149
  7. 2017SmoothGrad: removing noise by adding noisecited by 1,097
  8. 2017Interpretation of Neural Networks is Fragilecited by 443
  9. 2018Regularisation of Neural Networks by Enforcing Lipschitz Continuitycited by 311
  10. 2018On the Robustness of Interpretability Methodscited by 310
  11. 2019Explanations can be manipulated and geometry is to blamecited by 215
  12. 2021Towards a Unified Framework for Fair and Stable Graph Representation Learningcited by 128
3 more
  1. 2020Reliable Post hoc Explanations: Modeling Uncertainty in Explainabilitycited by 125
  2. 2021Probing GNN Explainers: A Rigorous Theoretical and Empirical Analysis of GNN Explanation Methodscited by 36
  3. 2021From Human Explanation to Model Interpretability: A Framework Based on Weight of Evidencecited by 17

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 7; 2023: 11; 2024: 15; 2025: 11; 2026: 8.

52 papers.

Sort:

By area:

  1. 2022OpenXAI: Towards a Transparent Evaluation of Model Explanationscited by 125
  2. 2022TalkToModel: Explaining Machine Learning Models with Interactive Natural Language Conversationscited by 65
  3. 2023The Meta-Evaluation Problem in Explainable AI: Identifying Reliable Estimators with MetaQuantuscited by 37
  4. 2022What Makes a Good Explanation?: A Harmonized View of Properties of Explanationscited by 21
  5. 2022Use-Case-Grounded Simulations for Explanation Evaluationcited by 19
  6. 2024Navigating the Maze of Explainable AI: A Systematic Approach to Evaluating Methods and Metricscited by 15
  7. 2023Efficient XAI Techniques: A Taxonomic Surveycited by 14
  8. 2024Sanity Checks Revisited: An Exploration to Repair the Model Parameter Randomisation Testcited by 9
  9. 2024A Fresh Look at Sanity Checks for Saliency Mapscited by 8
  10. 2024Benchmarking XAI Explanations with Human-Aligned Evaluationscited by 8
  11. 2022SoK: Modeling Explainability in Security Analytics for Interpretability, Trustworthiness, and Usabilitycited by 7
  12. 2022Analyzing Explainer Robustness via Probabilistic Lipschitzness of Prediction Functionscited by 6
38 more
  1. 2023Confident Feature Rankingcited by 6
  2. 2024From Movements to Metrics: Evaluating Explainable AI Methods in Skeleton-Based Human Activity Recognitioncited by 6
  3. 2024Provably Better Explanations with Optimized Aggregation of Feature Attributionscited by 6
  4. 2024FLEXtime: Filterbank learning to explain time seriescited by 6
  5. 2025Bridging the Gap in XAI-Why Reliable Metrics Matter for Explainability and Compliancecited by 6
  6. 2025Rethinking Explainability in the Era of Multimodal AIcited by 6
  7. 2025GrAInS: Gradient-based Attribution for Inference-Time Steering of LLMs and VLMscited by 6
  8. 2022On the Robustness of Explanations of Deep Neural Network Models: A Surveycited by 5
  9. 2023The Blame Problem in Evaluating Local Explanations, and How to Tackle itcited by 5
  10. 2024Revisiting the robustness of post-hoc interpretability methodscited by 5
  11. 2023A Survey of Explainable AI in Deep Visual Modeling: Methods and Metricscited by 4
  12. 2024Expected Grad-CAM: Towards gradient faithfulnesscited by 4
  13. 2023Saliency strikes back: How filtering out high frequencies improves white-box explanationscited by 3
  14. 2025On the Effectiveness of Methods and Metrics for Explainable AI in Remote Sensing Image Scene Classificationcited by 3
  15. 2023On the stability, correctness and plausibility of visual explanation methods based on feature importancecited by 2
  16. 2024Explaining Human Activity Recognition with SHAP: Validating Insights with Perturbation and Quantitative Measurescited by 2
  17. 2023Uncertainty Quantification for Local Model Explanations Without Model Accesscited by 1
  18. 2023Feature Importance Disparities for Data Bias Investigationscited by 1
  19. 2023Understanding cirrus clouds using explainable machine learningcited by 1
  20. 2023Rectifying Group Irregularities in Explanations for Distribution Shiftcited by 1
  21. 2024T-Explainer: A Model-Agnostic Explainability Framework Based on Gradientscited by 1
  22. 2024Evaluating Explainable AI Methods in Deep Learning Models for Early Detection of Cerebral Palsycited by 1
  23. 2024Meta-evaluating stability measures: MAX-Senstivity & AVG-Sensitivitycited by 1
  24. 2025Towards Biomarker Discovery for Early Cerebral Palsy Detection: Evaluating Explanations Through Kinematic Perturbationscited by 1
  25. 2025Towards an Evaluation Framework for Explainable Artificial Intelligence Systems for Health and Well-beingcited by 1
  26. 2025VARSHAP: Addressing Global Dependency Problems in Explainable AI with Variance-Based Local Feature Attributioncited by 1
  27. 2025Back to the Baseline: Examining Baseline Effects on Explainability Metricscited by 1
  28. 2026Exploring SAIG Methods for an Objective Evaluation of XAIcited by 1
  29. 2026CIRCUS: Circuit Consensus under Uncertainty via Stability Ensemblescited by 1
  30. 2024Stability of Explainable Recommendationcited by 0
  31. 2024Choose Your Explanation: A Comparison of SHAP and GradCAM in Human Activity Recognitioncited by 0
  32. 2025Value bounds and Convergence Analysis for Averages of LRP attributionscited by 0
  33. 2025Activation-Deactivation: A General Framework for Robust Post-hoc Explainable AIcited by 0
  34. 2025Rethinking Robustness: A New Approach to Evaluating Feature Attribution Methodscited by 0
  35. 2026Training for Trustworthy Saliency Maps: Adversarial Training Meets Feature-Map Smoothingcited by 0
  36. 2026No Single Metric Tells the Whole Story: A Multi-Dimensional Evaluation Framework for Uncertainty Attributionscited by 0
  37. 2026Foundation models for discovering robust biomarkers of neurological disorders from dynamic functional connectivitycited by 0
  38. 2026Measuring Explainer Stability via Attribution Separabilitycited by 0

50 most cited of 52 All 52 (JSON)