mlsanityData to 14 Oct 2026

On the Robustness of Removal-Based Feature Attributions

arXiv 2306.07462 ↗First posted 12 Jun 2023cs.LG stat.MLcited by 19

Not on the map. Why?

Abstract (arXiv)

To explain predictions made by complex machine learning models, many feature attribution methods have been developed that assign importance scores to input features. Some recent work challenges the robustness of these methods by showing that they are sensitive to input and model perturbations, while other work addresses this issue by proposing robust attribution methods. However, previous work on attribution robustness has focused primarily on gradient-based feature attributions, whereas the robustness of removal-based attribution methods is not currently well understood. To bridge this gap, we theoretically characterize the robustness properties of removal-based feature attributions. Specifically, we provide a unified analysis of such methods and derive upper bounds for the difference between intact and perturbed attributions, under settings of both input and model perturbations. Our empirical results on synthetic and real-world data validate our theoretical results and demonstrate their practical implications, including the ability to increase attribution robustness by improving the model's Lipschitz regularity.

Key facts. Cited by 19 papers in the map (data through 14 Oct 2026). Builds on 42 of its 64 references that are in the map. First posted on arXiv 12 Jun 2023.

Builds on

42 of 64 references.

References · 42

  1. 2014Explaining and Harnessing Adversarial Examplescited by 7,364
  2. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  3. 2013Intriguing properties of neural networkscited by 5,155
  4. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,070
  5. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  6. 2021Convex optimizationcited by 2,515
  7. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
  8. 2013Visualizing and Understanding Convolutional Networkscited by 2,290
  9. 2016Adversarial examples in the physical worldcited by 1,769
  10. 2017Learning Important Features Through Propagating Activation Differencescited by 1,458
  11. 2018Sanity Checks for Saliency Mapscited by 1,156
  12. 2018RISE: Randomized Input Sampling for Explanation of Black-box Modelscited by 838
30 more
  1. 2017Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networkscited by 744
  2. 2017Interpretation of Neural Networks is Fragilecited by 444
  3. 2019Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methodscited by 443
  4. 2018Lipschitz regularity of deep neural networks: analysis and efficient estimationcited by 415
  5. 2019The many Shapley values for model explanationcited by 344
  6. 2017The (Un)reliability of saliency methodscited by 333
  7. 2018Regularisation of Neural Networks by Enforcing Lipschitz Continuitycited by 314
  8. 2018On the Robustness of Interpretability Methodscited by 314
  9. 2016Not Just a Black Box: Learning Important Features Through Propagating Activation Differencescited by 297
  10. 2020Explaining by Removing: A Unified Framework for Model Explanationcited by 218
  11. 2019Explanations can be manipulated and geometry is to blamecited by 216
  12. 2019Explainable Machine Learning in Deploymentcited by 211
  13. 2019Explaining individual predictions when features are dependent: More accurate approximations to Shapley valuescited by 208
  14. 2017Spectral Norm Regularization for Improving the Generalizability of Deep Learningcited by 182
  15. 2018Semidefinite relaxations for certifying robustness to adversarial examplescited by 179
  16. 2022Algorithms to estimate Shapley value feature attributionscited by 115
  17. 2019Asymmetric Shapley values: incorporating causal knowledge into model-agnostic explainabilitycited by 114
  18. 2019Fooling Neural Network Interpretations via Adversarial Model Manipulationcited by 112
  19. 2020Shapley explainability on the data manifoldcited by 88
  20. 2020Fairwashing Explanations with Off-Manifold Detergentcited by 72
  21. 2023A Unified Algebraic Perspective on Lipschitz Neural Networkscited by 60
  22. 2021Towards the Unification and Robustness of Perturbation and Gradient Based Explanationscited by 47
  23. 2022Learning to Estimate Shapley Values with Vision Transformerscited by 45
  24. 2020Towards Robust Explanations for Deep Neural Networkscited by 43
  25. 2020Smoothed Geometry for Robust Attributioncited by 38
  26. 2021Probing GNN Explainers: A Rigorous Theoretical and Empirical Analysis of GNN Explanation Methodscited by 36
  27. 2019LS-Tree: Model Interpretation When the Data Are Linguisticcited by 7
  28. 2022Zonotope Domains for Lagrangian Neural Network Verificationcited by 5
  29. 2022Ensembling improves stability and power of feature selection for deep learning modelscited by 3
  30. 2022Shapley variable importance cloud for machine learning modelscited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 8; 2025: 6; 2026: 5.

19 papers.

Sort:

By area:

  1. 2024shapiq: Shapley Interactions for Machine Learningcited by 43
  2. 2024Stochastic Amortization: A Unified Approach to Accelerate Feature and Data Attributioncited by 21
  3. 2025Probabilistic Stability Guarantees for Feature Attributionscited by 9
  4. 2024Provably Better Explanations with Optimized Aggregation of Feature Attributionscited by 6
  5. 2024Efficient and Accurate Explanation Estimation with Distribution Compressioncited by 4
  6. 2025Pixel-level Certified Explanations via Randomized Smoothingcited by 4
  7. 2025Improving Perturbation-based Explanations by Understanding the Role of Uncertainty Calibrationcited by 3
  8. 2024On the Robustness of Global Feature Effect Explanationscited by 2
  9. 2025Attribution Explanations for Deep Neural Networks: A Theoretical Perspectivecited by 2
  10. 2024Assessing biomedical knowledge robustness in large language models by query-efficient sampling attackscited by 1
  11. 2024Error Analysis of Shapley Value-Based Model Explanations: An Informative Perspectivecited by 1
  12. 2025Fast Fourier Correlation is a Highly Efficient and Accurate Feature Attribution Algorithm from the Perspective of Control Theory and Game Theorycited by 1
7 more
  1. 2026Bound Propagation meets Constraint Simplification: Improving Logic-based XAI for Neural Networkscited by 1
  2. 2026Missingness Bias Calibration in Feature Attribution Explanationscited by 1
  3. 2026DePICT: Decision-Preserving Interface for Constrained Downstream Taskscited by 1
  4. 2024Comprehensive and Reliable Feature Attribution for Diverse Modalities and Models via Frequency-Domain Insightscited by 0
  5. 2025Why Uncertainty Calibration Matters for Reliable Perturbation-based Explanationscited by 0
  6. 2026Geometry-Adaptive Explainer for Faithful Dictionary-Based Interpretability under Distribution Shiftcited by 0
  7. 2026Improving Adversarial Robustness of Attribution via Implicit Regularizationcited by 0