mlsanityData to 7 Oct 2026

Stability Guarantees for Feature Attributions with Multiplicative Smoothing

arXiv 2307.05902 ↗First posted 12 Jul 2023cs.LG cs.AIcited by 14

Not on the map. Why?

Abstract (arXiv)

Explanation methods for machine learning models tend not to provide any formal guarantees and may not reflect the underlying decision-making process. In this work, we analyze stability as a property for reliable feature attribution methods. We prove that relaxed variants of stability are guaranteed if the model is sufficiently Lipschitz with respect to the masking of features. We develop a smoothing method called Multiplicative Smoothing (MuS) to achieve such a model. We show that MuS overcomes the theoretical limitations of standard smoothing techniques and can be integrated with any classifier and feature attribution method. We evaluate MuS on vision and language models with various feature attribution methods, such as LIME and SHAP, and demonstrate that MuS endows feature attributions with non-trivial stability guarantees.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 54 of its 64 references that are in the map. First posted on arXiv 12 Jul 2023.

Builds on

54 of 64 references.

References · 54

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  4. 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,266
  5. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  6. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,690
  7. 2017A Unified Approach to Interpreting Model Predictionscited by 6,488
  8. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
  9. 2021On the Opportunities and Risks of Foundation Modelscited by 3,961
  10. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  11. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
  12. 2013Visualizing and Understanding Convolutional Networkscited by 2,278
38 more
  1. 2017Understanding Black-box Predictions via Influence Functionscited by 1,714
  2. 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,324
  3. 2018Sanity Checks for Saliency Mapscited by 1,149
  4. 2017SmoothGrad: removing noise by adding noisecited by 1,097
  5. 2017Counterfactual Explanations without Opening the Black Box: Automated Decisions and the GDPRcited by 1,086
  6. 2016Adversarial Machine Learning at Scalecited by 934
  7. 2017ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Modelscited by 656
  8. 2020Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?cited by 590
  9. 2019ERASER: A Benchmark to Evaluate Rationalized NLP Modelscited by 523
  10. 2018A Benchmark for Interpretability Methods in Deep Neural Networkscited by 467
  11. 2017Interpretation of Neural Networks is Fragilecited by 443
  12. 2020TweetEval: Unified Benchmark and Comparative Evaluation for Tweet Classificationcited by 414
  13. 2019A Survey on Explainable Artificial Intelligence (XAI): Towards Medical XAIcited by 403
  14. 2019The many Shapley values for model explanationcited by 341
  15. 2017The (Un)reliability of saliency methodscited by 331
  16. 2022From Anecdotal Evidence to Quantitative Evaluation Methods: A Systematic Review on Evaluating Explainable AIcited by 299
  17. 2023TRAK: Attributing Model Behavior at Scalecited by 273
  18. 2018Learning to Explain: An Information-Theoretic Perspective on Model Interpretationcited by 248
  19. 2020A Survey on the Explainability of Supervised Machine Learningcited by 225
  20. 2022Datamodels: Predicting Predictions from Training Datacited by 186
  21. 2020Randomized Smoothing of All Shapes and Sizescited by 146
  22. 2021TorchXRayVision: A library of chest X-ray datasets and modelscited by 131
  23. 2022Towards Faithful Model Explanation in NLP: A Surveycited by 129
  24. 2020Reliable Post hoc Explanations: Modeling Uncertainty in Explainabilitycited by 125
  25. 2021Interpretable Deep Learning: Interpretation, Interpretability, Trustworthiness, and Beyondcited by 116
  26. 2021Counterfactual Explanations Can Be Manipulatedcited by 113
  27. 2022A Consistent and Efficient Evaluation Strategy for Attribution Methodscited by 100
  28. 2021Do Feature Attribution Methods Correctly Attribute Features?cited by 98
  29. 2019Benchmarking Attribution Methods with Relative Feature Importancecited by 89
  30. 2022Impossibility Theorems for Feature Attributioncited by 78
  31. 2022Which Explanation Should I Choose? A Function Approximation Perspective to Characterizing Post Hoc Explanationscited by 77
  32. 2021"Will You Find These Shortcuts?" A Protocol for Evaluating the Faithfulness of Input Salience Methods for Text Classificationcited by 66
  33. 2022A Rigorous Study of Integrated Gradients Method and Extensions to Internal Neuron Attributionscited by 64
  34. 2022Post hoc Explanations may be Ineffective for Detecting Unknown Spurious Correlationcited by 59
  35. 2021Certified Patch Robustness via Smoothed Vision Transformerscited by 44
  36. 2022Towards Formal XAI: Formally Approximate Minimal Explanations of Neural Networkscited by 42
  37. 2020Proper Network Interpretability Helps Adversarial Robustness in Classificationcited by 41
  38. 2021Improved, Deterministic Smoothing for L_1 Certified Robustnesscited by 39

50 most cited of 54 All 54 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 2; 2024: 1; 2025: 5; 2026: 6.

14 papers.

Sort:

By area:

  1. 2023SmoothLLM: Defending Large Language Models Against Jailbreaking Attackscited by 434
  2. 2025Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanationscited by 13
  3. 2026Formal Mechanistic Interpretability: Automated Circuit Discovery with Provable Guaranteescited by 10
  4. 2025Probabilistic Stability Guarantees for Feature Attributionscited by 9
  5. 2025SHAP Meets Tensor Networks: Provably Tractable Explanations with Parallelismcited by 8
  6. 2023Sum-of-Parts: Self-Attributing Neural Networks with End-to-End Learning of Feature Groupscited by 6
  7. 2026Interpretability Can Be Actionablecited by 5
  8. 2024Enhancing Model Interpretability with Local Attribution over Global Explorationcited by 4
  9. 2025Additive Models Explained: A Computational Complexity Approachcited by 3
  10. 2026Provably Explaining Neural Additive Modelscited by 1
  11. 2026Unifying Formal Explanations: A Complexity-Theoretic Perspectivecited by 1
  12. 2025When LRP Diverges from Leave-One-Out in Transformerscited by 0
2 more
  1. 2026Missingness Bias Calibration in Feature Attribution Explanationscited by 0
  2. 2026Measuring Explainer Stability via Attribution Separabilitycited by 0