mlsanityData to 14 Oct 2026

Trade-off Between Efficiency and Consistency for Removal-based Explanations

arXiv 2210.17426 ↗First posted 31 Oct 2022cs.LG cs.AI cs.CVcited by 6

Not on the map. Why?

Abstract (arXiv)

In the current landscape of explanation methodologies, most predominant approaches, such as SHAP and LIME, employ removal-based techniques to evaluate the impact of individual features by simulating various scenarios with specific features omitted. Nonetheless, these methods primarily emphasize efficiency in the original context, often resulting in general inconsistencies. In this paper, we demonstrate that such inconsistency is an inherent aspect of these approaches by establishing the Impossible Trinity Theorem, which posits that interpretability, efficiency, and consistency cannot hold simultaneously. Recognizing that the attainment of an ideal explanation remains elusive, we propose the utilization of interpretation error as a metric to gauge inefficiencies and inconsistencies. To this end, we present two novel algorithms founded on the standard polynomial basis, aimed at minimizing interpretation error. Our empirical findings indicate that the proposed methods achieve a substantial reduction in interpretation error, up to 31.8 times lower when compared to alternative techniques. Code is available at https://github.com/trusty-ai/efficient-consistent-explanations.

Key facts. Cited by 6 papers in the map (data through 14 Oct 2026). Builds on 45 of its 85 references that are in the map. First posted on arXiv 31 Oct 2022.

Builds on

45 of 85 references.

References · 45

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  4. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,968
  5. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  6. 2017A Unified Approach to Interpreting Model Predictionscited by 6,555
  7. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  8. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,070
  9. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  10. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
  11. 2017Learning Important Features Through Propagating Activation Differencescited by 1,458
  12. 2018Sanity Checks for Saliency Mapscited by 1,156
33 more
  1. 2017SmoothGrad: removing noise by adding noisecited by 1,109
  2. 2020Transformer Interpretability Beyond Attention Visualizationcited by 627
  3. 2017Interpretable Explanations of Black Boxes by Meaningful Perturbationcited by 612
  4. 2019The many Shapley values for model explanationcited by 344
  5. 2016Not Just a Black Box: Learning Important Features Through Propagating Activation Differencescited by 297
  6. 2018Consistent Individualized Feature Attribution for Tree Ensemblescited by 285
  7. 2021On Explainability of Graph Neural Networks via Subgraph Explorationscited by 281
  8. 2020Explaining by Removing: A Unified Framework for Model Explanationcited by 218
  9. 2019Explaining individual predictions when features are dependent: More accurate approximations to Shapley valuescited by 208
  10. 2020Improving KernelSHAP: Practical Shapley Value Estimation via Linear Regressioncited by 154
  11. 2019The Shapley Taylor Interaction Indexcited by 143
  12. 2021FastSHAP: Real-Time Shapley Value Estimationcited by 125
  13. 2019Explaining Deep Neural Networks with a Polynomial Time Algorithm for Shapley Values Approximationcited by 115
  14. 2019Asymmetric Shapley values: incorporating causal knowledge into model-agnostic explainabilitycited by 114
  15. 2020Explaining Explanations: Axiomatic Feature Interactions for Deep Networkscited by 103
  16. 2018L-Shapley and C-Shapley: Efficient Model Interpretation for Structured Datacited by 100
  17. 2019CXPlain: Causal Explanations for Model Interpretation under Uncertaintycited by 98
  18. 2020How does this interaction affect me? Interpretable attribution for feature interactionscited by 73
  19. 2017What Does Explainable AI Really Mean? A New Conceptualization of Perspectivescited by 72
  20. 2018Deep Neural Decision Treescited by 72
  21. 2020Attribution in Scale and Spacecited by 61
  22. 2020Shapley Flow: A Graph-based Approach to Interpreting Model Predictionscited by 51
  23. 2021Shapley Explanation Networkscited by 41
  24. 2017Interpretability via Model Extractioncited by 40
  25. 2015Improved SVRG for Non-Strongly-Convex or Sum-of-Non-Convex Objectivescited by 38
  26. 2018An Interpretable Model with Globally Consistent Explanations for Credit Riskcited by 33
  27. 2017Hyperparameter Optimization: A Spectral Approachcited by 24
  28. 2017Neural Decision Treescited by 17
  29. 2015Interpretable Two-level Boolean Rule Learning for Classificationcited by 10
  30. 2013Supersparse Linear Integer Models for Interpretable Classificationcited by 8
  31. 2021Axiomatic Explanations for Visual Search, Retrieval, and Similarity Learningcited by 6
  32. 2014Methods and Models for Interpretable Linear Classificationcited by 4
  33. 2015Or's of And's for Interpretable Classification, with Application to Context-Aware Recommender Systemscited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 1; 2024: 0; 2025: 4; 2026: 1.

6 papers.

  1. 2025ProxySPEX: Inference-Efficient Interpretability via Sparse Feature Interactions in LLMscited by 16
  2. 2025SPEX: Scaling Feature Interaction Explanations for LLMscited by 12
  3. 2025Existing LLMs Are Not Self-Consistent For Simple Taskscited by 5
  4. 2023Deep Tensor Networkcited by 0
  5. 2025Understanding Action Effects through Instrumental Empowerment in Multi-Agent Reinforcement Learningcited by 0
  6. 2026Fundamental Limitation in Explaining AIcited by 0