mlsanityData to 7 Oct 2026

Preemptively Pruning Clever-Hans Strategies in Deep Neural Networks

arXiv 2304.05727 ↗First posted 12 Apr 2023cs.LG cs.AI cs.CVcited by 9

Not on the map. Why?

Abstract (arXiv)

Robustness has become an important consideration in deep learning. With the help of explainable AI, mismatches between an explained model's decision strategy and the user's domain knowledge (e.g. Clever Hans effects) have been identified as a starting point for improving faulty models. However, it is less clear what to do when the user and the explanation agree. In this paper, we demonstrate that acceptance of explanations by the user is not a guarantee for a machine learning model to be robust against Clever Hans effects, which may remain undetected. Such hidden flaws of the model can nevertheless be mitigated, and we demonstrate this by contributing a new method, Explanation-Guided Exposure Minimization (EGEM), that preemptively prunes variations in the ML model that have not been the subject of positive explanation feedback. Experiments demonstrate that our approach leads to models that strongly reduce their reliance on hidden Clever Hans strategies, and consequently achieve higher accuracy on new data.

Key facts. Cited by 9 papers in the map (data through 7 Oct 2026). Builds on 80 of its 137 references that are in the map. First posted on arXiv 12 Apr 2023.

Builds on

80 of 137 references.

References · 80

  1. 2017Attention Is All You Needcited by 59,932
  2. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  3. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  4. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  5. 2020Language Models are Few-Shot Learnerscited by 32,199
  6. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  7. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  8. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 11,979
  9. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  10. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,690
  11. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
  12. 2014Neural Machine Translation by Jointly Learning to Align and Translatecited by 4,530
38 more
  1. 2019DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lightercited by 4,243
  2. 2013Auto-Encoding Variational Bayescited by 3,980
  3. 2021On the Opportunities and Risks of Foundation Modelscited by 3,961
  4. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  5. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  6. 2020Shortcut Learning in Deep Neural Networkscited by 1,928
  7. 2019Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AIcited by 1,543
  8. 2017Learning Important Features Through Propagating Activation Differencescited by 1,454
  9. 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,309
  10. 2018Sanity Checks for Saliency Mapscited by 1,149
  11. 2016The Mythos of Model Interpretabilitycited by 716
  12. 2020GANSpace: Discovering Interpretable GAN Controlscited by 612
  13. 2017Skin Lesion Analysis Toward Melanoma Detection: A Challenge at the 2017 International Symposium on Biomedical Imaging (ISBI), Hosted by the International Skin Imaging Collaboration (ISIC)cited by 549
  14. 2020A Survey of Deep Active Learningcited by 547
  15. 2021Just Train Twice: Improving Group Robustness without Training Group Informationcited by 522
  16. 2020Closed-Form Factorization of Latent Semantics in GANscited by 407
  17. 2022Last Layer Re-Training is Sufficient for Robustness to Spurious Correlationscited by 379
  18. 2019Unmasking Clever Hans Predictors and Assessing What Machines Really Learncited by 359
  19. 2020An Investigation of Why Overparameterization Exacerbates Spurious Correlationscited by 310
  20. 2020Environment Inference for Invariant Learningcited by 306
  21. 2019Interpretable machine learning: definitions, methods, and applicationscited by 303
  22. 2019The Non-IID Data Quagmire of Decentralized Machine Learningcited by 303
  23. 2017Right for the Right Reasons: Training Differentiable Models by Constraining their Explanationscited by 293
  24. 2020A Survey on Neural Network Interpretabilitycited by 291
  25. 2019BCN20000: Dermoscopic Lesions in the Wildcited by 224
  26. 2019Explanations can be manipulated and geometry is to blamecited by 215
  27. 2022Surgical Fine-Tuning Improves Adaptation to Distribution Shiftscited by 205
  28. 2021Simple data balancing achieves competitive worst-group-accuracycited by 177
  29. 2020Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scanscited by 158
  30. 2020BREEDS: Benchmarks for Subpopulation Shiftcited by 155
  31. 2019Interpretations are useful: penalizing explanations to align neural networks with prior knowledgecited by 144
  32. 2020Making deep neural networks right for the right scientific reasons by interacting with their explanationscited by 141
  33. 2020Higher-Order Explanations of Graph Neural Networks via Relevant Walkscited by 138
  34. 2018Explaining Explanations in AIcited by 136
  35. 2020Explainable Deep One-Class Classificationcited by 126
  36. 2020Survey of Deep Reinforcement Learning for Motion Planning of Autonomous Vehiclescited by 101
  37. 2022XAI for Transformers: Better Explanations through Conservative Propagationcited by 100
  38. 2019Pruning by Explaining: A Novel Criterion for Deep Neural Network Pruningcited by 89

50 most cited of 80 All 80 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 4; 2025: 2; 2026: 3.

9 papers.

  1. 2024The Clever Hans Effect in Unsupervised Learningcited by 24
  2. 2025Towards Desiderata-Driven Design of Visual Counterfactual Explainerscited by 7
  3. 2025Mitigating Clever Hans Strategies in Image Classifiers through Generating Counterexamplescited by 6
  4. 2024Reactive Model Correction: Mitigating Harm to Task-Relevant Features via Conditional Bias Suppressioncited by 5
  5. 2026Visual Disentangled Diffusion Autoencoders: Scalable Counterfactual Generation for Foundation Modelscited by 4
  6. 2024XpertAI: uncovering regression model strategies for sub-manifoldscited by 0
  7. 2024Automated Trustworthiness Oracle Generation for Machine Learning Text Classifierscited by 0
  8. 2026Symb-xMIL: Symbolic Explanations for Multiple Instance Learning in Digital Pathologycited by 0
  9. 2026Towards Fast and Disentangled Counterfactuals for Visual Foundation Modelscited by 0