mlsanityData to 7 Oct 2026

Monitoring Shortcut Learning using Mutual Information

arXiv 2206.13034 ↗First posted 27 Jun 2022cs.LG cs.AIcited by 6

Not on the map. Why?

Abstract (arXiv)

The failure of deep neural networks to generalize to out-of-distribution data is a well-known problem and raises concerns about the deployment of trained networks in safety-critical domains such as healthcare, finance and autonomous vehicles. We study a particular kind of distribution shift unicodex2013 shortcuts or spurious correlations in the training data. Shortcut learning is often only exposed when models are evaluated on real-world data that does not contain the same spurious correlations, posing a serious dilemma for AI practitioners to properly assess the effectiveness of a trained model for real-world applications. In this work, we propose to use the mutual information (MI) between the learned representation and the input as a metric to find where in training, the network latches onto shortcuts. Experiments demonstrate that MI can be used as a domain-agnostic metric for monitoring shortcut learning.

Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Builds on 19 of its 27 references that are in the map. First posted on arXiv 27 Jun 2022.

Builds on

19 of 27 references.

References · 19

  1. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,690
  2. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
  3. 2018Neural Tangent Kernel: Convergence and Generalization in Neural Networkscited by 2,597
  4. 2020Shortcut Learning in Deep Neural Networkscited by 1,928
  5. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
  6. 2015Deep Learning and the Information Bottleneck Principlecited by 793
  7. 2017Opening the Black Box of Deep Neural Networks via Informationcited by 494
  8. 2019Causality matters in medical imagingcited by 193
  9. 2014Qualitatively characterizing neural network optimization problemscited by 186
  10. 2021A Fine-Grained Analysis on Distribution Shiftcited by 171
  11. 2018Strike (with) a Pose: Neural Networks Are Easily Fooled by Strange Poses of Familiar Objectscited by 143
  12. 2018i-RevNet: Deep Invertible Networkscited by 137
7 more
  1. 2018Dark Model Adaptation: Semantic Image Segmentation from Daytime to Nighttimecited by 122
  2. 2018Unsupervised domain adaptation for medical imaging segmentation with self-ensemblingcited by 94
  3. 2020Evaluating Saliency Map Explanations for Convolutional Neural Networks: A User Studycited by 86
  4. 2018Entropy and mutual information in models of deep neural networkscited by 50
  5. 2018Deep Learning Predicts Hip Fracture using Confounding Patient and Healthcare Variablescited by 50
  6. 2020Are Visual Explanations Useful? A Case Study in Model-in-the-Loop Predictioncited by 36
  7. 2019Information in Infinite Ensembles of Infinitely-Wide Neural Networkscited by 14

Built on by

Cited by year

Citing papers per year: 2023: 2; 2024: 2; 2025: 2.

6 papers.

  1. 2023Auxiliary Losses for Learning Generalizable Concept-based Modelscited by 35
  2. 2024Demonstrating and Reducing Shortcuts in Vision-Language Representation Learningcited by 14
  3. 2023Beyond Distribution Shift: Spurious Features Through the Lens of Training Dynamicscited by 11
  4. 2024Exploiting Supervised Poison Vulnerability to Strengthen Self-Supervised Defensecited by 1
  5. 2025Trust Me, I Know the Way: Predictive Uncertainty in the Presence of Shortcut Learningcited by 1
  6. 2025Diagnosing Shortcut-Induced Rigidity in Continual Learning: The Einstellung Rigidity Index (ERI)cited by 0