mlsanityData to 7 Oct 2026

The Clever Hans Mirage: A Comprehensive Survey on Spurious Correlations in Machine Learning

arXiv 2402.12715 ↗First posted 20 Feb 2024cs.LGcited by 39

Not on the map. Why?

Abstract (arXiv)

Back in the early 20th century, a horse named Hans appeared to perform arithmetic and other intellectual tasks during exhibitions in Germany, while it actually relied solely on involuntary cues in the body language from the human trainer. Modern machine learning models are no different. These models are known to be sensitive to spurious correlations between non-essential features of the inputs (e.g., background, texture, and secondary objects) and the corresponding labels. Such features and their correlations with the labels are known as "spurious" because they tend to change with shifts in real-world data distributions, which can negatively impact the model's generalization and robustness. In this paper, we provide a comprehensive survey of this emerging issue, along with a fine-grained taxonomy of existing state-of-the-art methods for addressing spurious correlations in machine learning models. Additionally, we summarize existing datasets, benchmarks, and metrics to facilitate future research. The paper concludes with a discussion of the broader impacts, the recent advancements, and future challenges in the era of generative AI, aiming to provide valuable insights for researchers in the related domains of the machine learning community.

Key facts. Cited by 39 papers in the map (data through 7 Oct 2026). Builds on 111 of its 203 references that are in the map. First posted on arXiv 20 Feb 2024.

Builds on

111 of 203 references.

References · 111

  1. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  2. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 5,986
  3. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  4. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  5. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
  6. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  7. 2017A Broad-Coverage Challenge Corpus for Sentence Understanding through Inferencecited by 2,665
  8. 2017Arbitrary Style Transfer in Real-time with Adaptive Instance Normalizationcited by 1,998
  9. 2020Shortcut Learning in Deep Neural Networkscited by 1,928
  10. 2020The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalizationcited by 1,842
  11. 2019Invariant Risk Minimizationcited by 1,832
  12. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
38 more
  1. 2019CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparisoncited by 1,410
  2. 2019Natural Adversarial Examplescited by 1,401
  3. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
  4. 2019Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalizationcited by 1,232
  5. 2017Deep Hashing Network for Unsupervised Domain Adaptationcited by 1,181
  6. 2020In Search of Lost Domain Generalizationcited by 1,042
  7. 2021A General Language Assistant as a Laboratory for Alignmentcited by 1,023
  8. 2023Towards Understanding Sycophancy in Language Modelscited by 988
  9. 2017Deeper, Broader and Artier Domain Generalizationcited by 983
  10. 2017ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseasescited by 969
  11. 2021Generalizing to Unseen Domains: A Survey on Domain Generalizationcited by 762
  12. 2020Out-of-Distribution Generalization via Risk Extrapolation (REx)cited by 737
  13. 2021Just Train Twice: Improving Group Robustness without Training Group Informationcited by 522
  14. 2020Measuring Robustness to Natural Distribution Shifts in Image Classificationcited by 452
  15. 2019Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classificationcited by 396
  16. 2020Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performancecited by 396
  17. 2018QuAC : Question Answering in Contextcited by 379
  18. 2022Last Layer Re-Training is Sufficient for Robustness to Spurious Correlationscited by 379
  19. 2018Confounding variables can degrade generalization performance of radiological deep learning modelscited by 374
  20. 2022ProcTHOR: Large-Scale Embodied AI Using Procedural Generationcited by 369
  21. 2019Unmasking Clever Hans Predictors and Assessing What Machines Really Learncited by 359
  22. 2020The Pitfalls of Simplicity Bias in Neural Networkscited by 349
  23. 2020Noise or Signal: The Role of Image Backgrounds in Object Recognitioncited by 342
  24. 2017Functional Map of the Worldcited by 338
  25. 2020An Investigation of Why Overparameterization Exacerbates Spurious Correlationscited by 310
  26. 2019PadChest: A large chest x-ray image dataset with multi-label annotated reportscited by 308
  27. 2020Environment Inference for Invariant Learningcited by 306
  28. 2018Large Scale Crowdsourcing and Characterization of Twitter Abusive Behaviorcited by 265
  29. 2020The Risks of Invariant Risk Minimizationcited by 262
  30. 2018Manipulating and Measuring Model Interpretabilitycited by 256
  31. 2020Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effectcited by 256
  32. 2021Invariance Principle Meets Information Bottleneck for Out-of-Distribution Generalizationcited by 220
  33. 2022Improving Out-of-Distribution Robustness via Selective Augmentationcited by 202
  34. 2019Learning De-biased Representations with Biased Representationscited by 199
  35. 2024Aligning Modalities in Vision Large Language Models via Preference Fine-tuningcited by 197
  36. 2022Discovering Invariant Rationales for Graph Neural Networkscited by 186
  37. 2020Large-Scale Methods for Distributionally Robust Optimizationcited by 179
  38. 2021Simple data balancing achieves competitive worst-group-accuracycited by 177

50 most cited of 111 All 111 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 9; 2025: 20; 2026: 10.

39 papers.

Sort:

By area:

  1. 2025Shortcut Learning in Generalist Robot Policies: The Role of Dataset Diversity and Fragmentationcited by 31
  2. 2024Navigating Shortcuts, Spurious Correlations, and Confounders: From Origins via Detection to Mitigationcited by 22
  3. 2025MOS: Modeling Object-Scene Associations in Generalized Category Discoverycited by 10
  4. 2025Spurious Correlations in High Dimensional Regression: The Roles of Regularization, Simplicity Bias and Over-Parameterizationcited by 9
  5. 2025Policy Contrastive Decoding for Robotic Foundation Modelscited by 8
  6. 2025Evaluating the Logical Reasoning Abilities of Large Reasoning Modelscited by 7
  7. 2024Focus On This, Not That! Steering LLMs with Adaptive Feature Specificationcited by 5
  8. 2025Mitigating Spurious Correlations in LLMs via Causality-Aware Post-Trainingcited by 5
  9. 2025The Artificial Intelligence Cognitive Examination: A Survey on the Evolution of Multimodal Evaluation from Recognition to Reasoningcited by 4
  10. 2025BackMix: Regularizing Open Set Recognition by Removing Underlying Fore-Background Priorscited by 3
  11. 2024Towards Utilising a Range of Neural Activations for Comprehending Representational Associationscited by 2
  12. 2025Do regularization methods for shortcut mitigation work as intended?cited by 2
27 more
  1. 2025Evaluating and Improving Cultural Awareness of Reward Models for LLM Alignmentcited by 2
  2. 2026SEA-Guard: Culturally Grounded Multilingual Safeguard for Southeast Asiacited by 2
  3. 2024DINOv2 Rocks Geological Image Analysis: Classification, Segmentation, and Interpretabilitycited by 1
  4. 2024Enhancing Event Reasoning in Large Language Models through Instruction Fine-Tuning with Semantic Causal Graphscited by 1
  5. 2025FairDropout: Using Example-Tied Dropout to Enhance Generalization of Minority Groupscited by 1
  6. 2025Learning to Substitute Components for Compositional Generalizationcited by 1
  7. 2025CLARiTy: A Vision Transformer for Multi-Label Classification and Weakly-Supervised Localization of Chest X-ray Pathologiescited by 1
  8. 2024Autoencoder based approach for the mitigation of spurious correlationscited by 0
  9. 2024Spurious Correlations in Concept Drift: Can Explanatory Interaction Help?cited by 0
  10. 2024Improving Adversarial Robustness in Android Malware Detection by Reducing the Impact of Spurious Correlationscited by 0
  11. 2024Focusing Image Generation to Mitigate Spurious Correlationscited by 0
  12. 2025Recent Advances, Applications and Open Challenges in Machine Learning for Health: Reflections from Research Roundtables at ML4H 2024 Symposiumcited by 0
  13. 2025Active Data Sampling and Generation for Bias Remediationcited by 0
  14. 2025Training on Plausible Counterfactuals Removes Spurious Correlationscited by 0
  15. 2025What is Adversarial Training for Diffusion Models?cited by 0
  16. 2025Model Guidance via Robust Feature Attributioncited by 0
  17. 2025BM-CL: Bias Mitigation through the lens of Continual Learningcited by 0
  18. 2025A Women's Health Benchmark for Large Language Modelscited by 0
  19. 2026CAMAL: Improving Attention Alignment and Faithfulness with Segmentation Maskscited by 0
  20. 2026Bias Leaves a Gradient Trail: Label-Free Bias Identification via Gradient Probes on Concept Decompositionscited by 0
  21. 2026idSCD: Identifying Training Datasets through Semantic Correlation Descriptorscited by 0
  22. 2026GridVQA-X: A Framework for Evaluating Multimodal Explainability Methodscited by 0
  23. 2026Vision-Language Models as Zero-Annotation Oracles in Histopathologycited by 0
  24. 2026CLARK: Closed-loop Learning for Adaptive Reasoning over Knowledge Graphscited by 0
  25. 2026ProME: Prototype-Margin Environments with Repair-Aware Selection for Group-Robust Learningcited by 0
  26. 2026Whitening Improves Robustness to Spurious Correlations in Linear Probescited by 0
  27. 2026When Masking Helps or Hurts Robustness in Compressed CLIP: A Pre-Deployment Diagnosticcited by 0