mlsanityData to 7 Oct 2026

FACTS: First Amplify Correlations and Then Slice to Discover Bias

arXiv 2309.17430 ↗First posted 29 Sep 2023cs.CVcited by 26

Not on the map. Why?

Abstract (arXiv)

Computer vision datasets frequently contain spurious correlations between task-relevant labels and (easy to learn) latent task-irrelevant attributes (e.g. context). Models trained on such datasets learn "shortcuts" and underperform on bias-conflicting slices of data where the correlation does not hold. In this work, we study the problem of identifying such slices to inform downstream bias mitigation strategies. We propose First Amplify Correlations and Then Slice to Discover Bias (FACTS), wherein we first amplify correlations to fit a simple bias-aligned hypothesis via strongly regularized empirical risk minimization. Next, we perform correlation-aware slicing via mixture modeling in bias-aligned feature space to discover underperforming data slices that capture distinct correlations. Despite its simplicity, our method considerably improves over prior work (by as much as 35% precision@10) in correlation bias identification across a range of diverse evaluation settings. Our code is available at: https://github.com/yvsriram/FACTS.

Key facts. Cited by 26 papers in the map (data through 7 Oct 2026). Builds on 26 of its 43 references that are in the map. First posted on arXiv 29 Sep 2023.

Builds on

26 of 43 references.

References · 26

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  3. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,690
  4. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  5. 2020Shortcut Learning in Deep Neural Networkscited by 1,928
  6. 2016Semantics derived automatically from language corpora contain human-like biasescited by 894
  7. 2021ClipCap: CLIP Prefix for Image Captioningcited by 557
  8. 2021Just Train Twice: Improving Group Robustness without Training Group Informationcited by 522
  9. 2019Unsupervised Embedding Learning via Invariant and Spreading Instance Featurecited by 231
  10. 2019Hidden Stratification Causes Clinically Meaningful Failures in Machine Learning for Medical Imagingcited by 215
  11. 2022Domino: Discovering Systematic Errors with Cross-Modal Embeddingscited by 155
  12. 2021Salient ImageNet: How to discover spurious features in Deep Learning?cited by 114
14 more
  1. 2022Spread Spurious Attribute: Improving Worst-group Accuracy with Spurious Attribute Estimationcited by 88
  2. 2022NICO++: Towards Better Benchmarking for Domain Generalizationcited by 86
  3. 2021Leveraging Sparse Linear Layers for Debuggable Deep Networkscited by 82
  4. 2022Distilling Model Failures as Directions in Latent Spacecited by 82
  5. 2019Predictive Inequity in Object Detectioncited by 77
  6. 2023Diagnosing and Rectifying Vision Models using Languagecited by 58
  7. 2020Understanding Failures of Deep Networks via Robust Feature Extractioncited by 56
  8. 2018Deep Learning Predicts Hip Fracture using Confounding Patient and Healthcare Variablescited by 50
  9. 2023Model Agnostic Sample Reweighting for Out-of-Distribution Learningcited by 43
  10. 2022ModelDiff: A Framework for Comparing Learning Algorithmscited by 30
  11. 2020Deep Learning Applied to Chest X-Rays: Exploiting and Preventing Shortcutscited by 29
  12. 2021UDIS: Unsupervised Discovery of Bias in Deep Visual Recognition Modelscited by 23
  13. 2022Measures of Information Reflect Memorization Patternscited by 12
  14. 2022AGRO: Adversarial Discovery of Error-prone groups for Robust Optimizationcited by 10

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 1; 2024: 10; 2025: 11; 2026: 4.

26 papers.

Sort:

By area:

  1. 2024The Clever Hans Mirage: A Comprehensive Survey on Spurious Correlations in Machine Learningcited by 39
  2. 2025Integrating LLMs with ITS: Recent Advances, Potentials, Challenges, and Future Directionscited by 23
  3. 2024Fairness and Bias Mitigation in Computer Vision: A Surveycited by 12
  4. 2025HiBug2: Efficient and Interpretable Error Slice Discovery for Comprehensive Model Debuggingcited by 11
  5. 2024Efficient Bias Mitigation Without Privileged Informationcited by 8
  6. 2025On Large Multimodal Models as Open-World Image Classifierscited by 8
  7. 2024LADDER: Language-Driven Slice Discovery and Error Rectification in Vision Classifierscited by 7
  8. 2024Visual Data Diagnosis and Debiasing with Concept Graphscited by 6
  9. 2024What could go wrong? Discovering and describing failure modes in computer visioncited by 5
  10. 2025ConceptScope: Characterizing Dataset Bias via Disentangled Visual Conceptscited by 5
  11. 2024Multi-View Incongruity Learning for Multimodal Sarcasm Detectioncited by 4
  12. 2025The Impact of Coreset Selection on Spurious Correlations and Group Robustnesscited by 4
14 more
  1. 2025Subgroup Performance Analysis in Hidden Stratificationscited by 3
  2. 2024ViG-Bias: Visually Grounded Bias Discovery and Mitigationcited by 2
  3. 2025Superclass-Guided Representation Disentanglement for Spurious Correlation Mitigationcited by 2
  4. 2025FailureAtlas:Mapping the Failure Landscape of T2I Models via Active Explorationcited by 2
  5. 2023Language-assisted Vision Model Debugger: A Sample-Free Approach to Finding and Fixing Bugscited by 1
  6. 2024Distributionally robust self-supervised learning for tabular datacited by 1
  7. 2025Does Weak-to-strong Generalization Happen under Spurious Correlations?cited by 1
  8. 2025Let Samples Speak: Mitigating Spurious Correlation by Exploiting the Clusterness of Samplescited by 1
  9. 2024Towards Real-world Debiasing: Rethinking Evaluation, Challenge, and Solutioncited by 0
  10. 2025Bias-Aware Machine Unlearning: Towards Fairer Vision Models via Controllable Forgettingcited by 0
  11. 2026LatentDiff: Scaling Semantic Dataset Comparison to Millions of Imagescited by 0
  12. 2026CB-SLICE: Concept-Based Interpretable Error Slice Discoverycited by 0
  13. 2026Invisible Shortcuts: Why Vision Encoders Know Your Cameracited by 0
  14. 2026Beyond Benchmarks: Using VLMs to Reveal Systematic Classification Failures Under Real World Conditionscited by 0