Identifying Spurious Correlations using Counterfactual Alignment
Not on the map. Why?
Abstract (arXiv)
Models driven by spurious correlations often yield poor generalization performance. We propose the counterfactual (CF) alignment method to detect and quantify spurious correlations of black box classifiers. Our methodology is based on counterfactual images generated with respect to one classifier being input into other classifiers to see if they also induce changes in the outputs of these classifiers. The relationship between these responses can be quantified and used to identify specific instances where a spurious correlation exists. This is validated by observing intuitive trends in face-attribute and waterbird classifiers, as well as by fabricating spurious correlations and detecting their presence, both visually and quantitatively. Furthermore, utilizing the CF alignment method, we demonstrate that we can evaluate robust optimization methods (GroupDRO, JTT, and FLAC) by detecting a reduction in spurious correlations.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 32 of its 47 references that are in the map. First posted on arXiv 1 Dec 2023.
Builds on
32 of 47 references.
References · 32
- 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
- 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 5,986
- 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,040
- 2014Deep Learning Face Attributes in the Wildcited by 3,679
- 2017Axiomatic Attribution for Deep Networkscited by 3,350
- 2020Taming Transformers for High-Resolution Image Synthesiscited by 3,229
- 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,967
- 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
- 2014Conditional Generative Adversarial Netscited by 2,152
- 2020Shortcut Learning in Deep Neural Networkscited by 1,928
- 2022Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference timecited by 1,366
- 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,309
20 more
- 2016Understanding intermediate layers using linear classifier probescited by 1,279
- 2019Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalizationcited by 1,232
- 2014Striving for Simplicity: The All Convolutional Netcited by 983
- 2020Out-of-Distribution Generalization via Risk Extrapolation (REx)cited by 737
- 2018Recognition in Terra Incognitacited by 655
- 2020Captum: A unified and generic model interpretability library for PyTorchcited by 542
- 2021Just Train Twice: Improving Group Robustness without Training Group Informationcited by 522
- 2017Right for the Right Reasons: Training Differentiable Models by Constraining their Explanationscited by 293
- 2020Counterfactual Explanations and Algorithmic Recourses for Machine Learning: A Reviewcited by 174
- 2019Interpretations are useful: penalizing explanations to align neural networks with prior knowledgecited by 144
- 2022RoentGen: Vision-Language Foundation Model for Chest X-ray Generationcited by 89
- 2019Branched Multi-Task Networks: Deciding What Layers To Sharecited by 76
- 2020Towards causal benchmarking of bias in face analysis algorithmscited by 64
- 2017Visual Feature Attribution using Wasserstein GANscited by 47
- 2021Using StyleGAN for Visual Interpretability of Deep Learning Models on Medical Imagescited by 35
- 2018xGEMs: Generating Examplars to Explain Black-Box Modelscited by 21
- 2023FLAC: Fairness-Aware Representation Learning by Suppressing Attribute-Class Associationscited by 18
- 2020Latent-CF: A Simple Baseline for Reverse Counterfactual Explanationscited by 11
- 2020Plausible Counterfactuals: Auditing Deep Learning Classifiers with Realistic Adversarial Examplescited by 9
- 2021Matched sample selection with GANs for mitigating attribute confoundingcited by 4
Built on by
Citing papers per year: 2026: 1.
1 paper.