Contextual Reliability: When Different Features Matter in Different Contexts
Not on the map. Why?
Abstract (arXiv)
Deep neural networks often fail catastrophically by relying on spurious correlations. Most prior work assumes a clear dichotomy into spurious and reliable features; however, this is often unrealistic. For example, most of the time we do not want an autonomous car to simply copy the speed of surrounding cars -- we don't want our car to run a red light if a neighboring car does so. However, we cannot simply enforce invariance to next-lane speed, since it could provide valuable information about an unobservable pedestrian at a crosswalk. Thus, universally ignoring features that are sometimes (but not always) reliable can lead to non-robust performance. We formalize a new setting called contextual reliability which accounts for the fact that the "right" features to use may vary depending on the context. We propose and analyze a two-stage framework called Explicit Non-spurious feature Prediction (ENP) which first identifies the relevant features to use for a given context, then trains a model to rely exclusively on these features. Our work theoretically and empirically demonstrates the advantages of ENP over existing methods and provides new benchmarks for contextual reliability.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 29 of its 41 references that are in the map. First posted on arXiv 19 Jul 2023.
Builds on
29 of 41 references.
References · 29
- 2014Explaining and Harnessing Adversarial Examplescited by 7,287
- 2019Invariant Risk Minimizationcited by 1,832
- 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
- 2019Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalizationcited by 1,232
- 2020In Search of Lost Domain Generalizationcited by 1,042
- 2020Concept Bottleneck Modelscited by 1,022
- 2017Learning to Generalize: Meta-Learning for Domain Generalizationcited by 762
- 2013Domain Generalization via Invariant Feature Representationcited by 622
- 2015Causal inference using invariant prediction: identification and confidence intervalscited by 533
- 2021Large Scale Interactive Motion Forecasting for Autonomous Driving : The Waymo Open Motion Datasetcited by 532
- 2021Just Train Twice: Improving Group Robustness without Training Group Informationcited by 522
- 2018Confounding variables can degrade generalization performance of radiological deep learning modelscited by 374
17 more
- 2020An Investigation of Why Overparameterization Exacerbates Spurious Correlationscited by 310
- 2018Certified Defenses against Adversarial Examplescited by 305
- 2017Right for the Right Reasons: Training Differentiable Models by Constraining their Explanationscited by 293
- 2019Causal Confusion in Imitation Learningcited by 286
- 2020The Risks of Invariant Risk Minimizationcited by 262
- 2021MultiPath++: Efficient Information Fusion and Trajectory Aggregation for Behavior Predictioncited by 246
- 2017Invariant Causal Prediction for Nonlinear Modelscited by 140
- 2020Learning from Failure: Training Debiased Classifier from Biased Classifiercited by 133
- 2021The iWildCam 2021 Competition Datasetcited by 77
- 2020Decoupling Exploration and Exploitation for Meta-Reinforcement Learning without Sacrificescited by 59
- 2021Do Input Gradients Highlight Discriminative Features?cited by 52
- 2022Causal Confusion and Reward Misidentification in Preference-Based Reward Learningcited by 52
- 2022Active Learning Helps Pretrained Models Learn the Intended Taskcited by 33
- 2020Bayesian Robust Optimization for Imitation Learningcited by 28
- 2022CausalAgents: A Robustness Benchmark for Motion Forecasting using Causal Relationshipscited by 18
- 2021Policy Gradient Bayesian Robust Optimization for Imitation Learningcited by 11
- 2022Regularising for invariance to data augmentation improves supervised learningcited by 9
Built on by
Not cited yet.