Spuriosity Didn't Kill the Classifier: Using Invariant Predictions to Harness Spurious Features

Instead of discarding “spurious” features whose relationship with the label changes across domains, we show they can be safely exploited in the test domain without labels: pseudo-labels from stable features provide sufficient guidance when stable and unstable features are conditionally independent given the label. Our Stable Feature Boosting algorithm learns an asymptotically-optimal predictor without test-domain labels.