<ul><li>Spurious correlations pose a challenge for robust real-world generalization in machine learning.</li><li>Existing methods address this issue by maximizing group-balanced or worst-group accuracy, but they heavily rely on expensive bias annotations.</li><li>A new method is proposed to tackle spurious correlations by reframing them as imbalances or mismatches in class-conditional distributions, eliminating the need for bias annotations or predictions.</li><li>The proposed method achieves class-conditional distribution balancing and produces a debiased data distribution for classification, delivering state-of-the-art performance.</li></ul>

Class-Conditional Distribution Balancing for Group Robust Classification

Discover more