Mitigating Spurious Correlations with Memorization-Guided Dataset De-Biasing

In three linesTwo-stage sample scoring function to mitigate spurious correlations by disentangling learning dynamics of core and spurious features. Algorithm prioritizes informative samples; standard ERM trained on selected data achieves better performance than state-of-the-art debiasing with only 10% of original training data.

Summary generated by Claude — human-verified