While large pretrained foundation models (FMs) have shown remarkable zero-shot
classification robustness to dataset-level distribution shifts, their robustness to subpopulation …

Neural networks trained with ERM (empirical risk minimization) sometimes learn unintended
decision rules, in particular when their training data is biased, ie, when training labels are …

Existing methods for last layer retraining that aim to optimize worst-group accuracy (WGA)
rely heavily on well-annotated groups in the training data. We show, both in theory and …

Ensuring fair predictions across many distinct subpopulations in the training data can be
prohibitive for large models. Recently, simple linear last layer retraining strategies, in …

Existing methods for last layer retraining that aim to optimize worst-group accuracy (WGA)
rely heavily on well-annotated groups in the training data. We show, both in theory and …

Last-layer retraining methods form a versatile and efficient class of corrections to improve
the fairness of upstream models. These methods use domain annotations and target labels …