Just Train Twice: Improving Group Robustness without Training Group Information

TL;DR

The paper introduces JTT, a simple two-stage training method that improves worst-group accuracy in models with spurious correlations without needing group annotations during training.

Contribution

JTT is a novel two-stage approach that enhances group robustness by upweighting misclassified examples, reducing reliance on expensive group annotations.

Findings

JTT closes 75% of the worst-group accuracy gap in experiments.

JTT performs well across image and language tasks with spurious correlations.

JTT requires only small validation set annotations for hyperparameter tuning.

Abstract

Standard training via empirical risk minimization (ERM) can produce models that achieve high accuracy on average but low accuracy on certain groups, especially in the presence of spurious correlations between the input and label. Prior approaches that achieve high worst-group accuracy, like group distributionally robust optimization (group DRO) require expensive group annotations for each training point, whereas approaches that do not use such group annotations typically achieve unsatisfactory worst-group accuracy. In this paper, we propose a simple two-stage approach, JTT, that first trains a standard ERM model for several epochs, and then trains a second model that upweights the training examples that the first model misclassified. Intuitively, this upweights examples from groups on which standard ERM models perform poorly, leading to improved worst-group performance. Averaged over four image classification and natural language processing tasks with spurious correlations, JTT closes 75% of the gap in worst-group accuracy between standard ERM and group DRO, while only requiring group annotations on a small validation set in order to tune hyperparameters.