Simple data balancing achieves competitive worst-group-accuracy

We study the problem of learning classifiers that perform well across (known or unknown) groups of data. After observing that common worst-group-accuracy datasets suffer from substantial imbalances, we set out to compare state-of-the-art methods to simple balancing of classes and groups by either subsampling or reweighting data. Our results show that these data balancing baselines achieve state-of-the-art-accuracy, while being faster to train and requiring no additional hyper-parameters. Finally, we highlight that access to group information is most critical for model selection purposes, and not so much during training. All in all, our findings beg closer examination of both benchmarks and methods for future research in worst-group-accuracy optimization.