WILDS: A Benchmark of in-the-Wild Distribution Shifts

TL;DR

WILDS introduces a comprehensive benchmark of 10 real-world datasets exhibiting diverse distribution shifts, highlighting the challenges in achieving robust ML performance across varied practical scenarios.

Contribution

The paper presents WILDS, a new benchmark with datasets reflecting real-world distribution shifts, and provides tools for standardized evaluation and fostering development of robust models.

Findings

Standard models perform poorly on distribution-shifted data.

Existing methods do not fully mitigate performance gaps.

WILDS facilitates research on robustness to distribution shifts.

Abstract

Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity in the real-world deployments, these distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated benchmark of 10 datasets reflecting a diverse range of distribution shifts that naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training yields substantially lower out-of-distribution than in-distribution performance. This gap remains even with models trained by existing methods for tackling distribution shifts, underscoring the need for new methods for training models that are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at https://wilds.stanford.edu.