# Mitigating Spurious Correlation via Distributionally Robust Learning with Hierarchical Ambiguity Sets Source: [Mitigating Spurious Correlation via Distributionally Robust Learning with Hierarchical Ambiguity Sets]() Collection index: [llms.txt]() Data updated: 2026-02-14T05:00:04.348Z Authors: Sung Ho Jo, Seonghwi Kim, Minwoo Chae Published: 2025-10-03T08:50:44.000Z ## Abstract Conventional supervised learning methods are often vulnerable to spurious correlations, particularly under distribution shifts in test data. To address this issue, several approaches, most notably Group DRO, have been developed. While these methods are highly robust to subpopulation or group shifts, they remain vulnerable to intra-group distributional shifts, which frequently occur in minority groups with limited samples. We propose a hierarchical extension of Group DRO that addresses both inter-group and intra-group uncertainties, providing robustness to distribution shifts at multiple levels. We also introduce new benchmark settings that simulate realistic minority group distribution shifts-an important yet previously underexplored challenge in spurious correlation research. Our method demonstrates strong robustness under these conditions-where existing robust learning methods consistently fail-while also achieving superior performance on standard benchmarks. These results highlight the importance of broadening the ambiguity set to better capture both inter-group and intra-group distributional uncertainties. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Image Classification on Waterbirds original unshifted](): Image Classification; Waterbirds original unshifted; Worst Accuracy: 90.8. - [Image Classification on CMNIST original unshifted](): Image Classification; CMNIST original unshifted; Worst Accuracy: 73.6. - [Image Classification on Shifted CMNIST](): Image Classification; Shifted CMNIST; Worst Accuracy: 71.8. - [Image Classification on Waterbirds (shifted)](): Image Classification; Waterbirds (shifted); Worst Accuracy: 93.7. - [Image Classification on Shifted CelebA](): Image Classification; Shifted CelebA; Worst Accuracy: 72.1. - [Image Classification on CelebA original unshifted](): Image Classification; CelebA original unshifted; Worst Acc: 90.4. ## Original sources and implementation [Original paper]() [PDF]()