A benchmark study finds that combining multiple distribution shifts usually hurts accuracy, spurious correlations dominate, and simple data augmentations beat specialized robustness methods.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
An Analysis of Model Robustness across Concurrent Distribution Shifts
A benchmark study finds that combining multiple distribution shifts usually hurts accuracy, spurious correlations dominate, and simple data augmentations beat specialized robustness methods.