ID and OOD Performance Are Sometimes Inversely Correlated on Real-world Datasets
This reveals a potential trade-off in model selection for OOD generalization, impacting practices in computer vision and NLP where ID performance is often used as a proxy.
This paper demonstrates that inverse correlations between in-distribution (ID) and out-of-distribution (OOD) performance occur in real-world datasets, challenging the common assumption of a positive correlation, and explains this theoretically in a minimal linear setting.
Several studies have compared the in-distribution (ID) and out-of-distribution (OOD) performance of models in computer vision and NLP. They report a frequent positive correlation and some surprisingly never even observe an inverse correlation indicative of a necessary trade-off. The possibility of inverse patterns is important to determine whether ID performance can serve as a proxy for OOD generalization capabilities. This paper shows with multiple datasets that inverse correlations between ID and OOD performance do happen in real-world data - not only in theoretical worst-case settings. We also explain theoretically how these cases can arise even in a minimal linear setting, and why past studies could miss such cases due to a biased selection of models. Our observations lead to recommendations that contradict those found in much of the current literature. - High OOD performance sometimes requires trading off ID performance. - Focusing on ID performance alone may not lead to optimal OOD performance. It may produce diminishing (eventually negative) returns in OOD performance. - In these cases, studies on OOD generalization that use ID performance for model selection (a common recommended practice) will necessarily miss the best-performing models, making these studies blind to a whole range of phenomena.