Table 2.

Statistics of the datasets for the WILDS Camelyon17 Histopathology dataset

Dataset	Type	N images	% positives
Sites S1–S3	Reference set $D_{ref}$	33,560	50%
Site S4	Unseen	34,904	50%
Site S5	Unseen	85,054	50%

Source data are provided as a Source Data file.