Table 1:

Summary of datasets. Datasets in blue were used as training sets and for test sets when benchmarking SSL. Datasets in red were used as test sets when benchmarking sound attribution.

Name	# Samples
Speaker-4M-E1	70,914
Edison-4M-E1	266,877
GerbilEarbud-4M-E1	7,698
SoloGerbil-4M-E1	61,513
DyadGerbil-4M-E1	653
Hexapod-8M-E2	156,900
MouseEarbud-24M-E3	200,000
SoloMouse-24M-E3	549
DyadMouse-24M-E3	2,191