Class indices come from the CC BY 4.0 dataset CSV, not the CC BY-SA ontology

README.md

@@ -51,7 +51,8 @@ here: each vector is a normalised multi-hot indicator over a class group — zer
 everywhere, `1/n` at each of the `n` chosen indices — so a class list fully
 determines it.
 
-The AudioSet ontology is CC BY-SA 4.0, a different licence from this model's.
+Class indices follow AudioSet's `class_labels_indices.csv`, published by Google
+with the dataset under CC BY 4.0.
 
 ## Used by