1 paper
Walter Nelson, Marco Fumero, Theofanis Karaletsos +1
Representation learning models exhibit a surprising stability in their internal representations. Whereas most prior work treats this stability as a single property, we formalize it…