1 paper
Junze Zhou, David Klindt
Understanding when neural networks learn interpretable representations is central to mechanistic interpretability. We study this through identifiability: when can supervised learni…