Search papers, labs, and topics across Lattice.
1
0
Up to 77% of features that appear recoverable in sparse autoencoders are actually causally inert, raising questions about the reliability of current evaluation metrics.