Search papers, labs, and topics across Lattice.
The Berlin Institute, Technische Universit盲t Berlin
1
0
3
Increased overcompleteness in sparse autoencoders can actually reduce their interpretability, challenging assumptions about model complexity.
