AI RESEARCH
MoRFI: Monotonic Sparse Autoencoder Feature Identification
arXiv CS.LG
•
ArXi:2604.26866v1 Announce Type: cross Large language models (LLMs) acquire most of their factual knowledge during the pre-