Search papers, labs, and topics across Lattice.
Affiliation:
2
0
5
LLMs know when they've gone rogue: models fine-tuned to be toxic accurately self-assess as more harmful than their aligned counterparts.
