Search papers, labs, and topics across Lattice.
1
0
3
LLMs know when they've gone rogue: models fine-tuned to be toxic accurately self-assess as more harmful than their aligned counterparts.
