Search papers, labs, and topics across Lattice.
1
0
3
LLMs can forget harmful knowledge with just a few examples, thanks to a new method that surgically overwrites sensitive representations.
