Search papers, labs, and topics across Lattice.
Affiliation:
2
0
4
7
The red-teaming methodology and highlighting new attack vectors aim to help defenders evaluate their mitigations against the possibility of persistent malign coding agents.
Amplifying reasoning weights can uncover hidden model secrets up to 10 times more effectively than standard approaches.
