Search papers, labs, and topics across Lattice.
Affiliation:
1
0
3
4
Unsafe-response rates for Chinese LLMs can soar to over 30% under adversarial conditions, exposing vulnerabilities in automated safety judges.
