Search papers, labs, and topics across Lattice.
Northwestern University in
1
0
3
LLM safety evaluations in English miss critical vulnerabilities: GPT-4o's refusal rate on Kazakh safety prompts varies from 5.5% to 53.8% across categories.
