Search papers, labs, and topics across Lattice.
Affiliation:
1
0
Multi-turn attacks expose LLM vulnerabilities that single-turn prompts miss, with EvoFlint achieving up to 94.3% success in exploiting these weaknesses.
