Search papers, labs, and topics across Lattice.
Wuhan University
1
0
3
Scenario-wrapped prompts can significantly weaken LLM refusal safeguards, revealing shared vulnerabilities across model families that enhance attack success rates.
