Red teaming must continually adapt as frontier LLM safeguards evolve, yet existing approaches largely search for adversarial prompts or optimize within fixed…
机构:复旦
来源:arXiv 2610.23726 | AI4Papers 论文推荐平台