Frontier LLM agents can turn known vulnerabilities into working exploits, but are costly and often access-restricted due to their potential for malicious use…
机构:上智院
来源:arXiv 2611.04196 | AI4Papers 论文推荐平台