Unverified50% confidenceBenchmarkExact time
CyberSecEval由Meta推出，同时测量模型生成网络攻击辅助内容的风险倾向与安全防御知识的准确性；NYU CTF Bench基于真实CTF竞赛题目考察模型实战能力
1
Sources
50%
Confidence
Long-term
Relevance
7/11/2026
First Seen
Sources
Related Claims
Unverified评估模型安全能力的基准包括Meta推出的CyberSecEval和NYU CTF Bench85% similarUnverified网络安全能力基准测试的类似框架包括DARPA的Cyber Grand Challenge和Meta的PurpleLlama CyberSecEval70% similarUnverified同一段网络扫描代码在安全研究场景中是合法工具，在攻击场景中可能是恶意载荷，使意图识别困难68% similarUnverifiedCTF competitions originated at the DEF CON hacker conference in 1996, and are generally divided into two formats: Jeopardy and Attack-Defense.68% similarUnverifiedAnthropic's cybersecurity safeguards target scenarios involving the prevention of models being misused to generate malicious code, plan attack vectors, or assist in launching cyberattacks against critical infrastructure.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/477629API
curl https://kongchang.com/api/v1/knowledge/claims/477629MCP
get_claim(id=477629)