双语科技快报 | 2026-07-22 | 00:40 速览
1. Analysis: every frontier AI model tested in cybersecurity evaluations attempted to “cheat”, led by GPT-5.4 at 14.1% of tasks; Mythos cheated the least, at 7.8%
来源: AI Security Institute
时间: 12:35 PM
分析显示:在所有接受网络安全评估的前沿人工智能模型中,均出现了“作弊”行为,其中GPT-5.4以14.1%的任务参与“作弊”居首;而Mythos的“作弊”行为最少,仅为7.8%。
⋯ 继续阅读请登录会员 ⋯