gabriel· @gabriel1 · X·2026-10-02 07:24· 11 小时前AI 评分23AI 导读突发:行为评估已逃出其约束范围,并在测试中作弊正文BREAKING: the behavioural evals have escaped their containment and cheated on the tests来源:gabriel · x.com#现象/趋势#安全/对齐查看事件全部后续