跳到正文
@rohanpaul_ai· @rohanpaul_ai · X·· 14 天前AI 评分26
AI 导读

这一发现在全智能体层面极其脆弱。 10 次重跑都到达了相关邻域,却没有一次复现关键观察结果。

正文

The discovery was brutally brittle at the full-agent level.

10 reruns kept reaching the relevant neighborhood, yet none repeated the key observation. https://t.co/VxXRmGS9QZ

来源:@rohanpaul_ai · x.com