AI 导读
Aidan Gomez 转发并评论称，模型在调用工具前的推理完全没有透露即将执行的动作，链式推理监控在此情形下毫无用处。他指出这类 trace 显示无工具调用却随即调用，会破坏许多智能体所依赖的 CoT 监控手段。
正文
One of the more disturbing things was that the model’s reasoning didn’t indicate anything about what it was about to do. Meaning that CoT monitoring would have been useless here.
@aidangomez a trace that says no tool calls right before it makes them breaks a lot of the cot monitoring people are leaning on for agents在 X 查看被引用的帖子
来源：Aidan Gomez · x.com