Harrison Chase：来自这篇谷歌论文的好模式
中文全文 · AI 翻译
来自这篇谷歌论文的好模式
验证器从一个干净的上下文开始，所以探索者的推理无法说服它接受一个错误的证明
因此探索者可以尝试大胆的想法，而验证器会丢弃错误的那些
deepagents 的子代理获得自己的上下文也是同样的原因
对照原文
good pattern from this google paper the verifier starts from a clean context, so the explorer's reasoning can't talk it into a bad proof so the explorer can try wild ideas and the verifier tosses the wrong ones same reason deepagents subagents get their own context https://t.co/w177WWAdqY