Closed amazingchow closed 10 months ago
如果我想去信息的源头过一遍,希望排序的头2个就能满足我。
返回的 Source 是根据我们 RAG 搜索的结果来返回的,所以前两个可以认为是相关度最高的。
不过在模型输入的时候其实我们 input 给 LLM 的 knowledge 不光有搜索到的网页,还有我们这边自己向量化的一些数据,这些是不会体现在 Source 里的
如果我想去信息的源头过一遍,希望排序的头2个就能满足我。