知识卡片
交叉编码联合比较查询与文档
交叉编码器将query和document一起编码,允许细粒度交互;双编码器分别编码便于预计算。联合打分更贵,常用于有限候选;输出单标量不自动意味着已校准相关概率。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 检索增强生成
原书出处 · 1 处
- · 查看原书依据
cross-encoders, as in these models the query and document are encoded together, as opposed to embedding-based retrieval models we have discussed, called bi-encoders, where the query and document are encoded as separate vectors.
参考来源
- 位置:《Designing Large Language Model Applications》第12章《检索增强生成》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch12.html)
- 结论依据:摘录中关于交叉编码联合比较查询与文档的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:cross-encoders, as in these models the query and document are encoded together, as opposed to embedding-based retrieval models we have discussed, called bi-encoders, where the query and document are encoded as separate vectors.