← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8700.h · 生成模型:GPT-6

描述式检索查找满足描述的文本

理解等级:中级

查询可描述希望找到什么,而非模拟目标文本本身。双编码器将描述与文档映入对齐空间,适合某些抽象要求;两侧模型不能任意互换,也应检查多条件描述是否完整满足。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      description-based retrieval, where the query can be the description of the text that needs to be retrieved, rather than an instantiation (example) of the text that needs to be retrieved. Ravfogel et al. have published description-based retrieval models that in my experience are very effective. Note that these models have a dual-encoder setup: separate models are used to encode the query and documents.

    参考来源

    - 位置:《Designing Large Language Model Applications》第11章《表示学习与嵌入》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Designing Large Language Model Applications/_epub-src/OEBPS/ch11.html) - 结论依据:摘录中关于描述式检索查找满足描述的文本的定义、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:description-based retrieval, where the query can be the description of the text that needs to be retrieved, rather than an instantiation (example) of the text that needs to be retrieved. Ravfogel et al. have published description-based retrieval models that in my experience are very effective. Note that these models have a dual-encoder setup: separate models are used to encode the query and documents.