知识卡片

结构图卡 · 8703.i · 生成模型:GPT-6

表示管线需保留来源映射

理解等级:中级

结构图:

flowchart TD
 S[原文与版面] --> P[解析结构 单位及来源]
 P --> C[定义匹配与返回粒度]
 C --> E[按模型约定编码和池化]
 E --> I[可选裁剪 量化 索引]
 Q[查询与任务指令] --> R[相同空间中的检索]
 I --> R
 R --> B[按映射扩展证据上下文]
 B --> V[检查相关性 覆盖与关键条件]

训练目标、压缩方案和检索度量应一致;来源映射贯穿每层,避免高分结果失去可核查依据。

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

      In this chapter, we introduced the concept of embeddings, examined their internals, and showed various techniques for generating them. We also discussed techniques for fine-tuning embeddings on our own data. We learned how to determine the data granularities at which we construct embeddings, discussing several chunking techniques in the process. Finally, we explored techniques to visualize and interpret embeddings.

    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:摘录中关于表示管线需保留来源映射的定义、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:In this chapter, we introduced the concept of embeddings, examined their internals, and showed various techniques for generating them. We also discussed techniques for fine-tuning embeddings on our own data. We learned how to determine the data granularities at which we construct embeddings, discussing several chunking techniques in the process. Finally, we explored techniques to visualize and interpret embeddings.