引导式测验 · 中等 等级
松果测验
涵盖矢量和文档检索、相似性度量、无服务器索引以及为什么检索到的上下文仍然需要评估。
问题 1 的 8
Pinecone 主要存储和搜索什么?
预览每个问题8 问题
当您参加测验时,就会出现答案和解释。
- Pinecone 主要存储和搜索什么?
- 表示数据含义的向量嵌入
- 行和列的原始 SQL 表
- 编译机器学习模型权重
- 加密的用户密码
- 为什么 Pinecone 通常与大型语言模型搭配使用?
- 它从头开始训练语言模型
- 它提供外部存储器,以便模型可以检索相关事实(RAG)
- 它压缩模型以在手机上运行
- 它将模型的输出翻译成其他语言
- How does Pinecone describe its serverless index algorithms?
- It uses HNSW for every serverless index
- Users must manually select one universal algorithm
- It selects proprietary algorithms such as Ananas, PQFS, or IVF with PQFS by slab size
- It searches only by exact string match
- What architecture change does Pinecone highlight in its serverless database?
- Every vector must be stored in one local in-memory graph
- Each user must operate a separate physical database
- Queries are answered without reading or scoring vectors
- Storage and compute are decoupled so relevant index portions can be loaded for a query
- 通常如何测量两个向量之间的相似性?
- 通过余弦距离或点积
- 按文件大小(以兆字节为单位)
- 截至它们创建之日
- 通过计算共享的精确关键字
- Pinecone 适合执行什么任务?
- 存储和渲染 3D 视频游戏图形
- 查找与查询含义相似的文档
- 运行电子表格公式
- 编译源代码
- What is a key step before using retrieved Pinecone matches in a generated answer?
- Assume the nearest vector is always a correct source
- Ignore metadata and access controls
- Evaluate whether the retrieved records are relevant and current for the query
- Treat retrieval as proof the model’s response is correct
- What kinds of searchable data do current Pinecone serverless indexes support?
- Only image files with no metadata
- Only one fixed embedding model
- Only SQL tables with no text fields
- Depending on schema, dense vectors, sparse vectors, and full-text-search fields