引導式測驗 · 中等 等級

GraphRAG 知識圖測驗

測試 GraphRAG 實體提取、圖形社群、本地和全域查詢、來源驗證以及社群報告背後的索引工作。

相關引導路徑Graphrag 知識圖譜
問題 1 的 8

GraphRAG 所建構的哪些核心結構使其與標準 RAG 不同?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. GraphRAG 所建構的哪些核心結構使其與標準 RAG 不同?
    • 使用者的關係 SQL 架構
    • 像素神經網路
    • 實體及其關係的知識圖
    • 停用詞列表
  2. What query scope motivated community reports in Microsoft's 2024 GraphRAG work?
    • Only changing a sentence's spelling
    • Collection-wide themes and summaries whose evidence spans many documents
    • Only measuring a video file's frame rate
    • A guarantee that all narrow lookups must use graph traversal
  3. GraphRAG 通常使用什麼演算法將相關實體聚集成社群?
    • 冒泡排序
    • 梯度下降
    • RSA加密
    • 萊頓社區檢測
  4. What does GraphRAG's model-based extraction stage produce from source chunks?
    • Structured entity and relationship records with descriptions
    • Only an executable operating-system image
    • Only random names unrelated to the source
    • Only audio waveforms with no text records
  5. How do documented GraphRAG local and global query strategies use context differently?
    • Local means Wi-Fi and global means satellite connectivity
    • Local search uses entity-related and source context; global search aggregates community reports
    • Local works only with images while global excludes all documents
    • Both strategies discard extracted information before generation
  6. 哪個組織在 2024 年普及了 GraphRAG 方法?
    • 一家 20 世紀 70 年代的電話公司
    • 一家速食連鎖店
    • Microsoft 研究
    • 航空聯盟
  7. Why should an important GraphRAG relationship be checked against its source documents?
    • The extraction model can make errors, and a graph edge does not independently establish the relationship's truth
    • Every extracted edge is automatically a verified fact
    • Graph storage replaces the need for source evidence
    • Community detection proves the original documents cannot be wrong
  8. What additional upfront work can GraphRAG indexing require compared with simply preparing text chunks?
    • No model calls or derived records of any kind
    • Automatic retraining of every user's language model
    • Publishing every source document to social media
    • Model-based entity and relationship extraction and generation of community reports