Kedis ku xidhan hagaha · Dhexdhexaad Heerka

Kediska Soo Celinta Macnaha guud

Tijaabada soo celinta macnaha guud: diyaarinta macnaha guud ee dukumeentiga ilaa qaybo, ku-xidhka macnaha guud ee BM25, dib u habeynta, iyo kharashyada kaydinta degdega ah.

Wadooyinka haga ee la xidhiidhaSoo Celinta Macnaha Guud
Su'aal 1 ee 8

Dhibaato noocee ah ayuu xalliyaa soo celinta macnaha guud?

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Dhibaato noocee ah ayuu xalliyaa soo celinta macnaha guud?
    • Chunks lose information like company or time period when split from their document
    • Embeddings are too expensive to store
    • Language models cannot read tables
    • BM25 cannot handle long queries
  2. Sidee bay macnaha guud ee qayb kasta loo soo saaray?
    • By copying the document title only
    • A language model reads the whole document and the chunk and writes a short situating explanation
    • By a human editor for every chunk
    • By averaging neighboring embeddings
  3. Waa kuwee labada tusmo ee hela qoraalka kooban ee macnaha guud ee habka Anthropic?
    • A graph database and a cache
    • A SQL table and a spreadsheet
    • An embedding index and a BM25 keyword index
    • Two separate vector indexes with different models
  4. Marka loo eego tijaabooyinka la soo wariyay ee Anthropic, qiyaas ahaan intee in le'eg bay ku-xidhka ku-meel-gaadhka ah ee lagu daray BM25-ku-dhismay guul-darrooyinka-20-ka sare ee soo-celinta?
    • About 5 percent
    • About 20 percent
    • About 90 percent
    • About 49 percent
  5. Waa maxay sababta kaydinta degdega ah uga dhigayso soo celinta macnaha guud mid ka jaban?
    • It caches the document once and reuses it for every chunk's context call
    • It skips the embedding step
    • It removes the need for BM25
    • It shortens each chunk
  6. Why is keeping BM25 in the pipeline especially useful with added context?
    • BM25 understands synonyms better than embeddings
    • Contexts often add exact identifiers like names, dates or codes that keyword search matches precisely
    • BM25 requires no index
    • BM25 replaces reranking
  7. When might contextual retrieval be unnecessary?
    • When the documents are in English
    • When using a reranker
    • When the knowledge base is small enough to fit in the model's context window
    • When chunks are longer than 50 tokens
  8. Where does the main extra cost of contextual retrieval occur?
    • At query time for every user question
    • In the reranker only
    • In the user's browser
    • At indexing time, with one model call per chunk