Kediska Raadinta AI

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. In retrieval-augmented search, what may happen before an answer is generated?
   • Relevant records are retrieved and supplied as context
   • The model’s parameters are retrained on the query
   • The result is guaranteed to be correct
   • All documents are deleted from the index
2. When is keyword search especially useful?
   • Only for queries with synonyms
   • When no text is indexed
   • For exact names, codes, or phrases
   • When the system must create a summary
3. What does semantic retrieval help with?
   • Guaranteeing that two documents mean the same thing
   • Finding related ideas expressed with different wording
   • Replacing all exact-match search
   • Verifying the truth of a claim
4. What is hybrid search?
   • Searching two browsers at once
   • Using only a generated answer
   • Replacing ranking with user votes
   • Combining lexical and semantic signals
5. What should a generated answer’s citations allow the user to do?
   • Inspect whether a cited passage supports each claim
   • Assume the answer has no errors
   • Skip reading the source
   • Confirm that the source is the newest
6. A citation points to a broadly related page but not a passage supporting the claim. What is the issue?
   • The page is automatically false
   • Search ranking is impossible
   • The citation is weak evidence for that particular claim
   • The model must be retrained
7. How should a search system be evaluated?
   • Count how many results are shown
   • Measure whether users complete their task, not only clicks
   • Use the model’s confidence as the only metric
   • Check only the visual design
8. Why test queries with uncommon terms or recently changed documents?
   • They make semantic retrieval unnecessary
   • They guarantee perfect results
   • They test only the interface color
   • They can expose ranking or freshness failures hidden by common queries