3.4Standard
Retrieval looks good but final answers are wrong.
Separate failure layers.
Choose two evaluations.
× Incorrect
Measure retrieval recall, ranking, authorization, evidence support, omissions, safety, latency, and cost on the same question set.
Detailed explanation
IncorrectUse final answer accuracy alone to identify retrieval causes
Layers cannot be separated.
CorrectLink retrieved evidence to final answers and record latency, cost, and refusals
End-to-end tradeoffs are comparable.
CorrectMeasure retrieval recall, ranking, authorization, and generation support separately
Failure layers are separated.
IncorrectSkip evidence checks when retrieval looks plausible
Generation errors remain.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 3.4のRAG評価、検索指標、生成品質を確認する。
Official references
What to confirm
検索と生成のどの段階が失敗したかを分離して改善できる。
Key points
- RAG evaluation
- Retrieval
- Generation
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Compare the definition with the AIF-C01 objective and the service documentation before applying it.
- Keep the concept separate from a particular answer choice or implementation step.
Foundation review
End-to-end evaluation
Assessment of a complete workflow from input and retrieval through final output.
Failure isolation
Separating pipeline stages to identify where an error was introduced.
Question IDAWS-AIF-443
EnvironmentAWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
Last technical review2026-08-20
Report an error or rights issue