3.4Standard
Compare models and prompts over time.
Use expert-checked expected answers.
Choose two practices.
× Incorrect
Version representative, boundary, harmful, and multilingual cases with evidence and criteria, and protect the set from access, change, and training contamination.
Detailed explanation
CorrectControl access, changes, approval, leakage, and contamination of the evaluation set
Evaluation trust is protected.
IncorrectMix expected answers into public training data
The evaluation is contaminated.
IncorrectDiscard the set after each model release
Regressions cannot be compared.
CorrectVersion representative, boundary, harmful, and multilingual cases with evidence and criteria
Comparisons are consistent.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 3.4の評価データ、回帰、モデル比較を確認する。
Official references
What to confirm
信頼できる固定評価セットでモデル・プロンプトの退行を検出できる。
Key points
- Golden set
- Contamination
- Version
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
Golden set
A reviewed collection of inputs, expected results, and evidence used as a quality reference.
Evaluation contamination
Evaluation information that enters training or tuning and makes performance look too high.
Question IDAWS-AIF-381
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