2.3Standard
Auditors need to compare repeated outputs.
Control generation variation.
Choose two designs.
× Incorrect
Record model version, prompt, sampling settings, input, and tool results, then run regression tests with a fixed evaluation set.
Detailed explanation
CorrectUse a fixed evaluation set and low variation to compare changes
Differences become measurable.
IncorrectKeep only an unreproducible success example
Audit comparison is impossible.
CorrectRecord model version, prompt, sampling settings, input, and tool results
Conditions can be reproduced.
IncorrectA model name alone is enough; provider version and settings need not be saved
Actual conditions can differ.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 2.3・3.4の生成再現性、評価、モデル版管理を確認する。
Official references
What to confirm
生成条件を保存して、変更前後の出力を再現・比較できる。
Key points
- Reproducibility
- Model version
- Regression
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
Generation reproducibility
The ability to repeat a generation under recorded conditions.
Sampling setting
A parameter controlling how output tokens are selected.
Question IDAWS-AIF-438
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