3.3Standard
Two summaries express the same meaning with different wording.
Do not rely on automatic matching alone.
Choose two practices.
× Incorrect
Use string metrics such as ROUGE for regression comparisons but assess factuality, omission, and safety with human or additional evaluation.
Detailed explanation
IncorrectIgnore severe examples because the average is high
Serious individual failures can be hidden.
CorrectUse ROUGE or similar metrics for regression and separate factuality evaluation
Large comparisons and semantic quality are separated.
CorrectReview representative examples for evidence, omissions, contradictions, and safety
Meaning not captured by overlap is checked.
IncorrectWord overlap guarantees factuality
Matching words can still form false claims.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 3.3の生成AI評価、品質指標、人手評価を確認する。
Official references
What to confirm
文字列一致と意味・事実性の違いを説明できる。
Key points
- ROUGE
- Factuality
- Human review
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
ROUGE
A text-overlap metric that compares generated and reference word sequences or n-grams.
Factuality
The degree to which generated statements agree with reliable evidence or real-world facts.
Question IDAWS-AIF-283
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