3.3Standard
Compare two model responses.
Reduce rater bias and record why a response was preferred.
Choose two appropriate practices.
× Incorrect
Use the same inputs, anonymize model identity and order, define criteria with examples, and record reasons and conditions.
Detailed explanation
CorrectProvide criteria and examples for correctness, relevance, safety, and style
Raters can apply a more consistent standard.
IncorrectStore only the winner and no rationale
Results cannot be reproduced or improved.
CorrectAnonymize responses and compare them for the same input
This reduces brand and ordering bias.
IncorrectTell raters model names and prices and let preference decide
Extraneous information biases quality judgments.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01ガイドDomain 3.3の人手評価と生成AI品質評価を確認する。
Official references
What to confirm
人手比較評価のバイアス抑制と再現性向上を説明できる。
Key points
- Anonymization
- Criteria
- Agreement
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
Pairwise evaluation
A human comparison of two outputs for the same input against defined criteria.
Evaluator agreement
The degree to which independent reviewers give consistent judgments.
Question IDAWS-AIF-145
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