Compare generated summaries with reference summaries.
Use an automated metric for an initial evaluation.
Which metrics compare overlap with reference text?
ROUGE and BLEU compare generated text with reference text, but semantic quality and safety require additional evaluation.
Detailed explanation
A storage class describes storage characteristics, not model quality.
Policy versioning manages permissions rather than generated text quality.
These metrics use overlap with reference text for tasks such as summarization or translation.
CPU utilization describes resource use, not text quality.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01ガイドのDomain 3.4で、モデル評価指標と自動評価の限界を確認する。
Official references
What to confirm
ROUGE・BLEUを参照テキストとの比較に使う指標として説明できる。
Key points
- ROUGE
- BLEU
- Reference text
- Evaluation limits
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
Metric coverage
Overlap can miss valid paraphrases and cannot fully measure safety or factual correctness.
Composite evaluation
Combining automatic metrics, human review, and task tests gives broader evidence than one score.