5.1.8 · Lesson 8 of 14
Incorporate Human Review into Evaluation
Use human evaluation workflows for nuanced quality judgments that automated metrics miss.
WHAT
Incorporate human review into GenAI evaluation for nuanced quality and policy judgments.
HOW
- Bedrock Model Evaluation supports Human Evaluation with internal teams or AWS Managed Human Workers.
- Use humans for tone, policy nuance, and ambiguous groundedness cases.
Key points
- Human Evaluation in Bedrock Model Evaluation
- Internal reviewers or Managed Human Workers
- Nuanced tone/policy judgments
- Complement automated metrics
- Annotate edge cases
Exam focus
Human evaluation complements automated metrics for nuanced tasks — it is not replaced by ROUGE alone.