-
Notifications
You must be signed in to change notification settings - Fork 0
Evaluation
Dev Doshi edited this page Jun 18, 2026
·
1 revision
VaultMind uses RAGAS for retrieval and answer quality evaluation.
Custom benchmark dataset.
Contains:
- 14 Ground Truth Questions
- Expected Answers
Measures whether the answer is grounded in retrieved context.
Measures whether the answer addresses the user's question.
Measures retrieval quality.
Measures retrieval completeness.
| Metric | Score |
|---|---|
| Faithfulness | 0.993 |
| Answer Relevancy | 0.906 |
| Context Precision | 0.738 |
| Context Recall | 0.821 |
| Overall | 0.865 |
Maintain:
Overall Score > 0.85