Metrics: recall, groundedness, refusals
On This Lesson 4 topics
This lesson is planned. The outline below lists the topics it will cover. The full lesson has not been published yet.
In this lesson
measure the parts of a retrieval-based assistant.
Planned topics
- Retrieval recall: did the retrieved passages include the expected source? (Measure before blaming the model.)
- Groundedness (faithfulness): is every claim in the answer supported by the retrieved sources?
- Refusal correctness: refuses when the sources do not answer, answers when they do.
- Also track format validity and latency per step.
- Report per category so a weak area is visible.
Draft outline details
Demonstration
table: metric | how measured | threshold (for example recall at least 0.9, groundedness at least 0.9, refusal correctness at least 0.95).
Check your understanding
If recall is low, where do you look first?
Show answer
Retrieval: chunking, embeddings, search settings.
Practice
Apply this lesson with the module's lab, exercises and quiz.
Marking a lesson complete is optional and you can undo it any time. Your progress stays in this browser.