Scoring & evidence
Metric
A defined quantity used to assess an aspect of performance.
Definition
A metric converts observed system behaviour into a number, category or pass/fail result, such as accuracy, latency or citation correctness.
Why it matters
A useful metric must align with the real decision and penalise failures in proportion to their business impact.
Related concepts
- Rubric
Explicit criteria describing what a good answer or deliverable must contain.
- Exact match
Whether an output exactly equals the reference answer after normalisation.
- F1 score
A combined measure of precision and recall.
- Critical-error rate
The share of cases containing a failure with severe consequences.