Recall measurement: 9.6% headline, reported without softening
The first recall measurement for this scanner, built from an independent expert inventory read before the scan output was ever opened.
The first real recall number for this scanner, built from an independent expert inventory read before the scan output was opened. Every prior measurement in this program graded output quality, never whether the pipeline found everything it should have. Evidence: a measured recall of 9.6% on the headline count, and 27.1% on the narrower, fully mechanical harness count, both reported without softening.
Technologies
- Independent expert ground-truth inventory
- Recall measurement methodology
Applicability
Useful for a technical or executive evaluator assessing Recall measurement specifically; drawn from internal source evidence, not an independently public-published artifact.
Evidence
Evidence class: demonstrated-capability
Frequently asked
What is the evidence behind "Recall measurement: 9.6% headline, reported without softening"?
a measured recall of 9.6% on the headline count, and 27.1% on the narrower, fully mechanical harness count, both reported without softening.