Operate / Quality
How good is it, across everything?
Use shows one document at a time. Quality will show all of them.
PlannedRoadmap phase 8
What will be here
- Accuracy by document type and task, measured against your evaluation datasets.
- The confidence distribution, and how many pages went to a person.
- Metadata coverage and normalization, and the share of each tag.
- Where labelling would help most.
What already answers
- Every encode names each component that answered, in its trace. Nothing records runs or aggregates them yet, so there is no number to show.
Nothing on this page is sample data. The screen will show only what the API returns, once it can. Read the roadmap →