Reading results
What each label means, where the proof is, and how to see everything the models did.
Statement labels
Every checked statement carries a label in words.
| Label | Meaning |
|---|---|
| Confirmed | Checked by independent models against your sources, and the workspace has a signed-off calibration, so the error rate of Confirmed statements is measured. |
| Models agree | Backed by independent checks with a quote from your sources, but not certified: the workspace has no signed-off calibration yet. |
| Not settled | Backed, but the checks were split. Read the quote before relying on it. |
| Disputed | Checks mostly went against it. It is removed from the answer or flagged. |
| Not supported | No model could back it up with your sources. |
| Opinion | A judgement or recommendation, not a checkable fact. |
| Two models agree / Only one model | Quick check: both models said it, or only one did. |
Note: Confirmed is never shown without a valid calibration for that workspace, depth and model set. Without one, the strongest label is Models agree.
Quotes and sources
The exact words your answer rests on.
Select a statement to see the quote that backs it, with the document and page. Quotes are matched against your file exactly or nearly exactly; a statement whose quote does not match is not counted as supported.
Tip: With no files attached, statements rest on the models' general knowledge and are labelled accordingly.
Full details
Every model's answer, critiques and rankings.
Show full details on a result opens each model's own answer, how the models ranked each other, the fairness checks (no model judges its own claims) and the usage of the run.
The audit record
Proof of what was asked, answered and checked.
Every finished question gets a signed audit record (DAR) in a tamper-evident chain, kept for the retention your plan allows. Auditors can download and verify it.