The model comparison places competing explanations of the manuscript side by side, measured with the same statistics. The checks after it test the Hebrew reading model against the manuscript, control corpora and its own grammar; the protocols run client-side and are reproducible with a fixed seed. “Target met” there means internal consistency, not that a reading beats the controls. That question is answered by the model comparison.
2 Target met2 Warning zone
Status legend
- Target met
- Internal criterion met. Says nothing about whether a reading beats the controls.
- Warning zone
- Between target and critical threshold. Worth watching.
- Critical
- Critical threshold reached. Finding needs review.
- —
- Reference data or a candidate count, no pass/fail criterion.