Audit label recognition before blaming OCR digit quality

When a label-aware OCR pipeline recognizes identifier digits but misreads the field label or separates labels and values across lines, a downstream selector constrained to labeled candidates may correctly abstain even though a person can read the number. Diagnose recognition, label normalization, layout association, candidate generation, and selection as separate stages. Persist raw OCR, candidate evidence strength, resolution method, and per-field selector choice. Do not broadly relax digit grounding; extend label-confusion and layout rules only against labeled evaluations because permissive matching can silently swap fields.

Applies when: OCR extracts regulated identifiers using label-bound candidates and a model selector. Keywords: OCR label recognition candidate parser layout association model abstention