READER MODELong-form reading surface.

§8.2 · BUILDING · 5 MIN READ · EDITION 1.0

Building a golden reference set

A defensible reference set spans clean, difficult, borderline, incomplete-evidence and no-finding cases, each with an agreed ground truth.

A set built only from easy cases overstates performance; a set with no no-finding cases cannot reveal a tool that over-generates findings.

Ground truth should be agreed by a competent reviewer independent of the tool being evaluated, and disagreements resolved before testing begins.

PUT THIS INTO PRACTICE

  1. 01Include no-finding cases deliberately.
  2. 02Agree ground truth before running the tool.
  3. 03Track case composition against the target mix.

RELATED INSTRUMENTS

CITE THIS SECTION

§8.2 · Edition 1.0 · Building a golden reference set

English is the authoritative text of this edition. Where a translation differs, the English text governs.