measurement, published

We publish how the numbers were made.

The 92.5% and 96.5% on our product pages come from two internal measurements, July 2026. This page shows what was measured, how, what the pipeline missed — enumerated one by one — and where the measurement itself falls short.

92.5%
requirement-extraction recall — full golden set, 173 items
96.5%
machine-certain precision — after adjudicating every flagged case
18 / 19
defects caught in the adversarial judgment benchmark
100%
run-to-run stability of the final pass/block posture (15 packages × 2 runs)
The golden set

Two real, published negotiated-contract service tenders (8 documents, table-preserving parse). An adjudicator authored the full requirement set — 85 and 88 items, 41 of them critical — while barred from seeing any pipeline output, so the answer key could not be shaped by what the pipeline happens to extract.

Every item carries a verbatim anchor into the source text, and all 173 anchors pass machine verification. Table rows are included as items — tables are where Korean tender documents put the substance.

Measuring extraction

A single extraction pass scores 82.1% recall against the golden set. Adding one critic pass — a second call that only hunts for requirements the first pass missed — lifts it to 92.5%, beating an equal-cost union of two independent passes (88.4%). That is the configuration we ship: one extra LLM call per document.

The critic over-extracts by design; the deterministic grounding layer absorbs it — every added item, grounded rate 100%.

Measuring judgment

Real bid submissions are exempt from disclosure in Korea, so they cannot be obtained. Instead: for three real published tenders (engineering design, solar installation works, security services — HWP and PDF), we generated fully compliant bid packages, then injected 45 defects with seeded, fixed mutations — so every defect label is certain — plus 15 harmless-rewording traps.

Machine-certain precision: 96.5% after adjudicating every flagged case. The deterministic layer caught every missing document, every identifier tamper, and every total-amount tamper — without LLM involvement.

What it missed

Every miss was adjudicated individually — none are mysteries. Of 45 injected defects, the judgment layer missed 9:

Also disclosed: 2 of the 15 harmless-rewording traps drew a “partial” flag (false positives), and the extraction residue clusters in edge cases such as authoring instructions inside a proposal’s table-of-contents table.

Stability

All 15 packages were verified twice, independently. The final pass/block posture agreed 15/15. What flickers between runs is confined to borderline grades — “is this commitment strong enough” — never the defect-anchored findings (missing documents, identifier and total mismatches, dropped requirements).

Limits of the measurement

The reports the pipeline produces are public too — read one before sending anything.