a36956b1e1
The truth files used `determinableFromResponseDocument`; the schema column
and gateEvaluableFromResponse() use `determinableFromResponse`. A silent
name drift means the flag reads as its default (true), and R-B6C — "two bid
copies submitted", which no response document can ever establish — starts
failing bidders on a packaging detail.
Renamed in rfp.truth.json and validate-responses.py, and added the mirror
check to validate-truth.py so the drift cannot recur unnoticed:
gates 13, evaluable from the response 12, excluded 1
excluded: R-B6C Two bid copies submitted.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Corpus
Source documents for the synthesized response corpus (#30). Archived 2026-08-03 because #22 read them from live URLs and kept nothing — and its ninth hazard was that live procurement PDFs 403 or return bot-challenge HTML without warning.
All seven retrieved successfully; none of the 403s #22 hit recurred.
Authoring anchors — solicitations we can write responses to
| Path | Pages | Why |
|---|---|---|
friendship-pcs/rfp.pdf |
52 | Grocery, nested rubric, closest to the retail domain. Its price sheet is a separate Excel attachment, which exercises #22's second extraction hazard. |
slc-airport/rfp.pdf |
142 | Richest binding structure — paragraph headings A–M restated verbatim. Has a shortlist-for-interviews stage that exercises the finalist release level. |
Validation artifact — not a solicitation
tarrant-county/rfp.pdf (1pp) is an award tabulation, not an RFP. No
requirements, so nothing can be written against it. What it holds is the only
real published outcome in the set:
Evaluation Criteria Max Five Star Selrico Trinity
Management & Support Staff 25 15.84 15.00 16.36
Meal and Service Plan 25 15.20 16.30 17.30
Financial Stability 20 12.00 12.00 12.00
Meal Price 25 25.00 24.52 22.23
References 5 3.77 3.07 2.73
Total Score 100 71.81 70.89 70.62
PRIMARY SECONDARY ALTERNATE
Three vendors inside 1.19 points. This is what ranking arithmetic gets checked against — the only place in the corpus where our maths can be wrong in a way that is detectable rather than self-consistent. It is also concrete evidence for storing fractional scores: rounding erases this result entirely.
Reference — rubric and pricing shapes, not authored against
| Path | Pages | Why |
|---|---|---|
reference/salt-lake-county-scoring-criteria.pdf |
3 | A second 0–5 scale with incompatible labels — the evidence that scales are per-document. |
reference/nevada-asd-401-eval-worksheet.pdf |
1 | The generic default rubric procurement offices reach for. |
reference/maryland-dgs-mbe-forms.pdf |
27 | Nested percentage compliance — a constraint program, not a field. |
reference/arkansas-unit-price-bid-sheet.pdf |
4 | Rate-function pricing: unit × frequency × quantity. |
Corrections to #22's record
- SLC Airport is 142pp, not 55pp.
- Maryland MBE forms are 27pp, not 20pp.
- Tarrant County was described as a document with a score tabulation; it is only a tabulation, which changes its role in the corpus.