Work · Evidence record
What Evidence Gaps Can AI Find in an Accessibility Questionnaire — Completed Benchmark Result: 6/10
The completed WFT-020 synthetic field test stopped at 6/10: three of five Accessibility Review checks passed after one correction, but Accessibility Review rule accuracy [WFT-020] and Accessibility Review exception handling [WFT-020] remained unsupported.
- Exact prompts and outputs
- One correction only
- Synthetic inputs disclosed
01 · The assignment
The task
check a vendor accessibility questionnaire for missing evidence
02 · Scope before score
Test disclosures
Input disclosure
Synthetic blind-test input pack WFT-020: A compliance team will provide a completed questionnaire, supporting documents, and a mandatory evidence checklist. Source facts: controlled excerpts WFT-020-D01 through WFT-020-D05; clauses 2.1, 3.4, 6.2, and 8.7; effective dates 2026-09-01 and 2026-10-15; one defined-term conflict in WFT-020-D04; and a mandatory exception in clause 6.2. Governing rule card: the effective dates and the distinction between mandatory and optional language. Mandatory clauses override optional language; retain explicit exceptions; cite the controlling section for every output row; unresolved definitions or missing evidence stay flagged. The pack deliberately contains no answer key, target ranking, expected classification, expected calculation result, or row-level decision. Every person, organization, record, statement, file, and identifier is fictional; no private, production, or live data was supplied.
Run disclosure
We ran one text-only synthetic benchmark WFT-020 for “check a vendor accessibility questionnaire for missing evidence” in a Codex multi-agent session. We froze the first response, returned only its 3 detected check failures, accepted one corrected response, and scored both against the same five static checks. No external or live action occurred: nothing was sent, published, deployed, uploaded, submitted, purchased, booked, contacted, emailed, called, messaged, executed, or changed outside the fictional text fixtures. No external, live, or production action occurred. Per-case elapsed time was not instrumented during the batch session.
- Evidence mode
- Synthetic benchmark
- Run environment
- Codex multi-agent session
- Model disclosure
- Exact underlying model identifier not disclosed by the Codex session
03 · Verbatim input
Exact first prompt
The recorded session received the following prompt without silent additions.
Run bounded synthetic field test WFT-020. Task: check a vendor accessibility questionnaire for missing evidence. Context: A compliance team will provide a completed questionnaire, supporting documents, and a mandatory evidence checklist. Fictional source facts: controlled excerpts WFT-020-D01 through WFT-020-D05; clauses 2.1, 3.4, 6.2, and 8.7; effective dates 2026-09-01 and 2026-10-15; one defined-term conflict in WFT-020-D04; and a mandatory exception in clause 6.2. Governing policy, formula, or rubric: the effective dates and the distinction between mandatory and optional language. Mandatory clauses override optional language; retain explicit exceptions; cite the controlling section for every output row; unresolved definitions or missing evidence stay flagged. Produce a clause matrix, proposed output, and unresolved-source register. Derive every row from the source facts and rule card; show calculations or criterion paths, cite supplied identifiers, state assumptions, surface any violated constraint, ambiguity, missing datum, or source conflict without presuming which row should pass, and abstain where evidence is incomplete. Do not infer an expected answer from the evidence plan and do not claim any live or external action. Planned verification after the frozen response: An omissions register and a reviewer comparison against required fields will verify each identified gap.04 · Baseline preserved
First result
The first response is retained before scoring or correction.
Frozen first response WFT-020 produced a clause matrix, proposed output, and unresolved-source register for the task “check a vendor accessibility questionnaire for missing evidence.” It treated the supplied pack as fictional and proposed this central handling: trace the proposed treatment to clauses 2.1 and 6.2, preserve the mandatory exception, and mark the definition conflict in WFT-020-D04 for review. Concrete saved artifact row WFT-020-ROW1 reads: “WFT-020-D01 | trace the proposed treatment to clauses 2.1 and 6.2, preserve the mandatory exception, and mark the definition conflict in WFT-020-D04 for review | source: supplied fictional pack | review state: first-draft.” The draft preserved the named records and separated supplied facts from assumptions; its inspectable work product also stated that no message, transaction, system change, or learner outcome had occurred. On the five predeclared static checks, it passed Accessibility Review source traceability [WFT-020] and Accessibility Review handoff usability [WFT-020]. The audit found concrete failures: for Accessibility Review task fidelity [WFT-020], the saved draft did not connect WFT-020-D04 to the full boundary of “check a vendor accessibility questionnaire for missing evidence”; for Accessibility Review rule accuracy [WFT-020], the saved draft left the effective dates and the distinction between mandatory and optional language without an explicit verification row; for Accessibility Review exception handling [WFT-020], the saved draft did not resolve or clearly preserve the clause-6.2 exception and conflicting definition in WFT-020-D04. Those were evidence defects in the saved response, not inferred real-world failures, and they became the complete boundary of the single correction pass.Initial score: 4/10
05 · One pass only
Exact correction prompt
Only this single correction was allowed; there was no second repair pass.
Correct only these detected WFT-020 first-draft failures, using no new input or goal: 1) Accessibility Review task fidelity [WFT-020] — the draft did not connect WFT-020-D04 to the full boundary of “check a vendor accessibility questionnaire for missing evidence”; 2) Accessibility Review rule accuracy [WFT-020] — the draft left the effective dates and the distinction between mandatory and optional language without an explicit verification row; 3) Accessibility Review exception handling [WFT-020] — the draft did not resolve or clearly preserve the clause-6.2 exception and conflicting definition in WFT-020-D04.06 · Corrected output
Corrected final result
Corrected response WFT-020 retained the original fictional inputs, task boundary, and central decision: trace the proposed treatment to clauses 2.1 and 6.2, preserve the mandatory exception, and mark the definition conflict in WFT-020-D04 for review. Concrete corrected artifact row WFT-020-ROW1 reads: “WFT-020-D01 | trace the proposed treatment to clauses 2.1 and 6.2, preserve the mandatory exception, and mark the definition conflict in WFT-020-D04 for review | evidence locator: WFT-020-D01 | static status: 6/10 after one correction.” It changed only the detected failure areas, adding inspectable rows for Accessibility Review task fidelity [WFT-020]. The frozen final text passed Accessibility Review task fidelity [WFT-020], Accessibility Review source traceability [WFT-020], and Accessibility Review handoff usability [WFT-020] and still failed Accessibility Review rule accuracy [WFT-020] and Accessibility Review exception handling [WFT-020]. The final clause matrix, proposed output, and unresolved-source register therefore earned 6/10 from 3 passing checks; no second repair was attempted. This result reports only a bounded transcript-and-fixture evaluation. It does not claim that any workplace process, learner performance, account, device, service, or external system was actually changed, contacted, tested live, or improved.Final score: 6/10
07 · Five checks, two points each
Five-check record
The first and final statuses are textual as well as color coded. Each final pass is worth two points; the displayed verdict is tied to the final total.
| Check | First | Final | Evidence |
|---|---|---|---|
| Accessibility Review task fidelity [WFT-020] | Fail | Pass | WFT-020 static check 1 inspected the saved wording for “Accessibility Review task fidelity [WFT-020].” The first transcript did not contain enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-020-D04, the declared Accessibility Review rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Accessibility Review rule accuracy [WFT-020] | Fail | Fail | WFT-020 static check 2 inspected the saved wording for “Accessibility Review rule accuracy [WFT-020].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture WFT-020-D04, the declared Accessibility Review rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Accessibility Review exception handling [WFT-020] | Fail | Fail | WFT-020 static check 3 inspected the saved wording for “Accessibility Review exception handling [WFT-020].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture WFT-020-D04, the declared Accessibility Review rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Accessibility Review source traceability [WFT-020] | Pass | Pass | WFT-020 static check 4 inspected the saved wording for “Accessibility Review source traceability [WFT-020].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-020-D04, the declared Accessibility Review rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Accessibility Review handoff usability [WFT-020] | Pass | Pass | WFT-020 static check 5 inspected the saved wording for “Accessibility Review handoff usability [WFT-020].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-020-D04, the declared Accessibility Review rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
08 · No cleanup by omission
What worked—and what failed
What worked
- WFT-020 kept “check a vendor accessibility questionnaire for missing evidence” bounded to disclosed fictional inputs and preserved an auditable first-response snapshot.
- WFT-020 made the central handling—trace the proposed treatment to clauses 2.1 and 6.2, preserve the mandatory exception, and mark the definition conflict in WFT-020-D04 for review—inspectable rather than implying unseen work.
- WFT-020 earned final passes for Accessibility Review task fidelity [WFT-020] and Accessibility Review source traceability [WFT-020] under the same frozen scoring rules.
What failed or remained weak
- WFT-020 still lacked enough saved-text evidence for Accessibility Review rule accuracy [WFT-020]; the record leaves that final failure visible.
- WFT-020 still lacked enough saved-text evidence for Accessibility Review exception handling [WFT-020]; the record leaves that final failure visible.
09 · Inspectable record
Evidence notes
An omissions register and a reviewer comparison against required fields will verify each identified gap.
- WFT-020 preserves the exact synthetic prompt, frozen first-response account, failure-only correction, corrected-response account, and five boolean decisions together.
- WFT-020 scores are arithmetic: 2 first-pass checks × 2 = 4/10; 3 final-pass checks × 2 = 6/10.
- WFT-020 evaluated only the text/static portion of the declared evidence plan—An omissions register and a reviewer comparison against required fields will verify each identified gap.—and did not fabricate a live artifact, external validator, or observed outcome.
10 · Boundary of the claim
Limitations
- WFT-020 is a synthetic benchmark, so its mixed verdict measures fit to the disclosed fictional Accessibility Review fixtures rather than effectiveness in a real workplace or learning setting.
- WFT-020 used one text-only Codex multi-agent session whose underlying model identifier was not disclosed; different prompts, models, fixtures, or human reviewers could produce different results.