Work · Evidence record
What Should an AI Applicant Shortlist Get Right: Four or More Checks Passed After One Correction
The completed WFT-003 synthetic field test reached 8/10 after one failure-only correction: 4 of five static checks passed for Candidate Screening, while 1 check remained unresolved.
- Exact prompts and outputs
- One correction only
- Synthetic inputs disclosed
01 · The assignment
The task
shortlist applicants from a structured hiring rubric
02 · Scope before score
Test disclosures
Input disclosure
Synthetic blind-test input pack WFT-003: A hiring team will provide anonymized applications and preapproved criteria for one open role. Source facts: six fictional records WFT-003-C01 through WFT-003-C06; policy rules P1–P5; scores 31, 39, 53, 62, 74, and 91; a shared surname on C02/C05; and an incomplete identifier on WFT-003-C04. Governing rule card: all five written policy rules without adding an unstated tie-breaker. Apply the written rules in order; use descending disclosed scores only where ranking is requested; a shared name alone never proves identity; incomplete identifiers require abstention rather than invention. 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-003 for “shortlist applicants from a structured hiring rubric” in a Codex multi-agent session. We froze the first response, returned only its 2 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-003. Task: shortlist applicants from a structured hiring rubric. Context: A hiring team will provide anonymized applications and preapproved criteria for one open role. Fictional source facts: six fictional records WFT-003-C01 through WFT-003-C06; policy rules P1–P5; scores 31, 39, 53, 62, 74, and 91; a shared surname on C02/C05; and an incomplete identifier on WFT-003-C04. Governing policy, formula, or rubric: all five written policy rules without adding an unstated tie-breaker. Apply the written rules in order; use descending disclosed scores only where ranking is requested; a shared name alone never proves identity; incomplete identifiers require abstention rather than invention. Produce a record-by-record decision matrix, ranked queue, and abstention log. 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: A candidate matrix and a blind reviewer cross-check will verify how each criterion was applied.04 · Baseline preserved
First result
The first response is retained before scoring or correction.
Frozen first response WFT-003 produced a record-by-record decision matrix, ranked queue, and abstention log for the task “shortlist applicants from a structured hiring rubric.” It treated the supplied pack as fictional and proposed this central handling: rank WFT-003-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-003-C04 until its identifier can be resolved. Concrete saved artifact row WFT-003-ROW1 reads: “WFT-003-C01 | rank WFT-003-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-003-C04 until its identifier can be resolved | 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 Candidate Screening task fidelity [WFT-003], Candidate Screening rule accuracy [WFT-003], and Candidate Screening handoff usability [WFT-003]. The audit found concrete failures: for Candidate Screening exception handling [WFT-003], the saved draft did not resolve or clearly preserve the shared-name nonmatch C02/C05 and incomplete record WFT-003-C04; for Candidate Screening source traceability [WFT-003], the saved draft gave the central WFT-003-C04 decision no source-to-output locator. 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: 6/10
05 · One pass only
Exact correction prompt
Only this single correction was allowed; there was no second repair pass.
Correct only these detected WFT-003 first-draft failures, using no new input or goal: 1) Candidate Screening exception handling [WFT-003] — the draft did not resolve or clearly preserve the shared-name nonmatch C02/C05 and incomplete record WFT-003-C04; 2) Candidate Screening source traceability [WFT-003] — the draft gave the central WFT-003-C04 decision no source-to-output locator.06 · Corrected output
Corrected final result
Corrected response WFT-003 retained the original fictional inputs, task boundary, and central decision: rank WFT-003-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-003-C04 until its identifier can be resolved. Concrete corrected artifact row WFT-003-ROW1 reads: “WFT-003-C01 | rank WFT-003-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-003-C04 until its identifier can be resolved | evidence locator: WFT-003-C01 | static status: 8/10 after one correction.” It changed only the detected failure areas, adding inspectable rows for Candidate Screening exception handling [WFT-003]. The frozen final text passed Candidate Screening task fidelity [WFT-003], Candidate Screening rule accuracy [WFT-003], Candidate Screening exception handling [WFT-003], and Candidate Screening handoff usability [WFT-003] and still failed Candidate Screening source traceability [WFT-003]. The final record-by-record decision matrix, ranked queue, and abstention log therefore earned 8/10 from 4 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: 8/10
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 |
|---|---|---|---|
| Candidate Screening task fidelity [WFT-003] | Pass | Pass | WFT-003 static check 1 inspected the saved wording for “Candidate Screening task fidelity [WFT-003].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-003-C04, the declared Candidate Screening rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Candidate Screening rule accuracy [WFT-003] | Pass | Pass | WFT-003 static check 2 inspected the saved wording for “Candidate Screening rule accuracy [WFT-003].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-003-C04, the declared Candidate Screening rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Candidate Screening exception handling [WFT-003] | Fail | Pass | WFT-003 static check 3 inspected the saved wording for “Candidate Screening exception handling [WFT-003].” The first transcript did not contain enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-003-C04, the declared Candidate Screening rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Candidate Screening source traceability [WFT-003] | Fail | Fail | WFT-003 static check 4 inspected the saved wording for “Candidate Screening source traceability [WFT-003].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture WFT-003-C04, the declared Candidate Screening rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Candidate Screening handoff usability [WFT-003] | Pass | Pass | WFT-003 static check 5 inspected the saved wording for “Candidate Screening handoff usability [WFT-003].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-003-C04, the declared Candidate Screening 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-003 kept “shortlist applicants from a structured hiring rubric” bounded to disclosed fictional inputs and preserved an auditable first-response snapshot.
- WFT-003 made the central handling—rank WFT-003-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-003-C04 until its identifier can be resolved—inspectable rather than implying unseen work.
- WFT-003 earned final passes for Candidate Screening task fidelity [WFT-003] and Candidate Screening rule accuracy [WFT-003] under the same frozen scoring rules.
What failed or remained weak
- WFT-003 still lacked enough saved-text evidence for Candidate Screening source traceability [WFT-003]; the record leaves that final failure visible.
09 · Inspectable record
Evidence notes
A candidate matrix and a blind reviewer cross-check will verify how each criterion was applied.
- WFT-003 preserves the exact synthetic prompt, frozen first-response account, failure-only correction, corrected-response account, and five boolean decisions together.
- WFT-003 scores are arithmetic: 3 first-pass checks × 2 = 6/10; 4 final-pass checks × 2 = 8/10.
- WFT-003 evaluated only the text/static portion of the declared evidence plan—A candidate matrix and a blind reviewer cross-check will verify how each criterion was applied.—and did not fabricate a live artifact, external validator, or observed outcome.
10 · Boundary of the claim
Limitations
- WFT-003 is a synthetic benchmark, so its worked verdict measures fit to the disclosed fictional Candidate Screening fixtures rather than effectiveness in a real workplace or learning setting.
- WFT-003 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.