Work · Evidence record
Does AI Spot Duplicate Customers Without Merging Different People: The One-Pass Revision Reached 8/10
The completed WFT-065 synthetic field test reached 8/10 after one failure-only correction: 4 of five static checks passed for Entity Resolution, while 1 check remained unresolved.
- Exact prompts and outputs
- One correction only
- Synthetic inputs disclosed
01 · The assignment
The task
detect duplicate customer records without merging distinct people
02 · Scope before score
Test disclosures
Input disclosure
Synthetic blind-test input pack WFT-065: A data team will provide fictional records containing name variants, shared households, reused phone numbers, typos, and deliberate lookalikes. Source facts: six fictional records WFT-065-C01 through WFT-065-C06; policy rules P1–P5; scores 30, 38, 77, 62, 74, and 91; a shared surname on C02/C05; and an incomplete identifier on WFT-065-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-065 for “detect duplicate customer records without merging distinct people” 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-065. Task: detect duplicate customer records without merging distinct people. Context: A data team will provide fictional records containing name variants, shared households, reused phone numbers, typos, and deliberate lookalikes. Fictional source facts: six fictional records WFT-065-C01 through WFT-065-C06; policy rules P1–P5; scores 30, 38, 77, 62, 74, and 91; a shared surname on C02/C05; and an incomplete identifier on WFT-065-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 hidden match key and pairwise precision-recall report will verify merges, misses, and protected nonmatches.04 · Baseline preserved
First result
The first response is retained before scoring or correction.
Frozen first response WFT-065 produced a record-by-record decision matrix, ranked queue, and abstention log for the task “detect duplicate customer records without merging distinct people.” It treated the supplied pack as fictional and proposed this central handling: rank WFT-065-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-065-C04 until its identifier can be resolved. Concrete saved artifact row WFT-065-ROW1 reads: “WFT-065-C01 | rank WFT-065-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-065-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 Entity Resolution task fidelity [WFT-065], Entity Resolution source traceability [WFT-065], and Entity Resolution handoff usability [WFT-065]. The audit found concrete failures: for Entity Resolution rule accuracy [WFT-065], the saved draft left all five written policy rules without adding an unstated tie-breaker without an explicit verification row; for Entity Resolution exception handling [WFT-065], the saved draft did not resolve or clearly preserve the shared-name nonmatch C02/C05 and incomplete record WFT-065-C04. 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-065 first-draft failures, using no new input or goal: 1) Entity Resolution rule accuracy [WFT-065] — the draft left all five written policy rules without adding an unstated tie-breaker without an explicit verification row; 2) Entity Resolution exception handling [WFT-065] — the draft did not resolve or clearly preserve the shared-name nonmatch C02/C05 and incomplete record WFT-065-C04.06 · Corrected output
Corrected final result
Corrected response WFT-065 retained the original fictional inputs, task boundary, and central decision: rank WFT-065-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-065-C04 until its identifier can be resolved. Concrete corrected artifact row WFT-065-ROW1 reads: “WFT-065-C01 | rank WFT-065-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-065-C04 until its identifier can be resolved | evidence locator: WFT-065-C01 | static status: 8/10 after one correction.” It changed only the detected failure areas, adding inspectable rows for Entity Resolution rule accuracy [WFT-065]. The frozen final text passed Entity Resolution task fidelity [WFT-065], Entity Resolution rule accuracy [WFT-065], Entity Resolution source traceability [WFT-065], and Entity Resolution handoff usability [WFT-065] and still failed Entity Resolution exception handling [WFT-065]. 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 |
|---|---|---|---|
| Entity Resolution task fidelity [WFT-065] | Pass | Pass | WFT-065 static check 1 inspected the saved wording for “Entity Resolution task fidelity [WFT-065].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-065-C04, the declared Entity Resolution rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Entity Resolution rule accuracy [WFT-065] | Fail | Pass | WFT-065 static check 2 inspected the saved wording for “Entity Resolution rule accuracy [WFT-065].” The first transcript did not contain enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-065-C04, the declared Entity Resolution rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Entity Resolution exception handling [WFT-065] | Fail | Fail | WFT-065 static check 3 inspected the saved wording for “Entity Resolution exception handling [WFT-065].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture WFT-065-C04, the declared Entity Resolution rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Entity Resolution source traceability [WFT-065] | Pass | Pass | WFT-065 static check 4 inspected the saved wording for “Entity Resolution source traceability [WFT-065].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-065-C04, the declared Entity Resolution rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Entity Resolution handoff usability [WFT-065] | Pass | Pass | WFT-065 static check 5 inspected the saved wording for “Entity Resolution handoff usability [WFT-065].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture WFT-065-C04, the declared Entity Resolution 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-065 kept “detect duplicate customer records without merging distinct people” bounded to disclosed fictional inputs and preserved an auditable first-response snapshot.
- WFT-065 made the central handling—rank WFT-065-C06 first under P2, keep C02 and C05 separate despite the shared surname, and abstain on WFT-065-C04 until its identifier can be resolved—inspectable rather than implying unseen work.
- WFT-065 earned final passes for Entity Resolution task fidelity [WFT-065] and Entity Resolution rule accuracy [WFT-065] under the same frozen scoring rules.
What failed or remained weak
- WFT-065 still lacked enough saved-text evidence for Entity Resolution exception handling [WFT-065]; the record leaves that final failure visible.
09 · Inspectable record
Evidence notes
A hidden match key and pairwise precision-recall report will verify merges, misses, and protected nonmatches.
- WFT-065 preserves the exact synthetic prompt, frozen first-response account, failure-only correction, corrected-response account, and five boolean decisions together.
- WFT-065 scores are arithmetic: 3 first-pass checks × 2 = 6/10; 4 final-pass checks × 2 = 8/10.
- WFT-065 evaluated only the text/static portion of the declared evidence plan—A hidden match key and pairwise precision-recall report will verify merges, misses, and protected nonmatches.—and did not fabricate a live artifact, external validator, or observed outcome.
10 · Boundary of the claim
Limitations
- WFT-065 is a synthetic benchmark, so its worked verdict measures fit to the disclosed fictional Entity Resolution fixtures rather than effectiveness in a real workplace or learning setting.
- WFT-065 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.