Same-Artifact All-Reviewer Gate

These examples and illustrative results are independently authored teaching materials, not measured model results.

Use case

The teaching workflow requires two independent passes on one release note/rubric. A passes v1, B passes v2, which still contains an unsupported benchmark. Different-version passes cannot be combined.

Mechanism

Bind artifact/rubric and supply identical required inputs to required reviewers. Track completed/failed/unavailable separately. All valid same-candidate results must meet this workflow’s gate. Repair supported findings and perform bounded changed-scope checks; retain open items at budget expiry rather than missing reviews as success.

Bad example

Combine A/v1 PASS and B/v2 PASS as dual approval for v2, ignoring unsupported numbers/incomplete dimensions.

Good example

Fix teaching v2/rubric-r1 and independently inspect it twice with actual version/state records. Do not count v1. Repair unsupported benchmarks from evidence and recheck under the bounded protocol. Two qualifying v2 results satisfy the gate; failures/missing output remain open. Gate passage is not publication permission.

Why the change matters

Verdicts bind inputs/criteria. Mixing versions or missing dimensions creates success for an artifact never actually inspected.

Observable expectation

A/v1 and B/v2 cannot satisfy v2; valid A/v2/B/v2 can. Missing dimensions are incomplete, not empty findings equated with pass.

Confirm specific repairs affirmatively rather than absence of repeated complaints. No review/release is run.

Limits

AND applies when explicitly required, not universally. Reviewers may share errors and need verification. Frozen loop counts/ship labels grant no authority.