Input
What you provide
- Named model candidates and versions
- Locked rubric, thresholds, data slices, and measurement version
- Known regressions, release policy, approvers, and rollback owner
Models / Model Launch Check
Compare model candidates under the same checks, replay known failures, identify blockers, collect approvals, and produce a source-linked launch file.
Input
Work
Output
A stage advances only when its required evidence and owner are visible.
01
Confirm candidate versions, owners, intended release, and evaluation context.
Candidate manifest
02
Lock the rubric and source slice before scores or rank are displayed.
Comparison matrix and source runs
03
Replay retained failures and keep each failed case visible.
Regression outcomes and affected versions
04
Name every failed, missing, expired, or permission-restricted requirement.
Blocker owner and recovery
05
Show who can approve, who has approved, and what remains.
Approval chain and timestamps
06
Export the decision, evidence, destination, checksum, and rollback.
Signed release packet
Evidence packet preview
| Outcome | Work | Evidence | Boundary |
|---|---|---|---|
| Ship | All required checks pass, approvals are current, and rollback is present. | Source-linked results, approvers, launch manifest, and generated time. | Approval applies only to the named candidate, environment, checks, and release scope. |
| Hold | A required check fails, expires, or lacks sufficient evidence. | Blocking criterion, failed source run, consequence, owner, and next action. | Hold remains active until the required evidence is regenerated and approved. |
| Return for work | The candidate needs remediation, new evaluation, or a corrected release record. | Returned items, responsible owner, expected evidence, and review route. | The workflow does not imply a fix; it records the required work and gate. |
Inspectable evidence
The example uses descriptive states because final scores must come from the customer's actual source runs.
Representative product state
The states below illustrate the interaction contract. Customer decisions use current source runs, named owners, and actual approvals.
| Required check | Candidate A | Candidate B |
|---|---|---|
| Review required | Evidence ready | |
| Blocking issue | Evidence ready | |
| Review required | Review required | |
| Evidence ready | Evidence ready |
Same checksCandidates share the same rubric and measurement version.
Blocking evidenceA failed regression remains visible and source-linked.
Release effectEvery difference states how it changes the decision.