Comparative model evaluation

Evaluate candidate models with the same immutable dataset and policy.

freeze evaluation setevaluate model Aevaluate model Bevaluate model Ccompare metricsapprove candidate

Activity responsibilities

Activity	Responsibility
`freeze-evaluation-set`	Execute the freeze evaluation set stage and publish its declared outputs for downstream activities.
`evaluate-model-a`	Execute the evaluate model A stage and publish its declared outputs for downstream activities.
`evaluate-model-b`	Execute the evaluate model B stage and publish its declared outputs for downstream activities.
`evaluate-model-c`	Execute the evaluate model C stage and publish its declared outputs for downstream activities.
`compare-metrics`	Execute the compare metrics stage and publish its declared outputs for downstream activities.
`approve-candidate`	Execute the approve candidate stage and publish its declared outputs for downstream activities.

Inputs

Candidate models
Evaluation set
Acceptance policy

Outputs

Metric reports
Decision record
Approved candidate

Provenance to preserve

Model versions
Evaluation runtime
Approval rationale

Runnable example

This is the complete checked-in bundle for this pattern. Download the environment, scope, topology, workflow, input, container recipe, runner, and validation contract from this page before executing it.

Read setup instructions ↓Environment YAML ↓Execution scope YAML ↓Network topology YAML ↓Workflow YAML ↓Input data ↓Container recipe ↓Run the workflow ↓Validate outputs ↓Expected output manifest ↓

Verified local execution

These captures and the output manifest were produced by the fixture's local Docker run and validator. They are published with the same bundle as the runnable files.

Comparative model evaluation workflow execution evidence — Workflow evidence

Comparative model evaluation execution evidence — Execution evidence

Comparative model evaluation output evidence — Output evidence

Open verified output manifest ↓

Execution considerations

Use comparable resources when latency is part of the decision. Keep quality and execution performance as separate measurements.

AkôFlow boundary: the engine schedules, deploys, executes, transfers data, and records evidence. The ML or agent framework remains an implementation choice inside each activity.