Configure Run Metadata
Preserve the candidate and execution context needed to interpret a Run.
Prerequisites
The intended Benchmark Version and the managed or external execution workflow you will use.
Steps
- Identify the candidate Version or external candidate context before submission.
- Supply metadata supported by the chosen execution or import contract. Keep application labels distinct from Case input, scoring configuration, and immutable version identities.
- Inspect the accepted Run details and its recorded execution conditions.
- Use those records when comparing results so a descriptive label is not mistaken for proof that candidate or experiment conditions matched.
Object and state changes
Run metadata describes the recorded operation. It does not change the selected Case population, activate a Harness, or rewrite accepted results.
Success criteria
A teammate can identify what ran, against which Benchmark Version, and under which recorded conditions.
Common failure modes
Using a friendly candidate name as its only identity; assuming metadata overrides the actual submitted candidate or input.