Determine whether calibration and review budgets are stable across Jev releases

Determine whether the calibration performance and human-review budgets reported for Jev are properties of the Jev model and coding schema generally or only of the audited Jev release.

Background

The paper emphasizes that Jev is a closed commercial model whose behavior may change without notice. Although the study records a pinned model identifier and per-call identifiers, it cannot prevent vendor-side changes or guarantee reproducibility if a release is withdrawn. A repeated audit using the same reference set and metric code is therefore needed to establish whether the reported calibration and review-budget results persist across model versions.