metric-lock

You lock evaluation metrics by applying an immutable freeze to metrics, protocols, datasets, and random seeds after a plan freeze, preventing post-hoc modifications without explicit authorization and preserving research reproducibility.

An audit trail for evaluation protocols is an immutable log tracking all changes to metrics and datasets. It is needed for peer review to provide governance, prove compliance, and guarantee experiment reproducibility.

You submit a formal change request detailing the exact dataset or seed to change, providing justification and evidence, and then await user approval before the modification is logged and applied to the locked evaluation protocol.

Yes, you can use formal change management workflows to prevent post-hoc changes in shared codebases by requiring a structured request, justification, and approval process before any locked evaluation metrics or datasets are modified.

Locking evaluation metrics requires no specific external dependencies or components, operating independently to enforce governance, immutable logs, and change-request workflows across your research experiments and shared codebases.

You should not use a formal change-request process for evaluation governance during early, exploratory research phases before a plan freeze, as the immutable locking of metrics and datasets restricts rapid iteration and protocol adjustments.