End-to-End Grade Reproducibility (EEGR)

GMI-EGD-007Evidence-Governed DataInfrastructure-Dependent

Definition

The share of claims in a re-run sample that receive the same grade when the full pipeline, including retrieval, extraction, and normalization, is executed again against the same corpus under the same parameter version.

The Measurement Gap

There is no metric that tells you whether re-running the same corpus end to end under the same parameter version reproduces the same grades, which means determinism is verified at the grading stage and asserted for a composite whose retrieval, extraction, and normalization are stochastic.

Formula

EEGR = G_identical / G_rerun , reported with the grade transition matrix

Proposed formula, pending validation pilots.

At a Glance

Unit
Ratio (0 to 1)
Trend Direction
Higher
Category
Grading Integrity and Determinism

Decision Supported

Whether the program's determinism claim can be made about the system or only about the grading engine.

← Consequence-Weighted Grade Index (CWGI)Extraction Accuracy Stratum Divergence (EASD) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment