The Avenridge Evaluation and Publication Framework: Locked, Reproducible Evaluation with Calibration Assessment and Null-Result Parity

We introduce the Avenridge Evaluation and Publication Framework (AEPF), an operationalized methodology for evaluation-intensive computational domains. AEPF integrates established disciplines—clinical-trial pre-registration, high-energy-physics blind analysis, machine-learning calibration assessment, and open-science null reporting—into a single workflow with concrete operational consequences: cryptographic anchoring of methodology to a public commit chain, calibration assessment as a standing pre-registration requirement, null-result publication parity, constrained-interpretation reporting, and a pre-specification viability gate that retires structurally over-determined studies. We illustrate the framework with four case studies—spanning cosmological residual analysis, transformer calibration auditing, framing-robustness testing of an open-weight language model, and a viability-gated disposition—together with a methods-level validation of the calibration-audit instrument against synthetic ground truth. Within the parametric miscalibration families tested, the validation recorded zero false-acceptance of broken calibration, supporting the instrument's soundness on those families; its overall pre-committed bar nonetheless returned FAIL, driven by reduced statistical power under a heavily skewed but genuinely calibrated distribution rather than by any soundness failure. We report that outcome as-is and treat the soundness claim as bounded to the families validated, so that a verdict reported under AEPF reflects the evaluated system rather than an artifact of the tool within that scope. The novelty claimed is in the integration and artifact-level operationalization, not in any individual component.