Circulatory Fidelity: Quantifying Structural Coupling to Diagnose Mean-Field Failure in Hierarchical Models

Manuscript PDF: https://zenodo.org/records/18121821/files/Circulatory_Fidelity_Manuscript.pdf?download=1 Initial public release accompanying the preprint submission of "Circulatory Fidelity: Quantifying Structural Coupling to Diagnose Mean-Field Failure in Hierarchical Models." This release includes: Complete manuscript (LaTeX source and compiled PDF) Reference implementations in Python and Julia Validation datasets (32,000+ simulations) Jupyter notebooks reproducing all figures and statistical analyses CF is a pre-inference diagnostic for mean-field variational inference, computable from prior predictive samples before posterior inference begins. Key findings include the Proximal Dominance Principle (distal coupling causes exactly 1.0× degradation) and strong predictive correlation with inference quality (r = 0.86 for log-likelihood gap). Manuscript under review. Feedback welcome via GitHub Issues.