The Non-Separable Fiber Program — A Propensity Model of Quantum Correlations, the Three Screens, and the Case for Decomposition-Independence: Deposit v1

This record deposits, as one versioned set, three companion documents of the Non-Separable Fiber Program. 1. A Non-Separable Propensity Model of Quantum Correlations (v5.2). A constructive stochastic hidden-structure model of two-photon polarization correlations. The model distinguishes kinematic locality (no superluminal signaling) from statistical separability, and discards only the second: each entangled pair bears one token fiber state with two bearers — a definite point on the Poincaré sphere, an affine (Malus) propensity coupling, a measurement-triggered update rule, and, at the pair level, a presentation structure identifying how the point stamped at one wing is consulted at the other. The model reproduces the full quantum joint distribution for every two-qubit pure state exactly, in both actualization orders, with no-signaling resolved into two named, contingent conditions (the complementarity identity of the coupling; matched fiber equilibrium). Two theorems locate the quantum/super-quantum boundary — the same architecture reaches the Popescu–Rohrlich maximum under a threshold coupling, and the update dynamics identifies the quantum correlations with the harmonic purity of the local coupling — and the cos² coupling itself is derived from stated premises by non-contextual normalization, with Busch- and Wright–Weigert-type completions and a named residue. The presentation structure is identified explicitly as the quantum steering identity realized as token structure, and the paper states, in Bell/Jarrett vocabulary, exactly which locality conditions the model concedes at the hidden level and which it preserves operationally. 2. Three Screens: Why the Foliation Is Hidden (v1.4). The companion synthesis. On the paper's preferred (conditional) reading, the actualization order is supplied by a real foliation; this document explains that structure's empirical concealment as the joint consequence of three contingent, parameterized identities — exact Lorentz invariance (the Mansouri–Sexl α̂ bound), the complementarity identity (derived within the model), and matched fiber equilibrium — each with a named breaking parameter that experiment constrains, addressing the Balashov–Janssen "unexplained conspiracy" objection. 3. Why Decomposition-Independence: The Dispositional Case for P4-mix (v1.3). The companion essay on the derivation's load-bearing premise: its exact statement, the positive dispositional case, a stress-test against the strongest objections, and an analysis showing the premise is experimentally enforced for all realized devices within the model's propensity ontology, with the residual scope stated exactly. The three documents cross-reference by section and are deposited together as one record; this is version 1. The framework was developed with AI assistance under the disclosed-provenance practice described in the paper's Author's Note and Acknowledgments; its mathematics has been verified independently by multiple methods, including hostile cross-model review, as documented there. Full development transcripts, referenced in the Acknowledgments as the authoritative attribution record, are available from the author on request; archival deposit of the transcripts is planned to follow.Each document is provided as PDF and as its Markdown source.