Two-Tier Semantic Mixing and Disambiguation Budget: A Long-Form Study of Finite Core, Row-Gluing Obstructions, and Topologically Protected Resource Theory
Year: 2026
Type: preprint
Abstract: We study a finite discrete identification problem: given a finite set X, a target ρ:X→ℝ, and a two-tier observation π=(q,r):X→ A× B, when does a detector D with D∘π=ρ exist, and if not, how is the obstruction quantified? We separate local fiber capacities μ_a, the global fiberwise mixing degree μ_fib, the coarse aggregate mixing degree τ(q), and the natural row count ν_r. Exact detection is equivalent to μ_fib=1; this equivalence is a standard identifiability/rough-set starting point, not a new result, and is used as a unified language. Row structure is formalized via a conflict graph; ν_r is a set-partition number, and the atomic row count reduces to bin packing in the no-conflict case, where the decision problem is NP-complete and admits no PTAS. The product counting chain τ(q)≤μ_fibν_r(π) is unconditional; saturation is characterized by perfect row representations. For μ_fib>1 the unrestricted disambiguation budget equals μ_fib; restricted budgets admit universal, entropy, and span lower bounds. The Three-Layer Bridge. The core contribution is a unified logical chain: (1) Local Discrete Core: combinatorial constraints on row-labeling (natural row count ν_r); (2) Topological Obstruction: non-abelian Čech cohomology on the separable row-label groupoid, instantiated as a groupoid-valued presheaf on the Čech nerve in the sense of Giraud–Moerdijk, yields an explicit obstruction class [c] whose vanishing is equivalent to the existence of a global section; (3) Macroscopic Resource Transition: the Bridge Theorem (Theorem[thm:bridge], proven under the uniform rate definition and the monodromy-preserving operation set) proves that the non-trivial topological obstruction (H⁰=∅) forbids positive uniform distillation rates (R_dist=0) while preserving dilution (R_dil>0). The distillation rate spectrum (Proposition[prop:rate-spectrum]) refines this: the distillable tensor powers are exactly those n for which Mon_∞ⁿ has a fixed point, and the per-copy rate on this subsequence is logμ_fib; the uniform rate is zero because the complement of this set is infinite. The zero-rate bound is carried by the monodromy obstruction, not by monotonicity of logμ_fib: logμ_fib is not a monotone under monodromy-preserving operations (Proposition[prop:mufib-not-monotone]). This connects local combinatorics to macroscopic resource asymmetry via global topology. A verified |I|=3 no-descent-data obstruction is given in the separable layer (Proposition[prop:no-descent]), and a polynomial-time path-consistency criterion is obtained by instantiating the known majority-polymorphism bounded-width theorem on a median-polymorphism subclass (Theorem[thm:median-path]); the contribution is the verification of condition (L) on the separable row-label constraint language, not the bounded-width theorem itself. Theorem[thm:i2-elimination] eliminates the |I|=2 case on connected nerves under the standing restriction hypothesis H-restrict; without H-restrict, Proposition[prop:i2-counterexample] gives a connected 4-cycle counterexample, so the verified |I|=3 obstruction is minimal in the separable layer under H-restrict. The resource part establishes an obstruction-based resource dichotomy for ℤ-ladders under the monodromy-preserving operation set (Theorem[thm:correspondence]): H⁰≠∅ iff full distillation at rate logμ_fib, and H⁰=∅ iff zero uniform distillation, with dilution at least logμ_fib. Proposition[prop:z-ladder-no-fourth] shows that on a ℤ-ladder the fourth-type obstruction does not occur independently of monodromy, so the dichotomy is exhaustive. It is realized by Model D.1 (Abelian monodromy), Model D.2 (non-Abelian finite-order monodromy), and Model D.3 (syndrome/coset ladder with unbounded label space). Boundary of claims. The finite core is proved unconditionally. The gluing and resource parts contain one verified obstruction and an obstruction-based resource dichotomy on ℤ-ladders under monodromy-preserving operations, realized by Abelian and non-Abelian finite-order monodromy models. The continuous-limit section is conditional on H4/H6. The noncommutative and Postnikov sections are an outlook. The effective-descent hypothesis (H-ED) and edge-slack positivity (H-ESP) are not verified in general. H-restrict is a working hypothesis for the elimination statements; without it, |I|=2 obstructions can occur (Proposition[prop:i2-counterexample]). Genuine fourth-type obstructions on complexes with nontrivial fundamental group are outside the scope of this paper.
Author: Boris, Boomer
Language: English
Topic: Rough Sets and Fuzzy Logic
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7235022655
