We prove an improved lower bound of 0.097387436922 on the largest normalised Ingleton violation among four-variable almost-entropic vectors. Previously reported constructions reach approximately 0.0925, used as a rounded benchmark; the best published universal upper bound is 19/124. Conditional parity hashing combines local binary parity labels, one common linear hash, simultaneous conditional privacy amplification and joint conditioning. Applied to a 24-atom source of Matúš and Csirmaz, it subtracts a multiple of the uniform-matroid rank function U(3,4) from all fifteen entropy coordinates. The limiting signed score lies in [-0.097387436923, -0.097387436922], and finite-alphabet distributions with score below -0.09738 exist. This record contains the preprint PDF, manuscript sources, Lean proof sources, exact numerical checkers, build instructions and verification evidence. The main theorem, including the closed rate interval, score enclosure and finite-alphabet existence statement, was rebuilt and kernel-checked in Lean 4.35.0-rc2 with pinned Mathlib dependencies. The finite sampler and fixed-path optimality argument are written proofs rather than separate Lean declarations. Søren Riis is the creator and coordinator and directed the research. Astra/Codex and Claude contributed mathematical development, computation and verification; Claude contributed the Lean formalisation and adversarial review. The source construction is due to Matúš and Csirmaz; the paper gives full references to the earlier entropy-region and privacy-amplification work. AI checks do not constitute independent human peer review. Licences by component: the paper PDF, LaTeX sources and original figures are CC BY 4.0; Lean sources, Python checkers and build scripts are Apache-2.0. Third-party dependencies retain their own licences. Repository: https://github.com/SR123/conditional-parity-hashing
We prove an improved lower bound of 0.097387436922 on the largest normalised Ingleton violation among four-variable almost-entropic vectors. Previously reported constructions reach approximately 0.0925, used as a rounded benchmark; the best published universal upper bound is 19/124. Conditional parity hashing combines local binary parity labels, one common linear hash, simultaneous conditional privacy amplification and joint conditioning. Applied to a 24-atom source of Matúš and Csirmaz, it subtracts a multiple of the uniform-matroid rank function U(3,4) from all fifteen entropy coordinates. The limiting signed score lies in [-0.097387436923, -0.097387436922], and finite-alphabet distributions with score below -0.09738 exist. This record contains the preprint PDF, manuscript sources, Lean proof sources, exact numerical checkers, build instructions and verification evidence. The main theorem, including the closed rate interval, score enclosure and finite-alphabet existence statement, was rebuilt and kernel-checked in Lean 4.35.0-rc2 with pinned Mathlib dependencies. The finite sampler and fixed-path optimality argument are written proofs rather than separate Lean declarations. Søren Riis is the creator and coordinator and directed the research. Astra/Codex and Claude contributed mathematical development, computation and verification; Claude contributed the Lean formalisation and adversarial review. The source construction is due to Matúš and Csirmaz; the paper gives full references to the earlier entropy-region and privacy-amplification work. AI checks do not constitute independent human peer review. Licences by component: the paper PDF, LaTeX sources and original figures are CC BY 4.0; Lean sources, Python checkers and build scripts are Apache-2.0. Third-party dependencies retain their own licences. Repository: https://github.com/SR123/conditional-parity-hashing
We study independent one-message shuffling through a common channel with replacement adjacency and arbitrary deterministic backgrounds. Under one local Rényi budget, without a pointwise likelihood-ratio cap, the optimal uniform same-order rate is n−min{1,α−1} for fixed α>1; a local KL budget alone gives no uniform amplification. We determine the exact asymptotic worst-case approximate-DP envelope at each fixed positive privacy parameter and the leading inverse calibration at failure probability d/n for all fixed α>1. Matching upper bounds are uniform over channels and alphabets varying with n. At and above order three, the calibration coefficient is the square root of an attained local triple-variance frontier, realizable on at most nine output symbols. At order three we evaluate this frontier on a certified budget interval and identify the unique extremal ratio law. The same frontier determines the limiting central-window profile for α>2 and the logarithmic moderate-window exponent. Under local χ2 diameter C, the exact asymptotic neighboring coefficient is 2C(C+5)/(C+4); we also give explicit finite amplification bounds and a fixed-channel background reduction. Formal verification. All statements of the paper have been formalized in Lean 4 (version 4.35.0) with Mathlib (release v4.35.0), including the computer-assisted order-three theorem, for which all certificate computations are checked by the Lean kernel. The formal proofs contain no sorry and use only the standard axioms propext, Classical.choice and Quot.sound. The project builds from scratch, every module passes a replay by the Lean kernel (leanchecker), and all theorems of the interface file Statements.lean, together with everything they depend on (69 078 declarations in total), were rechecked by the independent type checker nanoda. ChatGPT was used to write the Lean code and Claude to audit it; the author takes full responsibility for all results. Files. The PDF is the paper (24 pages). The ZIP archive Uniform_Renyi_arXiv_source_2026-11-09.zip is the LaTeX source. Its folder anc/ contains the exact rational polynomial certificate for the computer-assisted order-three theorem (verify_order_three_frontier.py, needs only the Python standard library) and further exact checks. The ZIP archive UniformRenyiShuffle-Lean-1.0.0-2026-11-09.zip is the Lean project UniformRenyiShuffle 1.0.0; its file MAP.md gives the Lean name of every statement of the paper. Changes from version 1. The Lean formalization is added. The paper and its LaTeX source are unchanged.
For four finite discrete random variables, we prove that the negative of the Ingleton expression, divided by the joint entropy, is at most 19/124≈0.15323, improving the previously reported upper bound 3/19 by 11/2356. The proof uses a copy program with four copy steps and six auxiliary variables, certified by exact rational linear-programming certificates. We also show that two published information inequalities, due to Dougherty-Freiling-Zeger and Csirmaz, imply the bound 5/32 by an explicit nonnegative combination with Shannon inequalities, and that 5/32 is the exact optimum both of the union of the three published inequality catalogues considered here and of the full copy-lemma extensions of all 213 distinct copy programs occurring in them. The new program is not among these programs, and the bound 19/124 remains optimal when all of them and all catalogue inequalities are added. Ancillary files provide the parsed catalogues, exact primal and dual certificates, exact extension certificates, and verifiers that use only the Python standard library.
This paper characterizes the fixed finite positive semidefinite composition partners whose tensor powers test strong positivity of every finite Hermitian matrix: a partner is universal exactly when it has a nonreal entry in the specified history basis. Constructive proofs give explicit copy bounds, finite-depth spectral and tolerance certificates, a finite-family obstruction, and a quantum-history realization. For a noisy two-history family, all-event lower bounds and explicit exposing events establish sharp weak-imaginary resource scaling. Further results include an exact pure-phase onset and conditional finite-composition CHSH certificates. The accompanying ZIP contains the LaTeX source, six exact standard-library Python checkers and their results, a package verifier and file hashes. Priority remains unestablished. No derivation of a physical probability law is claimed.Robert Barritz directed the research and its preparation for publication. OpenAI’s ChatGPT and Codex systems contributed substantially to mathematical formulation, proof development and examination, examples, exact computational verification, literature comparison, and manuscript preparation. Anthropic’s Claude contributed mathematical, computational, and literature review that informed revisions. The AI contribution extended substantially beyond language editing. AI-assisted review does not constitute human expert or journal peer review.
The paper deals with linear information inequalities valid for entropy functions induced by discrete random variables. Specifically, the so-calledconditional Ingleton inequalitiesare in the center of interest: these are valid under conditional independence assumptions on the inducing random variables. We discuss five inequalities of this particular type, four of which has appeared earlier in the literature. Besides the proof of the new fifth inequality, simpler proofs of (some of) former inequalities are presented. These five information inequalities are used to characterize all conditional independence structures induced by four discrete random variables.
The motivating problem is agnostically learning parity functions, i.e., parity with arbitrary or adversarial noise. Specifically, given random labeled examples from an *arbitrary* distribution, we would like to produce an hypothesis whose accuracy nearly matches the accuracy of the best parity function. Our algorithm runs in time 2O(n/log n), which matches the best known for the easier cases of learning parities with random classification noise (Blum et al, 2003) and for agnostically learning parities over the uniform distribution on inputs (Feldman et al, 2006).
A source-inclusive coherence observable is useful only when its measurement resources are admitted. This preprint constructs a bounded local reader for the conserved exchange observable of a specified finite quantum-record apparatus. A degree-three graph uses one independently prepared probe, twenty relays with arbitrary joint states, and finite-duration two-register extraction and full-SWAP interactions. The original apparatus Hamiltonian and an independently added archive-gated process remain active throughout their respective implementations. A prefix-dependent control-frame comparison, symmetry-resolved effect estimate, and explicit gain, phase and finite-slew waveform bounds yield a sufficient response separation greater than 0.005139889 for the stated fixed, fully accessible sixteen-position source and remaining response-error budgets. This is a conditional mathematical result, not measured hardware calibration, a partial-source-access theorem, or a unique diagnosis or derivation of fundamental collapse. The package contains the main manuscript, technical supplement, editable LaTeX, direct numerical reproduction sources, and accepted returned Mac evidence for 56 paper-specific supporting Lean declarations. That subset verifies selected algebra, scalar inequalities and the actual-constant numerical expression; it does not formalize the full physical/continuous-time theorem end-to-end. Separate-session AI review, numerical reproduction and returned kernel evidence are distinguished. Author-owned manuscripts, documentation and mathematical data are CC BY 4.0; newly authored numerical and Lean software is MIT. Predecessor and third-party notices remain unchanged; the reproduction package contains the licence map. No external funding; no competing interests declared. Public preprint v1.0; edition date 18 October 2026; DOI 10.5281/zenodo.23442579.