Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Nonnegative binary rational hafnians: an exact reduction and approximation and sampling consequences

Kriebel Alec

Published Oct 6, 2026• 0 citations
Publisher

Abstract

This note records a self-contained exact reduction from every even-order symmetric matrix with binary nonnegative rational off-diagonal entries to a simple unweighted perfect-matching count, with polynomial encoding length, an explicit logarithmic integer-weight gadget, exact rational scaling, and weighted matching fibers. Citing the general-graph FPRAS of OpenAI, it derives a uniform rational hafnian FPRAS with exact zero detection and a rigorously specified bounded-bit total-variation approximate sampler. Logarithmic integer weight removal, binary rational weighted-to-unweighted equivalence, and counting self-reduction are established machinery and credited to earlier work, including Dell and collaborators, McQuillan, and Jerrum, Valiant, and Vazirani. The approximation breakthrough belongs to OpenAI. This is an attributed consequence and implementation note; no first-publication, new reduction, or independent solution of the base approximation problem is claimed. The supplement contains proofs, transparent code, exact finite enumeration and output-law checks, dependency and priority audits, and clean reproduction instructions. Finite checks do not certify the upstream FPRAS. The real upstream Lean statement was inspected, but no independent kernel rebuild was reproduced. Nonnegative rational inputs only; no signed or complex hafnian or unrestricted Gaussian-boson-sampling consequence. AI tools were used extensively in research, drafting and automated adversarial verification. The preprint has not undergone conventional human peer review or refereeing at publication. Automated reviews are not human peer review. Newly authored prose and data are CC BY 4.0; code is MIT as specified in the archive.

DOI
10.5281/zenodo.23205293

Related papers

Polynomial-Time Exact Learning of Fourier-Sparse Boolean Functions from Nearly Linear Quantum Samples

Daniel Fredriksen

Research manuscript with PDF, LaTeX and review records. An algorithm exactly recovers every k-Fourier-sparse Boolean function with success at least 2/3 using O(k log^3(2k)) independent supplied uniform quantum examples and fixed-polynomial total computation and live space. Fresh residual estimates construct a queryable sign-majority hypothesis; known-function Fourier search and original-target certification contract missing Fourier energy. The theorem uses no target membership query or state-preparation inverse. All samples, adaptive refits and exact rational bit costs are charged. This complements Ye's smaller sample bound with a polynomial computational guarantee. No optimal logarithmic factor, practical implementation, noise robustness, exhaustive novelty, independent peer review, formal certification or BPP/BQP result is claimed. OpenAI Codex assisted derivation, source comparison, drafting and internal checking. Separate model reviews were obtained through Grok, Claude and AGY (configured with Gemini), with findings dispositioned in the archive. These are AI reviews, not independent human peer review.

•Nov 10, 2026•0 citations

Shifted Anticoncentration for Real Gram Hafnians and Symmetric Gaussian Hafnians

Hongru Zhao

We prove a uniform shifted anticoncentration theorem for the hafnian of a real Gaussian Gram matrix under an explicit condition on the row dimension. After normalization by its root mean square, the law has a bounded continuous density, maximal at zero, and every interval has probability bounded by an explicit coefficient times its radius. Under suitable growth conditions on the row dimension, this coefficient grows at most polynomially in the hafnian order, meaning half the dimension of the Gram matrix. We also compute the exact second moment. The proof exploits the perfect matching structure, combining conditional Gaussian representations, row suspension, and bilinear interpolation to control an inverse moment of the conditional variance. At fixed hafnian order, a rescaled limit as the row dimension grows yields corresponding bounds for the hafnian of a real symmetric Gaussian matrix with independent entries above the diagonal.

math.PRCross-listedarXiv:2609.06526•Sep 8, 2026

Exact rational certification of upper and lower quantum bounds for the 𝐼3322 Bell inequality

Nidhal Mghirbi

This preprint constructs finite rational proof objects for both sides of the quantum value of I3322 Bell inequality. Let 𝑄t denote the supremum over finite-dimensional tensor-product binary measurements and𝑄c the commuting-operator value. On the complete word space of total length at most four, we construct a rational Gram certificate. The full signed setting-relabeling stabilizer is the dihedral group of order eight, reducing the certificate to five rational blocks. Ex-act affine reconstruction and a uniform diagonal shift of certified positive interior prove: 𝑄c (𝐼3322) ≤ 62718846253/250000000000 = 0.250875385012. Independently, a local-dimension-499 strategy specified by rational projectors and a nonzero unnormalized rational Schmidt vector has exact normalized value 𝐿499 = 0.25087538451397653548 . . .. Con-sequently, 𝐿499 ≤ 𝑄t(𝐼3322) ≤ 𝑄c (𝐼3322) ≤ 0.250875385012, with endpoint difference below 5 × 10−10. The lower certificate also proves 𝑄t(𝐼3322) > 1/4. Every theorem is checked using exact rational or integer arithmetic; floating-point computation is used only to discover candidates. Complete certificates, source code, tests, and reproducibility materials are archived at: doi:10.5281/zenodo.21462892.

•Jul 21, 2026•2 citations

Pure-loss bosonic dynamic capacities: EPnI consequences and consumed-key secrecy accounting

Kriebel Alec

The finite-energy multimode entropy photon-number inequality in the OpenAI mathematics release supplies the sharp entropy premise for the previously conditional pure-loss bosonic classical/quantum/entanglement capacity region under an average photon constraint, including arbitrary entanglement across channel uses. The one-mode pure-loss channel has transmissivity 1/2 ≤ η ≤ 1, with forward resource accounting without free feedback; the photon-number constraint is averaged over messages, key and encoder randomness, rather than a peak or per-codeword bound. This note gives the precise consequence and audits a private-resource operational defect: the printed catalytic secrecy condition in the cited coding theorem protects consumed key jointly with generated secrets and excludes the one-time pad. A zero-energy converse refutes the displayed private region under that literal condition. The public/private/secret-key formula is proved under an explicitly corrected generated-secret trace-secrecy convention, with corrected converse accounting and direct key-assisted coding. The formulas and conditional reductions are inherited; this is not an independent proof of EPnI or a full solution of the false unchanged private target. The upstream unrefereed EPnI proof was analytically audited; its Lean build was not reproduced. AI tools were used extensively in research, drafting and verification. Automated adversarial reviews are not conventional human peer review; this preprint has not undergone conventional human refereeing.

•Oct 6, 2026•0 citations

Constructing a rational cp-matrix whose every cp-factorization is irrational, with GPT-6 Astra

Yaroslav Shitov

•Sep 9, 2026•0 citations

A Nonstandard Variant of Learning with Rounding with Polynomial Modulus and Unbounded Samples

Hart William Montgomery

Lecture notes in computer science•Jan 1, 2018•5 citations
This note records a self-contained exact reduction from every even-order symmetric matrix with binary nonnegative rational off-diagonal entries to a simple unweighted perfect-matching count, with polynomial encoding length, an explicit logarithmic integer-weight gadget, exact rational scaling, and weighted matching fibers. Citing the general-graph FPRAS of OpenAI, it derives a uniform rational hafnian FPRAS with exact zero detection and a rigorously specified bounded-bit total-variation approximate sampler. Logarithmic integer weight removal, binary rational weighted-to-unweighted equivalence, and counting self-reduction are established machinery and credited to earlier work, including Dell and collaborators, McQuillan, and Jerrum, Valiant, and Vazirani. The approximation breakthrough belongs to OpenAI. This is an attributed consequence and implementation note; no first-publication, new reduction, or independent solution of the base approximation problem is claimed. The supplement contains proofs, transparent code, exact finite enumeration and output-law checks, dependency and priority audits, and clean reproduction instructions. Finite checks do not certify the upstream FPRAS. The real upstream Lean statement was inspected, but no independent kernel rebuild was reproduced. Nonnegative rational inputs only; no signed or complex hafnian or unrestricted Gaussian-boson-sampling consequence. AI tools were used extensively in research, drafting and automated adversarial verification. The preprint has not undergone conventional human peer review or refereeing at publication. Automated reviews are not human peer review. Newly authored prose and data are CC BY 4.0; code is MIT as specified in the archive.
Research manuscript with PDF, LaTeX and review records. An algorithm exactly recovers every k-Fourier-sparse Boolean function with success at least 2/3 using O(k log^3(2k)) independent supplied uniform quantum examples and fixed-polynomial total computation and live space. Fresh residual estimates construct a queryable sign-majority hypothesis; known-function Fourier search and original-target certification contract missing Fourier energy. The theorem uses no target membership query or state-preparation inverse. All samples, adaptive refits and exact rational bit costs are charged. This complements Ye's smaller sample bound with a polynomial computational guarantee. No optimal logarithmic factor, practical implementation, noise robustness, exhaustive novelty, independent peer review, formal certification or BPP/BQP result is claimed. OpenAI Codex assisted derivation, source comparison, drafting and internal checking. Separate model reviews were obtained through Grok, Claude and AGY (configured with Gemini), with findings dispositioned in the archive. These are AI reviews, not independent human peer review.
We prove a uniform shifted anticoncentration theorem for the hafnian of a real Gaussian Gram matrix under an explicit condition on the row dimension. After normalization by its root mean square, the law has a bounded continuous density, maximal at zero, and every interval has probability bounded by an explicit coefficient times its radius. Under suitable growth conditions on the row dimension, this coefficient grows at most polynomially in the hafnian order, meaning half the dimension of the Gram matrix. We also compute the exact second moment. The proof exploits the perfect matching structure, combining conditional Gaussian representations, row suspension, and bilinear interpolation to control an inverse moment of the conditional variance. At fixed hafnian order, a rescaled limit as the row dimension grows yields corresponding bounds for the hafnian of a real symmetric Gaussian matrix with independent entries above the diagonal.
This preprint constructs finite rational proof objects for both sides of the quantum value of I3322 Bell inequality. Let 𝑄t denote the supremum over finite-dimensional tensor-product binary measurements and𝑄c the commuting-operator value. On the complete word space of total length at most four, we construct a rational Gram certificate. The full signed setting-relabeling stabilizer is the dihedral group of order eight, reducing the certificate to five rational blocks. Ex-act affine reconstruction and a uniform diagonal shift of certified positive interior prove: 𝑄c (𝐼3322) ≤ 62718846253/250000000000 = 0.250875385012. Independently, a local-dimension-499 strategy specified by rational projectors and a nonzero unnormalized rational Schmidt vector has exact normalized value 𝐿499 = 0.25087538451397653548 . . .. Con-sequently, 𝐿499 ≤ 𝑄t(𝐼3322) ≤ 𝑄c (𝐼3322) ≤ 0.250875385012, with endpoint difference below 5 × 10−10. The lower certificate also proves 𝑄t(𝐼3322) > 1/4. Every theorem is checked using exact rational or integer arithmetic; floating-point computation is used only to discover candidates. Complete certificates, source code, tests, and reproducibility materials are archived at: doi:10.5281/zenodo.21462892.
The finite-energy multimode entropy photon-number inequality in the OpenAI mathematics release supplies the sharp entropy premise for the previously conditional pure-loss bosonic classical/quantum/entanglement capacity region under an average photon constraint, including arbitrary entanglement across channel uses. The one-mode pure-loss channel has transmissivity 1/2 ≤ η ≤ 1, with forward resource accounting without free feedback; the photon-number constraint is averaged over messages, key and encoder randomness, rather than a peak or per-codeword bound. This note gives the precise consequence and audits a private-resource operational defect: the printed catalytic secrecy condition in the cited coding theorem protects consumed key jointly with generated secrets and excludes the one-time pad. A zero-energy converse refutes the displayed private region under that literal condition. The public/private/secret-key formula is proved under an explicitly corrected generated-secret trace-secrecy convention, with corrected converse accounting and direct key-assisted coding. The formulas and conditional reductions are inherited; this is not an independent proof of EPnI or a full solution of the false unchanged private target. The upstream unrefereed EPnI proof was analytically audited; its Lean build was not reproduced. AI tools were used extensively in research, drafting and verification. Automated adversarial reviews are not conventional human peer review; this preprint has not undergone conventional human refereeing.