{"id":"13dd9824-c75f-4314-82bd-9c9a0da250ec","arxiv_id":"2509.05487","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For beta=L^2 even, correlation functions of beta-ensembles are expressed as Vandermonde times hyperpfaffians of Wronskian-built L-vectors.","lead":"A new exact formula expresses correlation functions of beta-ensembles, for beta an even square integer, as Vandermonde factors times hyperpfaffians of exterior-algebra vectors built from Wronskians. The main theorem generalizes the classic Pfaffian (beta=4) and determinant (beta=2) correlation formulas to a family of beta values, with explicit small-M pair correlations for beta=16, but several stated Wronskian and hyperpfaffian evaluations contain an exponent error and need c","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"False monomial Wronskian lemma, not the determinant step, is the real obstruction; central Theorem 4.1 survives, secondary hyperpfaffian evaluations fail as printed.","rationale":"The reader's conditional verdict is substantively correct, but for a more specific reason than stated. The allegedly problematic determinant in the proof of Theorem 4.1 is straightforwardly correct by block triangularity, so the main theorem's proof is not actually missing a hidden sign or Vandermonde factor. The real, demonstrable error is Lemma 3.1: the monomial Wronskian exponent is misstated, and this error propagates into at least Proposition 4.4 and Proposition 4.5, both of which fail already in the smallest nontrivial cases. These are secondary to the central hyperpfaffian correlation theorem, which does not depend on Lemma 3.1, but they are advertised results and must be fixed. The correct course is to keep the conditional verdict: the paper needs revision, and the monomial-based formulas need regeneration, while the core Theorem 4.1 appears sound.","tokens_in":17551,"tokens_out":28912,"duration_ms":286732,"concrete_test":"Recompute Proposition 4.4 for L=2, M=2, a=b=1 using the corrected monomial Wronskian: replace B(a+Σ_t,b) with B(a+Σ_t−L(L−1)/2,b)=B(a+Σ_t−1,b). Evaluate the hyperpfaffian of the resulting 2-vector and compare with the Selberg value 1/30. As printed the hyperpfaffian equals 1/240; with the shifted exponent it should equal 1/30. This single check isolates whether the false Lemma 3.1 is the cause of the numerical discrepancy and determines whether all monomial-based hyperpfaffian evaluations in §4.2 need regeneration.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader flagged the unproved determinant in §5.1, but that determinant is actually correct: with the q_y family, ordering the Lm rows as (j,ℓ), the coefficient matrix is block upper triangular, the j-th diagonal block is S_j(y_j)I_L with S_j(y_j)=∏_{i<j}(y_j−y_i)^L, so det(U)=∏_{j<k}(y_k−y_j)^{L^2}. Thus that concern does not land.\n\nThe load-bearing flaw is Lemma 3.1. With D^ℓ=1/ℓ! d^ℓ and monomials m_t, one has det[D^ℓ x^{t_k}] = eΔ_t x^{Σ_t − L(L−1)/2}, not eΔ_t x^{Σ_t}. The missing shift invalidates every monomial-based hyperpfaffian evaluation as printed. For example, Proposition 4.4 with L=2, M=2, a=b=1 gives, as stated, PF[Σ B(1+Σ_t,1)eΔ_t e_t] = 1/240, whereas the Selberg partition function is 1/30. Replacing B(a+Σ_t,b) by the shifted B(a+Σ_t−1,b) yields exactly 1/30. Proposition 4.5 similarly fails for L=2,M=2: the printed sum has PF 64, while the Mehta value is 6.\n\nThese are not merely cosmetic: the abstract advertises hyperpfaffian evaluations, and Section 4.2 presents them as results. The main Theorem 4.1 and the circular formulas that explicitly include the shift (e.g., δ_u definitions) appear to be unaffected, but the monomial Wronskian lemma and the propositions derived from it must be corrected before the paper can be accepted as a whole.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a hyperpfaffian formulation for correlation functions of beta-ensembles when beta = L^2 with L even. The main result, Theorem 4.1, expresses the m-th correlation function as a Vandermonde determinant in the external points times the hyperpfaffian of an L-vector gamma_y whose coefficients are integrals of Wronskians multiplied by prod_j (x-y_j)^beta. The circular case is developed using monomials, yielding explicit formulas for gamma_y and pair-correlation polynomials for beta=16 and small M. The paper also records hyperpfaffian evaluations based on Selberg, Mehta, and Dyson-Gunson integrals. The proof structure of Theorem 4.1 is coherent, but Lemma 3.1 is false as stated, and the monomial-based hyperpfaffian evaluations in Section 4.2 are consequently incorrect as printed.","tokens_in":17998,"tokens_out":24440,"duration_ms":230054,"significance":"If corrected, the framework would give exact correlation formulas for beta=L^2 even ensembles, generalizing the beta=4 Pfaffian point-process structure. The formulas involve no free parameters, and the circular pair-correlation polynomials are concrete, checkable outputs. The main determinant step in Section 5.1 is sound. However, the advertised hyperpfaffian evaluations in Section 4.2 are currently unreliable because of the Wronskian exponent error. The central theorem and the circular formulas appear to survive, but the secondary results must be corrected before the paper can be accepted.","major_comments":[{"comment":"Lemma 3.1 is false as stated. With D^ell = (1/ell!) d^ell/dx^ell, the matrix [D^ell x^{t_k}] has determinant eDelta_t x^{Sigma_t - L(L-1)/2}, not eDelta_t x^{Sigma_t}. This follows from factoring x^{t_k - ell} and using Vandermonde determinants. The missing shift propagates into every monomial-based computation that uses the lemma without an additional factor.","section":"Section 3.5, Lemma 3.1"},{"comment":"Proposition 4.4 is incorrect as printed. For L=2, M=2, a=b=1, the stated hyperpfaffian equals 1/240, whereas the Selberg partition function is 1/30. Replacing B(a+Sigma_t,b) by B(a+Sigma_t - L(L-1)/2,b) gives exactly 1/30 in this example. The proposition must be restated with the shifted Beta argument.","section":"Section 4.2, Proposition 4.4"},{"comment":"Proposition 4.5 is incorrect as printed. For L=2, M=2, the printed sum has hyperpfaffian 64, while the stated right-hand side is 6. There are two issues: the missing Wronskian shift from Lemma 3.1, and the claim that (2j)!! is the (2j)-th moment of a standard normal random variable. The even double factorial (2j)!! = 2^j j! is not the moment; the correct moment is (2j-1)!!. Both corrections are needed for the proposition to hold.","section":"Section 4.2, Proposition 4.5"}],"minor_comments":[{"comment":"The determinant evaluation det(U) = prod_{j<k}(y_k-y_j)^beta is asserted as 'an easy calculation.' Since this determinant is the sole source of the Vandermonde factor in Theorem 4.1, a short proof (e.g., block upper triangular with diagonal blocks S_j(y_j) I_L) should be included.","section":"Section 5.1"},{"comment":"The displayed formula for int_T Wr(m_t) u dmu is correct only because u includes the circular factor c^{(M-1)beta/2}; the Wronskian shift from the corrected Lemma 3.1 is absorbed by this factor. The text should say this explicitly to avoid confusion.","section":"Section 4.1"},{"comment":"There is a typo in the summation condition: 'Sigma_t = Sigma_t' should be 'delta_t = 0' or equivalently 'Sigma_t = L(N-1)/2'.","section":"Section 4.2, Proposition 4.7"},{"comment":"The double-factorial notation should be defined consistently with standard normal moments, or the moments should be written as (2j-1)!!.","section":"Section 4.2, Proposition 4.5"}],"recommendation":"major_revision","confidential_remarks":"The central theorem and the circular correlation formulas appear sound, and the flaws are localized to the monomial Wronskian lemma and the non-circular hyperpfaffian evaluations. These are fixable within the scope of the paper, but the concrete counterexamples for Propositions 4.4 and 4.5 mean the current version should not be accepted as is."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real new content here is Theorem 4.1: for beta = L^2 with L even, the m-th correlation function is a Vandermonde factor times a hyperpfaffian of an L-vector built from Wronskian integrals. That generalizes the beta = 4 Pfaffian story and is worth having. The proof via an averaged characteristic polynomial / exterior algebra is a nice method, and the circular specialization (Theorem 4.2, Corollary 4.3) gives a usable formalism, including exact pair-correlation polynomials for beta = 16 and small M. Those are concrete, checkable outputs and the paper is honest about the fact that the hyperpfaffian partition function was already known from Sinclair and Wells' thesis.\n\nThe soft spot is real and specific. Lemma 3.1 is false as stated: for monomials, the Wronskian has total power Sigma_t - L(L-1)/2, not Sigma_t. The missing shift propagates directly into Proposition 4.4, Proposition 4.5, and Proposition 4.9 (and the surrounding text), because they all feed monomial Wronskians into hyperpfaffian evaluations. The stress-test note is right: the determinant in §5.1 that the reader worried about is actually fine; the block-triangular computation gives the Vandermonde factor. So the main theorem's proof structure is sound, and Theorem 4.1 probably survives. But the printed hyperpfaffian evaluations in Section 4.2 are not reliable until the Wronskian exponent is corrected and all monomial-based formulas are regenerated. The circular results look less affected because the shift is already folded into the delta_u definitions, but the paper should explicitly re-derive them after fixing Lemma 3.1.\n\nThis is a solid subfield advance with a fixable error, not a fatal flaw. The authors need to correct Lemma 3.1, re-run the evaluations, and either prove or carefully state the determinant identity in §5.1. The abstract and Section 4.2 currently overclaim. I would send this to a serious referee: the central theorem is important enough and the mistakes are localized enough that a revision could make this a genuinely useful paper.","headline":"The genuinely new correlation-function theorem (4.1) and the circular specialization look salvageable, but the monomial Wronskian lemma is misstated and several advertised hyperpfaffian evaluations fail as printed.","tokens_in":18502,"tokens_out":1320,"would_cite":true,"duration_ms":15736,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["15B52","60B20","60G55","82B23","15A15"],"pacs":[],"model":"deepseek-v4-flash","headline":"For beta equal to an even square, the correlation functions of beta-ensembles are hyperpfaffians of L-vectors built from Wronskian integrals, generalizing the beta=4 Pfaffian point process.","keywords":["random matrices","beta ensembles","exterior algebra","correlation functions","pfaffians","hyperpfaffians","hyperpfaffian evaluations","Wronskians"],"falsifier":"Compute the determinant of U for the smallest nontrivial cases (L=2,m=2 or L=4,m=1) with a computer-algebra system; if it is not prod_{j<k}(y_k-y_j)^beta, Theorem 4.1's Vandermonde factor is wrong. Independently, evaluate Wr(m_t;x) from Lemma 3.1 for a small index set such as t={0,1,2} to check the printed exponent before trusting the circular coefficient formulas.","tokens_in":17426,"feed_emoji":"📐","tokens_out":8844,"duration_ms":90324,"temperature":0.7,"pith_summary":"The paper studies beta-ensembles, random point processes whose joint density is proportional to the product of pairwise distances raised to a power beta times a weight, and asks for their m-th correlation functions, the marginal densities that control gaps, clustering, and spacing statistics. It claims that whenever beta = L^2 with L an even positive integer, the m-th correlation function has an exact hyperpfaffian form: a Vandermonde product in the observed points y times weights times the hyperpfaffian of an L-vector gamma_y. The coefficients of gamma_y are integrals of Wronskians of shifted monic polynomials against the ensemble weight, so the whole correlation function is reduced to one hyperpfaffian evaluation. If correct, this gives beta=4 its classical Pfaffian process as the L=2 case and opens beta=16,36,... to the same exact treatment, with the circular ensembles worked out in concrete polynomial form.","feed_headline":"For beta = 4, 16, 36, ... correlations are exact hyperpfaffians","feed_subtitle":"Exact formulas extend the beta=4 Pfaffian machinery to every even-square beta.","key_machinery":"The machinery is the hyperpfaffian of an L-vector: for omega in Lambda^L V, PF(omega) = *(omega^{wedge M}/M!), read off from the coefficient of the volume form after wedging the L-vector with itself M times. The paper builds omega(x) = p(x) wedge D^1 p(x) wedge ... wedge D^{L-1} p(x) from a complete family of monic polynomials; its Grassmann coordinates are Wronskians Wr(p_t;x). The key identity is the confluent Vandermonde formula, *omega(x_1) wedge ... wedge omega(x_M) = prod_{m<n}(x_n-x_m)^{L^2}, which turns the beta power into a wedge of L-vectors. The Gram L-vector gamma = int_W u(x) omega(x) dmu(x) carries the partition function, and the correlation hyperpfaffian gamma_y is the same ob","core_discovery":"The paper's central claim is Theorem 4.1: for beta=L^2 with L even, the m-th correlation function of the ensemble is R_m(y) = (1/Z) prod_{j<k}(y_k-y_j)^beta prod_n u(y_n) PF(gamma_y), where gamma_y is the L-vector over an (L(M-m))-dimensional space whose coefficient for each index set u is integral_W [prod_{j=1}^m (x-y_j)^beta] Wr(p_u;x) u(x) dmu(x). Here Wr(p_u;x) is the renormalized Wronskian of the L monic polynomials selected by u, and PF is the hyperpfaffian, the natural L-vector extension of the Pfaffian. The partition function itself is Z = PF(gamma) for the Gram L-vector gamma = int tilde omega dmu. The theorem is proved by inserting the averaged characteristic polynomial identity an","pith_inferences":["The determinant identity for the auxiliary family q_y is asserted as an easy calculation but not proved; a symbolic check for small L and m would settle whether the Vandermonde factor in Theorem 4.1 is exactly as stated.","If a diagonalizing family exists for classical weights, the construction would produce an L-vector kernel analogous to the beta=4 matrix kernel, opening a generalized Pfaffian point-process analysis; the paper does not construct such families.","The same Laurent-coefficient mechanism used for beta=16 should extend to beta=36,64,... and larger M, though the multinomial enumeration in E grows quickly and would need efficient recursive or parallel computation.","The printed monomial Wronskian formula, Lemma 3.1, should be re-verified before relying on the circular coefficient expressions numerically; any exponent error there would propagate into Theorem 4.2 and Corollary 4.3."],"forward_implications":["Every beta=L^2 even ensemble gains an exact hyperpfaffian formula for all correlation functions, with beta=4 as the Pfaffian special case and beta=16,36,... accessible by the same calculation.","The circular ensembles have an explicit, sparse gamma_y: coefficients vanish except when the sum of selected degrees matches a central value, and the surviving coefficients are Laurent-polynomial coefficients, making R_2(theta) a polynomial in cos(theta).","The framework yields hyperpfaffian evaluations from known partition functions such as Selberg, Mehta, and Dyson integrals, giving independent checks and new identities for exterior-algebra computations.","Finding monic families 'sympathetic' to the weight makes gamma_y sparse, a direct generalization of skew-orthogonal polynomials; such families are the route to closed-form kernels."],"supporting_citations":[{"why":"Established the hyperpfaffian partition-function formulation for beta a square integer that this paper extends to correlations.","marker":"[24]"},{"why":"Prior thesis on solvability of beta square integer, foundational for the Gram L-vector method.","marker":"[32]"},{"why":"Supplies the confluent Vandermonde identity that converts beta powers into wedges of L-vectors.","marker":"[31]"},{"why":"Provides the beta=4 skew-orthogonal polynomials that the new sympathetic families generalize.","marker":"[1]"},{"why":"Supplies the Selberg integral background used to produce Jacobi-weight hyperpfaffian evaluations.","marker":"[14]"},{"why":"Gives Dyson's constant conjecture for circular ensembles, used in the circular hyperpfaffian evaluation.","marker":"[8]"},{"why":"Proves Dyson's conjecture, anchoring the circular hyperpfaffian evaluation used in the paper.","marker":"[16]"}],"fun_headline_variants":["Even-square beta: correlations as hyperpfaffians","Hyperpfaffians crack beta-ensembles for L^2","Exact correlations for beta = 4, 16, 36 via hyperpfaffians","Beta = L^2: hyperpfaffian formulas for all correlations"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The proof depends on an unproved 'easy calculation': that the auxiliary complete family q_y makes the block matrix U triangular with determinant exactly prod_{j<k}(y_k-y_j)^beta; that determinant is the only source of the Vandermonde factor in Theorem 4.1, so if it is wrong every correlation formula built on it shifts. The circular branch additionally relies on the monomial Wronskian formula of Lemma 3.1, whose printed exponent should be re-verified before the circular coeffi","fun_headline_variants_meta":{"raw":{"variants":["Even-square beta: correlations as hyperpfaffians","Hyperpfaffians crack beta-ensembles for L^2","Exact correlations for beta = 4, 16, 36 via hyperpfaffians","Beta = L^2: hyperpfaffian formulas for all correlations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000487,"raw_usage":{"total_tokens":2369,"prompt_tokens":1009,"completion_tokens":1360,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":753,"completion_tokens_details":{"reasoning_tokens":1281}},"tokens_in":753,"tokens_out":1360,"duration_ms":10377,"temperature":1.0,"reasoning_tokens":1281,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T05:26:24.687355+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the determinant of U for the smallest nontrivial cases (L=2,m=2 or L=4,m=1) with a computer-algebra system; if it is not prod_{j<k}(y_k-y_j)^beta, Theorem 4.1's Vandermonde factor is wrong. Independently, evaluate Wr(m_t;x) from Lemma 3.1 for a small index set such as t={0,1,2} to check the printed exponent before trusting the circular coefficient formulas.","supporting_citations":[{"cited_title":"Ensemble averages whenβis a square integer.Monatshefte f¨ ur Mathematik, 166:121–144, 2012","cited_arxiv_id":null,"evidence_quote":"Established the hyperpfaffian partition-function formulation for beta a square integer that this paper extends to correlations."},{"cited_title":"Wells.On the Solvability of Beta-ensembles when Beta Is a Square Integer","cited_arxiv_id":null,"evidence_quote":"Prior thesis on solvability of beta square integer, foundational for the Gram L-vector method."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the confluent Vandermonde identity that converts beta powers into wedges of L-vectors."},{"cited_title":"Adler, P","cited_arxiv_id":null,"evidence_quote":"Provides the beta=4 skew-orthogonal polynomials that the new sympathetic families generalize."},{"cited_title":"Forrester and S","cited_arxiv_id":null,"evidence_quote":"Supplies the Selberg integral background used to produce Jacobi-weight hyperpfaffian evaluations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives Dyson's constant conjecture for circular ensembles, used in the circular hyperpfaffian evaluation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves Dyson's conjecture, anchoring the circular hyperpfaffian evaluation used in the paper."}],"review_version":1}