Pith. sign in

REVIEW 3 major objections 4 minor 32 references

Distribution of Eigenvalues of Random Real Symmetric Block Matrices

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Mixing two random matrix ensembles yields a new universal law

desk verdict Novel block-ensemble construction with a solid d=1 analysis, but the general-d universality theorem rests on an unproved variance bound. read the letter →

arxiv 1908.03834 v1 pith:BE4DAZPY submitted 2019-08-11 math.PR math-phmath.MPmath.NT

Keller Blackwell , Neelima Borade , Charles Devlin VI , Noah Luntzlara , Renyuan Ma , Steven J. Miller , Mengxi Wang , Wanqiao Xu This is my paper · ORCID
classification math.PRmath-phmath.MPmath.NT MSC 15A5260F9962H10
keywords randommatrixtheorylimitingspectralmeasureToeplitzmatricessemicirclelawGaussiandistributionmethodofmomentsblockuniversal
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper introduces the 'disco' construction: starting from two real symmetric matrix ensembles, it alternates copies of the first and second in a block matrix, then repeats the pattern at larger scales to form D_d. It claims that for each fixed d, the normalized eigenvalue distribution of D_d converges weakly to a new distribution that is universal, meaning it does not depend on the entry distribution p, only on the two component ensembles; the d=1 case is also proved almost surely. In the central example, palindromic Toeplitz matrices (whose limit is Gaussian) are mixed with full real symmetric matrices (whose limit is semicircle), and the resulting d=1 law has moments squeezed between the two component laws, such as an exact fourth moment of 9/4, with unbounded support. As d grows, the law approaches that of the second ensemble with moment discrepancies of order $2^{{-d}}$. A sympathetic reader should care because this gives a concrete mechanism for interpolating between any two random matrix spectral laws.

What carries the argument

The central object is the block matrix $$D_1(A,B)=\begin{bmatrix} A&B\\B&A\end{bmatrix},$$ whose iterated refinement gives D_d. The key identity is $$E[\operatorname{Tr}(D_1^k)]=E[\operatorname{Tr}((A+B)^k+(A-B)^k)],$$ which reduces moment calculations to pairing entries of A and B. In the limit, A-entries may be paired in any of the (2k-1)!! Gaussian ways, B-entries must be paired in the non-crossing Catalan ways of the semicircle, and mixed pairings survive only when no b-pair crosses an a-pair or another b-pair. The paper encodes these restrictions as a spanning-tree formula P(α,β) counting allowed pairings of 2α red (A) and 2β blue (B) points on a circle. A generalized Hölder trace inequality bounds arbitrary mixed products by powers of the component traces, and a degree-of-freedom count shows the variance of the empirical moments vanishes, enabling Markov's method of moments.

What would settle it

Enumerate the index-pairing equations in E[M_k(D_d,N)^2] - E[M_k(D_d,N)]^2 for a small case such as d=2, k=4 and look for one crossover configuration whose equations retain 2k+2 degrees of freedom. If such a configuration exists and contributes with full $N^{{2k+2}}$ weight, the variance does not vanish and weak convergence fails; a direct enumeration or a small-N simulation comparing the variance's N-scaling would settle it.

Watch

Extended reading notes

Core claim

For fixed d, take A from one real symmetric ensemble with limiting spectral measure μ_A and take B_1, B_2, ... from another ensemble with limit μ_B, all entries i.i.d. from a distribution p with mean 0, variance 1, and finite higher moments. Form the d-disco block matrix D_d by placing A on the diagonal blocks and the B_i on off-diagonal blocks in the recursive ABBA pattern. The paper's main theorem states that the normalized empirical spectral measure of D_d converges weakly, as the base size N tends to infinity, to a new probability measure μ_d that depends only on the two ensembles and on d, not on p. Moreover, in the PST/RS case the moments of μ_d are finite and lie between the moments of the two component measures, and μ_d has unbounded support. As d tends to infinity, μ_d converges weakly to μ_B, the limiting measure of the B ensemble, with moment discrepancies of order $2^{{-d}}$.

Load-bearing premise

The whole argument assumes that in the variance expansion every 'crossover' configuration, where indices from the two copies of the matrix pair with each other, loses at least one degree of freedom, so those terms vanish as N grows; the general-d proof borrows this from earlier work rather than proving it in the disco setting.

Editorial extensions

If this is right

  • For PST and RS components, every even moment of the d=1 law satisfies S_{2k} ≤ M_{2k}(D_1) < G_{2k}, and M_{2k}/G_{2k} → 0, so the hybrid is genuinely distinct from both Gaussian and semicircle.
  • The construction applies to any pair of real symmetric sub-ensembles with finite moments: finite d yields a new universal law, and taking d → ∞ recovers the B ensemble's law with O(2^{-d}) moment convergence.
  • The d=1 law has unbounded support even though it is sandwiched between the bounded semicircle and unbounded Gaussian laws; its moments grow slower than Gaussian moments.
  • Exact low moments provide quantitative benchmarks: M_2 = 1, M_4 = 9/4, M_6 = 7, and M_8 = 27.5 in the PST/RS case.
  • The paper's moment bounds support its conjecture that the normalized eigenvalue spacings of D_1 lie between those of the component ensembles, a statement left open.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the main theorem holds, the d-disco supplies a discrete interpolation path between any two spectral laws; one could view d as a 'mixing depth' and define an interpolating family of laws by moment interpolation, which the paper does not attempt.
  • The spanning-tree formula P(α,β) suggests the moments of the d=1 law may admit a generating-function description as weighted plane trees; extracting that closed form would give exact higher moments rather than bounds.
  • The O(2^{-d}) convergence rate suggests a quantitative notion of distance between ensembles: two ensembles are close if their disco at moderate d already matches B's moments to prescribed precision, a property that could be tested numerically for several entry distributions.
  • Because the construction only needs finite moments and mean-zero variance-one entries, it may extend to non-symmetric or complex ensembles, where the Gaussian and semicircle roles are replaced by other universal laws; the paper does not pursue this.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper introduces a 'disco' construction D_d(A,B) that intertwines two random real symmetric matrix ensembles in block form, with D_1 = [[A,B],[B,A]] and higher d built recursively. The main results are: (i) for d=1, with A a palindromic symmetric Toeplitz (PST) matrix and B a real symmetric (RS) matrix with i.i.d. entries from a mean-zero, variance-one law p, the normalized empirical spectral measure converges weakly and almost surely to a new universal measure whose even moments lie between the semicircle and Gaussian moments; (ii) for finite d, D_d(A,B) converges weakly to a new universal distribution independent of p; (iii) as d tends to infinity, the limiting spectral distribution converges to that of the B ensemble. The proof uses the method of moments, expansion of Tr((B_d+C_d)^m), a generalized Hölder trace inequality, and combinatorial counting of chord pairings (Theorem 3.3).

Significance. If the main theorem (Theorem 7.3) is correct, the construction provides a mechanism for interpolating between arbitrary random-matrix spectral laws, which is a novel and potentially useful idea with connections to number-theoretic convolution constructions. The paper's strengths include explicit low-moment computations for the d=1 PST/RS case (M_4=9/4, M_6=7, M_8=27.5), a general combinatorial formula for contribution counts (Theorem 3.3), and a clean generalized Hölder bound for mixed products (Theorem 4.4). The upper and lower moment bounds for d=1 are established with explicit inequalities. However, the central convergence claims for general d rest on deferred degree-counting arguments that are not supplied in the disco setting, and Theorem 6.1 has a serious proof gap and a missing independence hypothesis.

major comments (3)
  1. [Section 7, Theorem 7.3 and Eq. (7.16)] The proof of the variance bound for finite d is not supplied. The text states that (7.16) follows by 'arguments analogous to those of Theorem 6.2', but Theorem 6.2's proof in turn concludes that '[t]he generality of the arguments in [HM] are immediately applicable here' without giving the degree-counting argument in the disco setting. This matters because in D_d the same random variable (an entry of some B_i) appears at many global positions: for d=1, the variable B_{u,v} appears in both off-diagonal blocks, and more generally each B_i is repeated throughout the block structure. A crossover between two occurrences of such a repeated entry can be satisfied by an identity of the underlying variable, and the induced constraint on the global summation indices may be weaker than the asserted 'loss of at least one degree of freedom'. No counting lemma for these configurations is provided, so the claim that the variance tends to zero, and hence the universality conclusion, is not established by the manuscript.
  2. [Section 6, Theorem 6.1 and Eq. (6.7)] The moment identity M_k(D_1(X,B_1)) = M_k(X) is not proven by the argument given. Equation (6.7) asserts that each pairing configuration of D_1 entries gives 2^k configurations from X, but the expansion (6.4)-(6.5) contains mixed products such as Tr(X^2 B_1^2), which contribute at the same order in N; indeed the authors' own calculation (2.29) for the PST/RS case has E[Tr(A^2 B^2)] = N^3. The factorization in (6.7) therefore does not follow from the displayed counting. Moreover, the theorem does not state whether X and B_1 are independent; if X = B_1 as a matrix, D_1 = [[X,X],[X,X]] has normalized 2kth moment 2^{k-1} M_{2k}(X), not M_{2k}(X). Since Theorem 6.2 and Corollary 6.3 depend on Theorem 6.1, this gap is load-bearing for the general-d claims.
  3. [Section 2.5, Eq. (2.61)] The almost sure convergence claim is also deferred. The proof says that 'the analogues of the proofs in [HM] hold in the disco case as well' and that crossovers involving B entries lose 'more degrees of freedom', but no computation is given. The same repeated-variable issue as in Theorem 7.3 arises already for D_1, since B appears in both off-diagonal blocks; the [HM] Toeplitz counting assumes each random variable occurs in a single pattern of positions, so the transfer is not automatic. Without a proof of the fourth-moment bound (2.61), the almost sure statement in the abstract and Section 2.5 is unsupported.
minor comments (4)
  1. [Section 4, Theorem 4.4 proof] The proof says 'Taking p_i = 1/K in Theorem 4.3', but this choice violates the hypothesis sum p_i^{-1}=1 for the n factors. The claimed inequality is the standard Hölder bound and is correct if one takes p_i = K/I_i for A-factors and p_i = K/J_i for B-factors, so this is a fixable presentation error.
  2. [Definition 1.1 and Section 7] The dimension of D_d is written inconsistently as 2dN in several places, while the iterative construction and equations such as (7.7)-(7.8) suggest the size is 2^d N. Please standardize the notation throughout.
  3. [Section 2.4, Lemma 2.15] In the proof of Lemma 2.15, the phrase 'a pair of b's cross over' is used loosely; for symmetric matrices, b_{s,s+1} = b_{t,t+1} means {s,s+1} = {t,t+1}, not a crossover in the sense of distinct index pairs. Clarify the terminology for the b-pairings.
  4. [Section 8, Conjecture 8.1] The numerical data in Table 3 support the conjecture for the displayed moments, but the table heading says 'supporting (8.1)' while Conjecture 8.1 is numbered 8.1; update the reference to avoid ambiguity.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular reduction of the central claims; the finite-d variance step rests on an unproved analogy to [HM], which is a proof gap rather than a circular reuse of the target result.

full rationale

The derivation chain is not circular. The component inputs (PST moments from [MMS], RS/semicircle moments from Wigner, and the Diophantine degree-counting toolkit from [HM]) are established external results used as ingredients, not restatements of the paper's target theorems. The d=1 moment computations in Section 2 and the variance lemmas (Lemmas 2.15 and 2.16) are carried out directly in the paper by explicit degree-of-freedom counting, and the bounds in Sections 4-5 use the generalized Holder trace inequality rather than assuming the conclusion. Theorem 7.2's d-to-infinity convergence is a direct estimate using those bounds. The one genuinely load-bearing soft spot is Theorem 7.3's finite-d variance claim, which is dispatched by 'arguments analogous to those of Theorem 6.2' (Section 7) and, in turn, by 'the generality of the arguments in [HM] are immediately applicable here' (Section 6). Because [HM] is a prior paper by one of the present authors and its arguments are asserted, not re-proved, for the disco setting, this is a missing-support and correctness concern, not a definitional or fitted-input circularity: no equation in the paper defines an input as the predicted quantity, and no fitted parameter is renamed as a prediction. The score 1 reflects the presence of self-citations in the variance argument without treating them as circular reductions.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard moment-convergence machinery, the known Gaussian and semicircle limits of the two component ensembles, the i.i.d. entry assumption, and the generalized Hölder inequality. No free parameters are introduced. The paper's own proof is self-contained apart from the cited component results.

assumptions (5)
  • standard math General Moment Convergence Theorem with uniqueness via Carleman's condition
    Used to conclude weak convergence from convergence of moments (Theorems 2.17 and 6.2). The stated Theorem 1.3 only covers the normal case; the paper relies on the general form.
  • domain assumption Limiting spectral measure of the symmetric palindromic Toeplitz ensemble is the standard Gaussian
    Quoted from [MMS] and used to compute moments of the A component throughout Section 2.
  • domain assumption Limiting spectral measure of the full real symmetric ensemble is the semicircle law
    Quoted from [Wig2] and used for the B component throughout Section 2.
  • domain assumption Component entries are i.i.d. with mean 0, variance 1, and finite higher moments
    Assumed at the start of Sections 1.2 and 2.1; ensures vanishing of unpaired and high-multiplicity terms in the trace expansions.
  • standard math Generalized Hölder trace inequality
    Proved in the appendix using von Neumann's trace inequality and Hölder's inequality; used in Lemma 5.2 and Theorem 7.2 to bound mixed products.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distribution of Eigenvalues of Random Real Symmetric Block Matrices." pith.science (2026). https://pith.science/paper/BE4DAZPY

@misc{pith2026190803834,
  author       = {Pith},
  title        = {Pith review of: Distribution of Eigenvalues of Random Real Symmetric Block Matrices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BE4DAZPY}},
  note         = {Machine review of arXiv:1908.03834}
}
abstract

Random Matrix Theory (RMT) has successfully modeled diverse systems, from energy levels of heavy nuclei to zeros of $L$-functions. Many statistics in one can be interpreted in terms of quantities of the other; for example, zeros of $L$-functions correspond to eigenvalues of matrices, and values of $L$-functions to values of the characteristic polynomials. This correspondence has allowed RMT to successfully predict many number theory behaviors; however, there are some operations which to date have no RMT analogue. The motivation of this paper is to try and find an RMT equivalent to Rankin-Selberg convolution, which builds a new $L$-functions from an input pair. For definiteness we concentrate on two specific families, the ensemble of palindromic real symmetric Toeplitz (PST) matrices and the ensemble of real symmetric (RS) matrices, whose limiting spectral measures are the Gaussian and semicircle distributions, respectively; these were chosen as they are the two extreme cases in terms of moment calculations. For a PST matrix $A$ and a RS matrix $B$, we construct an ensemble of random real symmetric block matrices whose first row is $\lbrace A, B \rbrace$ and whose second row is $\lbrace B, A \rbrace$. By Markov's Method of Moments, we show this ensemble converges weakly and almost surely to a new, universal distribution with a hybrid of Gaussian and semicircle behaviors. We extend this construction by considering an iterated concatenation of matrices from an arbitrary pair of random real symmetric sub-ensembles with different limiting spectral measures. We prove that finite iterations converge to new, universal distributions with hybrid behavior, and that infinite iterations converge to the limiting spectral measures of the component matrices.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 29 canonical work pages

  1. [1]

    Bai, Methodologies in Spectral Analysis of Large Dimensional Random Matrices, A Review, Statistica Sinica 9, 1999, 611--677

    Z. Bai, Methodologies in Spectral Analysis of Large Dimensional Random Matrices, A Review, Statistica Sinica 9, 1999, 611--677

    work page 1999

  2. [2]

    Barrett, F

    O. Barrett, F. W. K. Firk, S. J. Miller, and C. Turnage-Butterbaugh, From Quantum Systems to L -Functions: Pair Correlation Statistics and Beyond, in Open Problems in Mathematics (editors John Nash Jr. and Michael Th. Rassias), Springer-Verlag, 2016, pages 123--171

    work page 2016

  3. [3]

    A. Bose, S. Chatterjee, and S. Gangopadhyay, Limiting Spectral Distributions of Large Dimensional Random Matrices, J. Indian Statist. Assoc. 41 (2003), 221--259

    work page 2003

  4. [4]

    W. Bryc, A. Dembo, and T. Jiang, Spectral Measure of Large Random Hankel, Markov and Toeplitz Matrices, Annals of Probability 34 (2006), no. 1, 1--38

    work page 2006

  5. [5]

    Burkhardt, P

    P. Burkhardt, P. Cohen, J. Dewitt, M. Hlavacek, S. J. Miller, C. Sprunger, Y. N. Truong Vu, R. Van Peski, and Kevin Yang, Random Matrix Ensembles with Split Limiting Behavior (with an appendix joint with Manuel Fernandez and Nicholas Sieger), Random Matrices: Theory and Applications 7 (2018), no. 3, 1850006 (30 pages), DOI: 10.1142/S2010326318500065

    work page doi:10.1142/s2010326318500065 2018

  6. [6]

    J. B. Conrey, L -Functions and Random Matrices, Mathematics unlimited -- 2001 and Beyond, Springer-Verlag, Berlin, 2001, 331--352

    work page 2001

  7. [7]

    Cheung, Generalizations of H\"older's Inequality, International Journal of Mathematics and Mathematical Sciences, 26, 2001, no

    W. Cheung, Generalizations of H\"older's Inequality, International Journal of Mathematics and Mathematical Sciences, 26, 2001, no. 1

    work page 2001

  8. [8]

    R. C. Chen, Y. H. Kim, J. D. Lichtman, S. J. Miller, S. Sweitzer, E. Winsor, Spectral Statistics of Non-Hermitian Random Matrix Ensembles, Random Matrices: Theory and Applications 8 (2019), no. 1, 1950005 (40 pages)

    work page 2019

Show all 32 references
  1. [9]

    Due\ nez and S

    E. Due\ nez and S. J. Miller, The low lying zeros of a GL (4) and a GL (6) family of L -functions , Compositio Mathematica 142 (2006), no. 6, 1403--1425

    2006

  2. [10]

    Due\ nez and S

    E. Due\ nez and S. J. Miller, The effect of convolving families of L -functions on the underlying group symmetries, Proceedings of the London Mathematical Society, 2009; doi: 10.1112/plms/pdp018

    2009 doi

  3. [11]

    P. D. T. A. Elliot, Probabilistic Number Theory I: Mean Value Theorems, Springer-Verlag, Berlin-New York, 1980

    1980

  4. [12]

    Firk and S

    F. Firk and S. J. Miller, Nuclei, Primes and the Random Matrix Connection, Symmetry, 2009, 1, 64--105

    2009

  5. [13]

    Goldmakher, C

    L. Goldmakher, C. Khoury, S. J. Miller, and K. Ninsuwan, The Expected Eigenvalue Distribution of Large, Weighted d -regular Graphs, Random Matrices: Theory and Applications. 3 (2014), no. 4, 1450015

    2014

  6. [14]

    R. A. Horn and C. R. Johnson, Topics in Matrix Analysis, Cambridge University Press, 1991, 172 -- 177

    1991

  7. [15]

    Hammond and S

    C. Hammond and S. J. Miller, Eigenvalue Spacing Distribution for the Ensemble of Real Symmetric Toeplitz Matrices, Journal of Theoretical Probability 18, 2005, no. 3, 537--566

    2005

  8. [16]

    Hejhal, On the triple correlation of zeros of the zeta function, Internat

    D. Hejhal, On the triple correlation of zeros of the zeta function, Internat. Math. Res. Notices (1994), no. 7, 294--302

    1994

  9. [17]

    Iwaniec and E

    H. Iwaniec and E. Kowalski, Analytic Number Theory, AMS Colloquium Publications, Vol. 53, AMS, Providence, RI, 2004

    2004

  10. [18]

    J. L. W. V. Jensen Sur les fonctions convexes et les inégalités entre les valeurs moyennes, Acta Math 30, 1906, 175--193

    1906

  11. [19]

    Jakobson, S

    D. Jakobson, S. D. Miller, I. Rivin, and Z. Rudnick, Eigenvalue spacings for regular graphs. Pages 317--327 in Emerging Applications of Number Theory (Minneapolis, 1996), The IMA Volumes in Mathematics and its Applications, Vol. 109, Springer, New York, 1999

    1996

  12. [20]

    G. Kopp, M. Kolo g lu, S. J. Miller, F. Strauch, and W. Xiong, The Limiting Spectral Measure for Ensembles of Symmetric Block Circulant Matrices, Journal of Theoretical Probability 4 (2013), 1020--1060

    2013

  13. [21]

    A. J. Laub, Matrix Analysis for Scientists and Engineers, AMS-SIAM, 2005, 139 -- 150

    2005

  14. [22]

    McKay, The expected eigenvalue distribution of a large regular graph, Linear Algebra Appl

    B. McKay, The expected eigenvalue distribution of a large regular graph, Linear Algebra Appl. 40 (1981), 203--216

    1981

  15. [23]

    Massey, S

    A. Massey, S. J. Miller, and J. Sinsheimer, Distribution of Eigenvalues of Real Symmetric Palindromic Toeplitz Matrices and Circulant Matrices, Journal of Theoretical Probability 20, 2007, no. 3, 637--662

    2007

  16. [24]

    Mehta, Random Matrices, 2 nd edition, Academic Press Inc., Boston, 1991

    M. Mehta, Random Matrices, 2 nd edition, Academic Press Inc., Boston, 1991

    1991

  17. [25]

    Morrison, Modeling Convolutions of L -Functions, Senior Thesis Williams College 2010 (advisor Steven J

    R. Morrison, Modeling Convolutions of L -Functions, Senior Thesis Williams College 2010 (advisor Steven J. Miller), https://arxiv.org/abs/1011.0229

    2010 arXiv

  18. [26]

    Rudnick, and P

    Z. Rudnick, and P. Sarnak, Zeros of principal L -functions and random matrix theory, Duke Math. J. 81 (1996), 269--322

    1996

  19. [27]

    Sarnak, S

    P. Sarnak, S. W. Shin, and N. Templier, Families of L -functions and their symmetry, in Families of automorphic forms and the trace formula (editors Werner M\"uller, Sug Woo Shin and Nicolas Templier), Simons Symp., Springer, 2016, pages 531--578

    2016

  20. [28]

    Ya. G. Sinai and A. B. Soshnikov, A Refinement of Wigner's Semicircle Law in a Neighborhood of the Spectrum Edge for Random Symmetric Matrices, Functional Analysis and Its Applications 32, 1998, no. 2, 114--130

    1998

  21. [29]

    Takacs, A Moment Convergence Theorem, The American Mathematical Monthly 98 (Oct., 1991), no

    L. Takacs, A Moment Convergence Theorem, The American Mathematical Monthly 98 (Oct., 1991), no. 8, 742--746

    1991

  22. [30]

    von Neumann, Some matrix-inequalities and metrization of matrix-space, Tomsk Univ

    J. von Neumann, Some matrix-inequalities and metrization of matrix-space, Tomsk Univ. Rev. 1, 1937, 286 -- 300. (In: A. H. Taub, Ed., John von Neumann Collected Works, Vol. IV, Pergamon, Oxford, 1962, pp. 205-218.)

    1937

  23. [31]

    Wigner, On the statistical distribution of the widths and spacings of nuclear resonance levels, Proc

    E. Wigner, On the statistical distribution of the widths and spacings of nuclear resonance levels, Proc. Cambridge Philo. Soc. 47, 1951, 790--798

    1951

  24. [32]

    Wigner, Statistical Properties of real symmetric matrices, Canadian Mathematical Congress Proceedings, University of Toronto Press, 1957, 174--184.'

    E. Wigner, Statistical Properties of real symmetric matrices, Canadian Mathematical Congress Proceedings, University of Toronto Press, 1957, 174--184.'

    1957

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.