{"id":"59dd0b66-3e45-4170-a8a8-290fd868af89","arxiv_id":"2502.06143","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For products of bi-invariant random matrices over non-archimedean fields, singular numbers obey an SLLN and CLT with limits given by the corners, extending known type-A/type-C results to all split reductive groups.","lead":"This paper proves that the sizes of products of random matrices over prime-based number systems follow the same Gaussian statistics for every 'split' symmetry type. It gives the first law-of-large-numbers and central-limit-theorem result for such products that covers all root systems at once.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper conflates the coroot lattice Q∨ with the coweight lattice X_*(T)=Hom(G_m,T), so the proof covers only split reductive groups with X_*(T)=Q∨ (e.g. simply connected semisimple), not every split reductive group.","rationale":"The reader identified the coroot/coweight lattice conflation as the weakest assumption, and my independent read agrees. The whole non-asymptotic and asymptotic machinery is indexed by R∨_+ and C[R∨]^W, but for a general split reductive group the Cartan decomposition and Satake isomorphism live on X_*(T). The examples in the paper (SL_{n+1}, Sp_{2n}) are simply connected semisimple, where Q∨=X_*(T), so the gap does not surface. No other step in the proof appears to require a different correction: Lemmas 4.1-4.3 and Theorem 1.5 are consistent once the lattice issue is resolved. Therefore the paper is conditionally correct, either with the stated generality amended to X_*(T)=Q∨ or with R∨ replaced by X_*(T) throughout. The reader's CONDITIONAL verdict already captures this, so no verdict adjustment is needed.","tokens_in":127,"tokens_out":9552,"duration_ms":148575,"concrete_test":"Check Proposition 2.1(1) for G=GL_2(Q_p) and A=diag(π,1). The paper defines R∨=Q∨={(a,-a):a∈Z}, but the standard Cartan decomposition of this A has dominant coweight (1,0), which is not in Q∨. Thus no λ∈R∨_+ satisfies A∈Kπ^λK, directly refuting the claimed setting for all split reductive groups. This simple verification settles whether the coroot/coweight conflation is load-bearing.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 1 item (2) asserts R∨=ΣZα_i^∨ and a 'canonical isomorphism R∨≅Hom(G_m,T)', identifying the coroot lattice with the coweight lattice X_*(T). This equality holds for simply connected semisimple groups such as SL_{n+1} and Sp_{2n}, but it fails for many split reductive groups: for G=GL_n(F), Q∨={(a_i)∈Z^n:Σa_i=0} while X_*(T)=Z^n; the quotient X_*(T)/Q∨ is also nontrivial for PGL_n, SO_{2n+1}, adjoint E6/E7, and others. Consequently Λ={π^λ:λ∈R∨} omits elements of T whose valuation lies in X_*(T)∖Q∨, so Proposition 2.1(1), the Cartan decomposition with λ∈R∨_+, is false as stated. For instance, A=diag(π,1)∈GL_2(Q_p) has singular number (1,0), which is not in Q∨. Because the Satake isomorphism (2.10), the volume formula (3.3), Theorems 1.6 and 1.7, and the asymptotic Theorems 1.3 and 1.5 are all formulated on R∨, they are not established for X_*(T). The paper's examples are exactly the isogeny class where Q∨=X_*(T), so this is an unflagged extra hypothesis rather than a harmless notational choice. Replacing R∨ by X_*(T) throughout, and re-verifying identities such as (2.12), (3.3), and (3.5), would fix the gap and likely preserve the main theorems.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies i.i.d. products of K-bi-invariant random elements in a split reductive group over a non-archimedean local field. The author defines singular numbers via the Cartan decomposition and corners via the Iwasawa decomposition, then derives explicit product and corner transition probabilities from the Satake isomorphism and Hall-Littlewood polynomials. On this basis, the paper claims a strong law of large numbers and a central limit theorem for singular numbers of products, together with a polynomial-in-⟨λ(k),ρ⟩ bound on the discrepancy between singular numbers and corners. The intended contribution is a uniform treatment of all root systems, extending previous results of Van Peski and of the author.","tokens_in":21006,"tokens_out":13808,"duration_ms":134140,"significance":"The results would be a significant unification: they place Gaussian universality for p-adic matrix products on a representation-theoretic footing and would cover exceptional split groups where no linear-algebraic proof is available. The non-asymptotic formulas are explicit and appear to be derived, not fit, from the Satake isomorphism; I see no circularity in the use of principal specialization or Hall-Littlewood structure constants. The main gap is that the paper proves the theorems only for the isogeny class with X_*(T)=Q^∨, while its title and statements claim all split reductive groups; the examples (SL and Sp) are exactly in this class. With the lattice identification corrected and the supporting lemmas repaired, the strategy is plausible, so the manuscript merits revision rather than rejection.","major_comments":[{"comment":"The identification R∨=Hom(G_m,T) is false for a general split reductive group. The coroot lattice Q∨=∑Zα_i^∨ is generally a proper sublattice of the coweight lattice X_*(T). For GL_n, X_*(T)=Z^n but Q∨={a∈Z^n:Σa_i=0}; for GL_2(Q_p), the matrix diag(π,1) has singular number (1,0), which is not in Q∨. Hence Proposition 2.1(1) as stated is false, and the Satake isomorphism (2.10), the volume formula (3.3), and the resulting formulas in Theorems 1.6 and 1.7 are only proved when X_*(T)=Q∨, e.g. for simply connected semisimple groups. The paper's examples SL_{n+1} and Sp_{2n} are exactly in this class, so the claimed generality over all split reductive groups is not established. The natural repair is to take R∨ to be X_*(T), define the positive cone using Q∨_+, define the dominant cone using X_*(T)_+, and re-verify identities such as (2.12), (3.3), and (3.5) in that setting.","section":"Section 1, item (2); Proposition 2.1(1); Theorem 2.15; Eq. (3.3)"},{"comment":"The Weyl denominator is written as δ=∏_{α∈Π+}(e^{α∨/2}-e^{-α∨/2})=e^ρ∏_{α∈Π+}(1-e^{-α∨}). The exponent ∑_{α∈Π+}α∨/2 is the half-sum of positive coroots, not the half-sum of positive roots ρ, except in the simply-laced case. In type C_2, for example, the two half-sums are different. As a consequence, the characters χλ defined by this δ need not lie in C[R∨]^W for non-simply-laced coweight lattices, and the use of (2.4) and (2.6) in Lemma 4.2 is not justified as written. The authors should either redefine the denominator using the half-sum of positive coroots and prove the character basis in the coweight lattice, or explicitly restrict the argument to simply-laced groups.","section":"Definition 2.5; Eqs. (2.4), (2.6); Lemma 4.2"},{"comment":"Lemma 4.3 asserts the existence of a uniform η>0 such that E⟨Cor(A),ρ⟩>η for every nonzero λ∈R∨_+. The proof, however, only shows positivity for each fixed λ and gives an asymptotic statement as ⟨λ,ρ⟩→∞. It does not rule out the infimum over the infinite set of nonzero λ with bounded ⟨λ,ρ⟩ being zero. Proposition 4.4 and Theorem 1.5 require a fixed positive η that is independent of k and λ, so this is a load-bearing gap. A separate lower-bound argument over all nonzero dominant coweights is needed.","section":"Lemma 4.3; Proposition 4.4; Theorem 1.5"}],"minor_comments":[{"comment":"The line 'R∨ = ∑ Zα_i^∨ be the coweights' should say that R∨ is the coroot lattice; the term 'coweights' is reserved for elements of Hom(G_m,T), and this wording contributes to the false identification discussed in Major Comment 1.","section":"Section 1, item (2)"},{"comment":"In the Iwasawa decomposition display, 'πν1,...,ν νn+1' appears to contain a typographical error and should presumably read 'π^{ν1},...,π^{ν_{n+1}}'.","section":"Example 1.1"},{"comment":"The displayed positive octant R∨_0 for Sp_{2n} is written as {λ_1+...+λ_i≥0 for all i}, which is the type A positive cone; for type C the positive cone generated by the simple coroots is {λ_1≥...≥λ_n≥0}.","section":"Example 2.2"},{"comment":"In the second line of (3.2), the expression '∫ cµ(y)cν(πλ y^{-1}) dzdy' retains a leftover dz after the z-integral has been evaluated by G-invariance; the displayed formula should involve only dy.","section":"Eq. (3.2)"},{"comment":"The estimate for ∑ P(C_k) should be framed as a conditional probability given the past value λ(k−1), with the unconditional bound obtained by taking expectation; as written, the conditioning step is skipped.","section":"Proof of Theorem 1.5"},{"comment":"The word 'homogenous' should be 'homogeneous'.","section":"Example 3.3"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about the coroot/coweight identification lands and is the central issue. The paper's proofs are internally coherent for the simply connected semisimple examples it treats, but the claimed scope over all split reductive groups is not supported. The fix appears to be within the manuscript's scope: replace the lattice R∨ by X_*(T), correct the Weyl denominator and the uniform lower bound in Lemma 4.3, and re-verify the displayed identities. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nRead Shen's arXiv:2502.06143. The core idea is real: route the product convolution and corner process through Hall-Littlewood polynomials and the Satake isomorphism, then get SLLN and CLT by showing singular numbers and corners are asymptotically interchangeable. The probabilistic machinery in Sections 3 and 4 is coherent, and the non-asymptotic formulas in Theorems 1.6 and 1.7 are honest consequences of standard identities, not fitted or circular.\n\nThat said, the paper as written overreaches. In Section 1, item (2), it identifies the coroot lattice R∨ with Hom(G_m,T)=X_*(T). This equality only holds for certain isogeny classes, e.g. simply connected semisimple groups. For a general split reductive group, X_*(T) is the coweight lattice P∨ and strictly contains Q∨. For GL_n, diag(π,1) has singular number (1,0), which is not in Q∨. So the Cartan decomposition indexed by R∨ in Proposition 2.1(1) omits valid singular numbers, and every formula built on R∨—including Theorems 1.3, 1.5, 1.6, and 1.7—is not established for the full group. The examples SL_{n+1} and Sp_{2n} are exactly the cases where Q∨=X_*(T), which is why the error is invisible there.\n\nThis is a load-bearing flaw, but it is also addressable. The author should either restrict the statement to simply connected semisimple groups, or redo the Hall-Littlewood and Satake computations with X_*(T) in place of the coroot lattice. The main identities (2.7), (3.3), (3.5) are likely to survive if one works with the full root datum, but that is work the paper has not done.\n\nWhat is genuinely new and valuable: the unified framework covers arbitrary root systems and, once corrected, would give the first CLT for p-adic products outside types A and C, including exceptional groups. The paper deserves refereeing, not desk rejection, but the referee should insist on either a corrected treatment of the coweight lattice or an explicit isogeny-class restriction. I would not cite this version as a theorem; I would cite the corrected version.","headline":"A genuine unified method for Gaussian universality via Satake isomorphism, but the paper's stated claim for all split reductive groups fails because it conflates the coroot lattice with the coweight lattice.","tokens_in":21520,"tokens_out":3570,"would_cite":false,"duration_ms":36559,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["15B52","15B30","60B15"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves Gaussian universality for products of bi-K-invariant random matrices in any split reductive group over a non-archimedean local field, with singular numbers and corners governed by Hall-Littlewood polynomials through the…","keywords":["non-archimedean random matrices","split reductive groups","singular numbers","corners","Hall-Littlewood polynomials","Satake isomorphism","Gaussian universality","strong law of large numbers"],"falsifier":"Take a split reductive group whose cocharacter lattice strictly contains the coroot lattice, such as $\\mathrm{PGL}_n(F)$, choose a bi-K-invariant distribution supported on a double coset indexed by a coweight not in $R^\\vee_+$, and test whether the transition probability formula of Theorem 1.6 reproduces the direct double-coset count for a small residue field of size $q$; a mismatch at any $q$ would show the Hall-Littlewood machinery only applies to the coroot-indexed isogeny class.","tokens_in":20367,"feed_emoji":"🎲","tokens_out":8888,"duration_ms":73999,"temperature":0.7,"pith_summary":"The paper aims to show that, in any split reductive group over a non-archimedean local field, the singular numbers of products of random bi-K-invariant matrices are asymptotically Gaussian after centering and scaling, with the same limit structure regardless of the root system. The proof routes the Cartan and Iwasawa decompositions through Hall-Littlewood polynomials via the Satake isomorphism, so the product process and the corner process are determined by the same algebraic data. The key probabilistic step is a separation bound: the singular numbers and the sum of corners of a long product differ by an amount smaller than any positive power of the singular-number size. If the argument is correct, the Gaussian universality previously known for GL_n and Sp_{2n} holds uniformly for all split reductive groups, including exceptional types.","feed_headline":"Satake isomorphism yields Gaussian limits for p-adic matrix products","feed_subtitle":"Singular numbers and corners in every split reductive group obey one Hall-Littlewood law, yielding SLLN and CLT.","key_machinery":"The mechanism is the Satake isomorphism between the spherical Hecke algebra of $G$ and the $W$-invariant Laurent polynomials, sending the double coset $c_\\lambda$ to $q^{\\langle\\lambda,\\rho\\rangle}P_\\lambda(q^{-1})$, with $P_\\lambda(t)$ the Hall-Littlewood polynomial attached to the dominant coweight $\\lambda$. Combined with the principal specialization formula $P_\\lambda(\\theta;t)=W(t)/W_\\lambda(t)\\,t^{-\\langle\\lambda,\\rho\\rangle}$, this converts product convolution of double cosets into multiplication of Hall-Littlewood polynomials and converts the corner distribution into the expansion coefficients $u_{\\lambda,\\nu}(t)$ of $P_\\lambda$ in the monomial basis. The volume formula $V(K\\pi^\\lambda K)=q^{2\\langle\\lambda,\\rho\\rangle}W(t)/W_\\lambda(t)$ then gives the explicit transition probabilities in Theorem 1.6 and Theorem 1.7, and the positivity and growth estimates for the $u_{\\lambda,\\nu}(q^{-1})$ coefficients drive the Borel-Cantelli separation bound.","core_discovery":"On the paper's own terms, the central discovery is Theorem 1.3: for i.i.d. K-bi-invariant random matrices $A_1,A_2,\\dots$ in a split reductive group $G$ over a non-archimedean local field, the singular numbers $\\lambda(k)=\\mathrm{SN}(A_1\\cdots A_k)$ satisfy $\\lambda(k)/k \\to E\\,\\mathrm{Cor}(A_1)$ almost surely and $(\\langle\\lambda(k),\\alpha_i\\rangle-kE\\langle\\mathrm{Cor}(A_1),\\alpha_i\\rangle)/\\sqrt{k} \\Rightarrow N(0,\\Sigma)$, where $\\Sigma$ is the covariance matrix of the corner coordinates. The engine is Theorem 1.6 and Theorem 1.7, which express the transition law of singular numbers and the conditional law of corners in terms of Hall-Littlewood structure coefficients, plus Theorem 1.5, which shows that $0\\le\\langle\\lambda(k)-\\nu(k),\\rho\\rangle\\le\\langle\\lambda(k),\\rho\\rangle^\\varepsilon$ eventually almost surely. The upshot is that the singular numbers of a long product are asymptotically interchangeable with the sum of independent corner increments, so classical LLN/CLT applies.","pith_inferences":["If the separation bound holds with only a divergence condition on $P(\\mathrm{SN}(A_k)\\neq 0)$, as the author conjectures, the SLLN and CLT should extend to non-identical bi-K-invariant sequences, making the Gaussian limit a genuine universality phenomenon rather than an i.i.d. artifact.","The same Satake dictionary may yield explicit Markov transition probabilities for random walks on Bruhat-Tits buildings of exceptional type, where linear-algebraic definitions of corners do not naturally apply.","The covariance matrix $\\Sigma$ can be singular when the corners concentrate on a proper sublattice; in such cases the multivariate CLT holds with a degenerate limit, and tests comparing projected coordinates would be more sensitive than the full vector."],"forward_implications":["Theorem 1.3 gives a strong law of large numbers and a central limit theorem for singular numbers of products in every split reductive group; in type A it reduces to the previously known GL_{n+1} Gaussian universality, and in type C to the Sp_{2n} result.","Theorem 1.5 implies that for bi-K-invariant products, singular numbers and corners are asymptotically interchangeable: the discrepancy $\\langle\\lambda(k)-\\nu(k),\\rho\\rangle$ is eventually smaller than any positive power of $\\langle\\lambda(k),\\rho\\rangle$.","Because the corner sequence has independent increments in distribution, the asymptotic means and covariance matrix of singular numbers are explicitly $E\\langle\\mathrm{Cor}(A_1),\\alpha_i\\rangle$ and $\\mathrm{Cov}(\\langle\\mathrm{Cor}(A_1),\\alpha_i\\rangle,\\langle\\mathrm{Cor}(A_1),\\alpha_j\\rangle)$, computable from the one-step corner law.","The framework provides a unified treatment for exceptional root systems, where direct linear-algebraic methods are unavailable.","The strong law in Theorem 1.3 gives the Lyapunov exponents $\\lim_{k\\to\\infty}\\langle\\lambda(k),\\alpha_i\\rangle/k=E\\langle\\mathrm{Cor}(A_1),\\alpha_i\\rangle$ in the non-archimedean setting."],"supporting_citations":[{"why":"Supplies the GL_n(F) Gaussian universality law and the Hall-Littlewood transition formula that this paper generalizes from type A to all root systems.","marker":"[30]"},{"why":"Establishes the corner-based proof of Gaussian universality for SL_{n+1} and Sp_{2n}, which this paper recovers from the Satake perspective.","marker":"[28]"},{"why":"Provides the Hall-Littlewood spherical-function theory, including the orbit volume and principal specialization formulas used in the transition probabilities.","marker":"[22]"},{"why":"Gives the Cartan and Iwasawa decompositions and the inequalities SN(A)≥Cor(A) and SN(AB)≤SN(A)+SN(B) that control the singular-number/corner gap.","marker":"[8]"},{"why":"Supplies the root-system facts on dominance and Weyl-group ordering used in the corner-law estimates.","marker":"[6]"},{"why":"Provides the principal specialization identity from which $P_\\lambda(\\theta;t)=W(t)/W_\\lambda(t)\\,t^{-\\langle\\lambda,\\rho\\rangle}$ is derived.","marker":"[27]"}],"fun_headline_variants":["Satake isomorphism yields Gaussian limits for p-adic products","Hall-Littlewood law drives Gaussian asymptotics in p-adic products","Universal Gaussian limits for p-adic split reductive groups","SLLN and CLT for p-adic matrix products via Satake","From Satake to Gaussian: p-adic matrix products obey SLLN and CLT"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the paper's identification of the cocharacter lattice of the maximal torus with the coroot lattice $R^\\vee$; this equality holds for some isogeny classes but not for groups such as PGL_n or odd special orthogonal groups, whose coweight lattice strictly contains the coroot lattice.","fun_headline_variants_meta":{"raw":{"variants":["Satake isomorphism yields Gaussian limits for p-adic products","Hall-Littlewood law drives Gaussian asymptotics in p-adic products","Universal Gaussian limits for p-adic split reductive groups","SLLN and CLT for p-adic matrix products via Satake","From Satake to Gaussian: p-adic matrix products obey SLLN and CLT"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002045,"raw_usage":{"total_tokens":7972,"prompt_tokens":957,"completion_tokens":7015,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":573,"completion_tokens_details":{"reasoning_tokens":6920}},"tokens_in":573,"tokens_out":7015,"duration_ms":43742,"temperature":1.0,"reasoning_tokens":6920,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T16:38:00.758127+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a split reductive group whose cocharacter lattice strictly contains the coroot lattice, such as $\\mathrm{PGL}_n(F)$, choose a bi-K-invariant distribution supported on a double coset indexed by a coweight not in $R^\\vee_+$, and test whether the transition probability formula of Theorem 1.6 reproduces the direct double-coset count for a small residue field of size $q$; a mismatch at any $q$ would show the Hall-Littlewood machinery only applies to the coroot-indexed isogeny class.","supporting_citations":[{"cited_title":"Limits and fluctuations of p-adic rando m matrix products","cited_arxiv_id":null,"evidence_quote":"Supplies the GL_n(F) Gaussian universality law and the Hall-Littlewood transition formula that this paper generalizes from type A to all root systems."},{"cited_title":"Spherical functions on a p-adic ch evalley group","cited_arxiv_id":null,"evidence_quote":"Provides the Hall-Littlewood spherical-function theory, including the orbit volume and principal specialization formulas used in the transition probabilities."},{"cited_title":"Groupes r´ eductifs s ur un corps local: I","cited_arxiv_id":null,"evidence_quote":"Gives the Cartan and Iwasawa decompositions and the inequalities SN(A)≥Cor(A) and SN(AB)≤SN(A)+SN(B) that control the singular-number/corner gap."},{"cited_title":"Lie groups and Lie algebras , volume 1","cited_arxiv_id":null,"evidence_quote":"Supplies the root-system facts on dominance and Weyl-group ordering used in the corner-law estimates."},{"cited_title":"Compact lie groups","cited_arxiv_id":null,"evidence_quote":"Provides the principal specialization identity from which $P_\\lambda(\\theta;t)=W(t)/W_\\lambda(t)\\,t^{-\\langle\\lambda,\\rho\\rangle}$ is derived."}],"review_version":1}