{"id":"6a584b34-981d-45a9-a11f-bc605bbacf3b","arxiv_id":"2505.14198","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For balanced Polya urns, the p-th moments of centered ball counts obey explicit polynomial-in-n bounds matching the urn's dominant eigenvalues, yielding moment convergence whenever a central limit theorem is known.","lead":"This paper proves bounds on the spread of the ball-count distribution in a classic random process called a Polya urn, where balls are drawn and replaced according to color-dependent rules. The result helps show when limit theorems for such urns automatically include convergence of all moments, which matters for anyone using urn models in applied probability.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified","rationale":"I re-read the proof of Theorem 3.1 (Section 4) and its supporting lemmas. The central representation (4.14)-(4.16) is correct: balance gives w_n = w0 + nb, so F_{ℓ,n} are nonrandom; Y_ℓ are martingale differences; (4.8)-(4.10) give the Lp bound needed for Lemma 5.1. The matrix estimates in Lemma 6.1 follow from the Jordan decomposition and scalar-product bounds; Lemma 6.2's summation splits correctly into the three eigenvalue regimes, including the critical log power 1 + 2ν_λ. Theorems 3.2 and A.1 use standard uniform-integrability arguments; no hidden circularity. The balance condition is the essential structural assumption, but it is explicit (PU4) and the paper honestly states the unbalanced case as open (Problem 1.1). I therefore find no load-bearing objection; the ACCEPT verdict stands.","tokens_in":13046,"tokens_out":21929,"duration_ms":206601,"concrete_test":"Verify the critical-case exponent analytically: for γ = 1/2 and ν > 0, directly evaluate S_n = Σ_{i=1}^n (1 + log(n/i))^{2ν} / i and confirm S_n = O((log n)^{1+2ν}); then re-derive Theorem 3.3's log power ν + 1/2 from Lemma 5.1. If the log power differs, the critical regime bound in (3.2) would be wrong.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the proof line by line, I find no internal inconsistency or gap that threatens the central claim. Theorem 3.1/3.3 rest on four ingredients: (i) balance makes w_n = w0 + nb deterministic; (ii) the recurrence (4.11) gives the deterministic product representation (4.14)-(4.16); (iii) Y_n is a martingale difference with bounded Lp norms by (4.8)-(4.10); (iv) Lemma 5.1 plus the matrix estimates in Lemmas 6.1-6.2 yield the three regimes. Each of these steps checks out. The balance condition (PU4) is genuinely load-bearing: without it, w_n is random, the matrices F_{ℓ,n} are not nonrandom, and the martingale-coefficient argument fails; the paper explicitly flags unbalanced urns as open (Problem 1.1). Since this is an explicit standing assumption rather than a hidden one, it does not constitute a defect in the theorem as stated. The only residual risk is the cited Lemma 2.2, but the theorems assume λ1 = b directly, so the proof does not depend on the lemma's correctness.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies balanced generalized Pólya urns in discrete time. Theorem 3.1 gives, under tenability, balance, and λ1=b, upper bounds for ||X_n - E X_n||_p for p≥2 in three regimes determined by Re λ2: C_p n^{1/2}, C_p n^{1/2}(log n)^{ν2+1/2}, or C_p n^{Re λ2/λ1}(log n)^{ν2}. Theorem 3.3 refines this to individual spectral projections Pλ(X_n - E X_n), and Theorem 3.4 shows that the λ1 component has zero centered fluctuation when λ1 is simple. Theorem 3.2 converts the bounds into convergence of all moments under previously known asymptotic normality results; Appendix A provides a uniform-integrability version when only a single p≥2 is assumed. The proofs use a martingale difference decomposition Y_n, the balance condition to obtain a deterministic recurrence and matrix-product representation F_{ℓ,n}, Burkholder-type inequalities (Lemma 5.1), and spectral estimates for those matrix products (Lemmas 6.1 and 6.2).","tokens_in":13276,"tokens_out":12917,"duration_ms":133270,"significance":"If the result holds, it is a useful unification: it recovers and extends earlier moment bounds for balanced urns with a simpler proof, and it upgrades known central limit theorems to moment convergence in the small-urn and critical regimes. The proof is checkable and essentially self-contained, with complete proofs included for the key lemmas. All constants are unspecified but not fitted, and the load-bearing structural assumption (balance, PU4) is explicit; the paper also honestly states the open problem of unbalanced urns. The theorems assume λ1=b directly, so the proof does not depend on the correctness of the cited Lemma 2.2. Overall this is a solid, useful contribution to the Pólya urn literature.","major_comments":[],"minor_comments":[{"comment":"The step concluding uniform integrability is too terse: Theorem 3.3 is stated for p≥2, and boundedness of L^r norms for all r would be needed to upgrade convergence in distribution to moment convergence of order p. The argument works by applying Theorem 3.3 at an exponent q>p, but this should be stated explicitly.","section":"Section 4, proof of Theorem 3.2"},{"comment":"Theorem A.1 invokes Lemma A.2 without verifying its hypothesis that |Y_i|^p is uniformly integrable. The verification is short: under (3.1), the conditional law of ΔX_n is a mixture of the finitely many fixed laws of ξ_j, each with finite p-th moment, and the conditional expectation term in Y_i is a.s. bounded by C^{1/p}; adding this one-line check would make the appendix self-contained.","section":"Appendix A, Theorem A.1"},{"comment":"The notation ΔX_n is overloaded: it denotes both the generic replacement vector in the paragraph after (2.1) and the actual increment X_{n+1}-X_n in (4.1). Using different symbols for the generic replacement and the realized increment would avoid confusion.","section":"Section 2.1 and (4.1)"},{"comment":"The passage from the case i≥i0 to all i≥1 is asserted in a single sentence. The finite number of omitted factors are indeed bounded, but spelling out this boundedness would make the proof fully explicit.","section":"Lemma 6.1"},{"comment":"There are small typos: 'irrdeucible' should be 'irreducible' in Section 1, and 'strictly less that p' in Remark 3.5 should be 'strictly less than p'.","section":"Section 1 and Remark 3.5"}],"recommendation":"minor_revision","confidential_remarks":"This is a clean, incremental contribution from an established expert. The self-references are appropriate and prior related results are clearly credited, and the open problem about unbalanced urns is honestly stated. The only reason for suggesting minor revision rather than immediate acceptance is that a few expositional details, especially the verification of the hypotheses in Appendix A and the uniform-integrability step in Theorem 3.2, should be made explicit for the reader."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid, careful paper. Janson proves Lp bounds for balanced Polya urns under minimal assumptions—no irreducibility, random replacements—and uses them to upgrade known distributional limits to moment convergence. The three regimes in Theorem 3.1 (n^{1/2}, log correction at critical, n^{Re lambda2/lambda1}) are exactly what you'd hope for, and the proof is genuinely simple: balance makes the total weight deterministic, the recurrence becomes a fixed matrix product plus a martingale difference, then Burkholder plus matrix estimates do the rest. I checked the chain (4.11)-(4.16), Lemma 5.1, Lemmas 6.1-6.2, and the uniform integrability appendix; it holds together.\n\nThe main improvement over the earlier work by the same author and Pouyanne is the lack of irreducibility and the allowance of random replacements. The paper also states Theorem 3.2's moment convergence upgrade directly. It honestly says much is not new; the novelty is in the generality and the simpler proof. That is a fair self-assessment.\n\nSoft spots: the balance condition (PU4) is load-bearing. Without it, the weight sequence is random and the whole deterministic matrix product argument collapses. Janson flags this as Problem 1.1, so it is not hidden, but it does mean the paper's scope is limited to balanced urns. The proof for p >= 2 is complete; the extension down to 1 <= p < 2 is waved at rather than shown, and Appendix A only partially fills that gap. The reliance on [13, Lemma A.1] for lambda1 = b is fine because the theorems assume lambda1 = b directly. Constants Cp are unspecified, which is normal for this type of bound. One tiny typo in the proof of Theorem 3.1 writes a norm inequality with the expectation norm omitted, but it is obvious and harmless.\n\nWho is this for? Urn theory specialists and anyone applying Polya urns who needs moment convergence. It is a real gap being filled, and the proofs are checkable. I would send it to a serious referee.","headline":"Janson proves clean Lp moment bounds for balanced Polya urns without irreducibility; the proof is checkable and the limitations are honestly stated.","tokens_in":13792,"tokens_out":1398,"would_cite":true,"duration_ms":13623,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-07T15:38:23.982660+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}