{"id":"e558df03-8f31-48dd-b94d-ff676be95e2b","arxiv_id":"1908.06555","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"The partition functions for directed polymers on diamond graphs with b=s converge in distribution to a unique limit law under a fine-tuned critical inverse-temperature scaling.","lead":"This paper proves a conjectured limit theorem for random polymer partition functions on diamond-shaped hierarchical graphs in the critical, marginally relevant case. The result gives a rigorous description of the weak-disorder limit and introduces a perturbative Stein's method that may transfer to (2+1)-dimensional polymers.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central proof depends on unproved external characterizations of R_b(r) and R_b^(m)(r); if the r→−∞ asymptotics from [10] fail, the Wasserstein contraction argument collapses.","rationale":"Read in good faith, the paper’s own chain from regular Q-pyramidic arrays through Proposition 9.1 and Lemmas 9.7–9.9 is detailed, and no circularity or internal inconsistency is apparent once the quoted moment functions are granted. The weak point is the foundation: Lemma 2.3 and Theorem 2.4 are imported from [10] without proof, and the paper explicitly says the proof of Proposition 6.14 is omitted. These results are not decorative; they supply the variance asymptotics, the higher-moment asymptotics, and the positivity/summability estimates that make the Wasserstein contraction work. If those quoted statements are correct, the central claim appears well-supported. If they are not, the proof of Theorem 6.23 fails at multiple points. The reader’s CONDITIONAL verdict already flags this dependence, so I see no basis to move the verdict. In particular, I would not reject the paper: the reliance is on prior published results, not on an obvious internal error. The appropriate action is to keep conditional acceptance pending verification of the quoted lemmas and the omitted proof of Proposition 6.14.","tokens_in":85781,"tokens_out":8088,"duration_ms":88394,"concrete_test":"Independently re-derive Lemma 2.3(II) from the recursion M(R(r)) = R(r+1) with M(x) = b^{-1}[(1+x)^b − 1], and verify the coefficient η_b = (b+1)/(3(b−1)) together with the stated O(log^2(−r)/|r|^3) remainder. Then check that the constant δ in Proposition 9.1, built from the series ∑(R(s−ℓ) − κ^2/(ℓ−s)) and ∑ R(s−ℓ)^2, is positive and finite using that expansion. If the logarithmic coefficient differs, recompute Lemma 9.7’s bound (11.4) and the rates in Lemmas 9.8–9.9; a mismatch would break the proof of Theorem 6.23. Separately, supply the omitted proof of Proposition 6.14 by adapting the proof of Theorem 3.3 of [10], so that moment convergence for minimally regular Q-pyramidic arrays is fully justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper’s main convergence theorem, Theorem 6.23 / Theorem 2.7, is a Wasserstein-2 argument whose constants and error terms are controlled by the variance function R_b(r) and the higher-moment functions R_b^(m)(r). These are imported from [10] as Lemma 2.3 and Theorem 2.4: the r→−∞ expansion R_b(r) = −κ_b^2/r + κ_b^2 η_b log(−r)/r^2 + O(log^2(−r)/|r|^3), the derivative formula, and the moment asymptotics are quoted without proof. This is load-bearing in at least four places: (1) Proposition 9.1’s proof needs R(s−k) positive with summable differences to define δ > 0; (2) Lemma 9.7’s telescoping estimate (11.4) uses the same expansion; (3) Lemma 11.3’s variance and fourth-moment bounds invoke Lemma 2.3(II) and Proposition 12.1, whose proof relies on Lemma 15.7 from [10]; and (4) Corollary 9.12 and the uniqueness argument use R(r−N) ≈ κ^2/N. In addition, Proposition 6.14, which transfers moment convergence to arbitrary minimally regular arrays, is stated without proof. These are not internal contradictions, but they mean the central claim’s rigor depends on external lemmas that are not reproduced or independently verified in this paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proves the conjectured weak-disorder distributional limit for directed polymer partition functions on diamond hierarchical graphs at marginal criticality (b=s). The main result, Theorem 2.7 (refined as Theorem 6.23), establishes convergence in law of W_n^ω(β_{n,r}) to a limit law L_r uniquely determined by its variance R_b(r), centered moments R_b^(m)(r), the recursive identity W_{r+1} = (1/b)∑_{i=1}^b ∏_{j=1}^b W_r^{(i,j)}, and Gaussian tails as r→−∞. The proof represents W as Q^n applied to an array of i.i.d. random variables, decomposes Q = L + E, and controls Wasserstein-2 distance through a contractive estimate (Proposition 9.1) and Stein-type Gaussian approximation lemmas (Lemmas 9.7–9.9). A parallel theorem for vertex disorder (Theorem 3.1) is proved by reducing the site-disorder model to a bond-disorder model at logarithmic scale. Several variance and moment characterizations are imported from the prior paper [10], and one rate lemma for the sharp-regularity theorem is left unproved. The paper also proves an explicit scaling form (2.5), including third- and fourth-cumulant corrections, and derives a rate version (Theorem 7.3) under stronger moment assumptions.","tokens_in":86096,"tokens_out":5718,"duration_ms":64425,"significance":"If the result is correct, it confirms a conjecture in [10] and provides one of the few rigorous critical weak-disorder scaling limits for a marginally relevant disordered system. The proof is a substantial technical contribution: the Q-pyramidic array formalism, the contractive L2 bound in Proposition 9.1, and the perturbative Stein-method estimates in Section 11 are genuine new machinery, and the reduction of the site-disorder model in Section 14 is an elegant and useful device. The main theorem makes a falsifiable prediction: the limit law depends only on r and the low-order cumulants of the disorder, not on the full distribution. The principal caveat is that the proof leans heavily on the variance and moment asymptotics of [10], which are quoted without proof; the rate theorem 7.3 additionally depends on an unproved Lemma 13.5. These omissions are the main obstacle between the manuscript and a fully self-contained proof.","major_comments":[{"comment":"The entire r→−∞ asymptotic regime is imported from [10] without proof: Lemma 2.3 gives R_b(r) = −κ_b²/r + κ_b²η_b log(−r)/r² + O(log²(−r)/|r|³) and the derivative formula, and Theorem 2.4 gives the higher-moment asymptotics R_b^(m)(r) ∼ κ_b^m m!/(2^{m/2}(m/2)! |r|^{m/2}) for even m. These estimates are load-bearing in at least four places: the definition of δ in Proposition 9.1 uses summability of R(s−ℓ)−κ²/(ℓ−s); the telescoping estimate (11.4) in Lemma 9.7 uses the same expansion; Lemma 11.3’s variance and fourth-moment bounds invoke Proposition 12.1 and Lemma 15.7, both from [10]; and Corollary 9.12 and Example 7.6 use R(r−N) ≈ κ²/N. If any of these asymptotics were false, the proof of Theorem 6.23 would collapse at a central step. The manuscript should either reproduce the proofs (at least of the precise estimates used) or state Lemma 2.3 and Theorem 2.4 as explicit standing hypotheses, cleanly separated from the new results proved here.","section":"§2.4, Lemma 2.3 and Theorem 2.4; used throughout §§9–11"},{"comment":"Proposition 6.14 is stated without proof: the text says only that the proof is the same as part (i) of Theorem 3.3 of [10], or that the proof of that theorem implicitly proves Proposition 6.14. This proposition is not a peripheral remark: it is used in Lemma 6.15(III) to obtain the moment convergence E[(X_a^{(k,n)})^m] → R^(m)(r−k), which in turn is used in the tightness construction in Section 8 and in the Wasserstein estimates of Section 11. Since the proposition generalizes Theorem 2.4 from the specific disorder variables (6.1) to arbitrary minimally regular arrays, it is exactly the transfer principle needed by Theorem 6.23. Please provide a proof, or state it as a quoted theorem with a precise reference and the exact moment conditions required.","section":"§6.2, Proposition 6.14"},{"comment":"Theorem 7.3 is a main result giving an explicit Wasserstein-2 rate n^{−υ}, and Corollary 7.5 plus Example 7.6 depend on it. The proof of Theorem 7.3 is chained through Lemma 13.5, whose proof is omitted with the explanation that it is a 'lengthy near-repetition' of the arguments in Section 11. As written, the rate claim is therefore unsubstantiated: the omitted lemma is not a routine variation but a full analogue of Lemmas 9.7–9.9 under α-sharp regularity, including the control of the error terms ξ_N(n), ξ'_N(n), and ξ''_N(n). Either the proof of Lemma 13.5 should be included with the same level of detail as the proofs in Section 11, or Theorem 7.3 should be stated as conditional on Lemma 13.5, with the omitted proof explicitly marked as an assumption.","section":"§13.1–13.2, Lemma 13.5 and Theorem 7.3"}],"minor_comments":[{"comment":"There are minor typos: 'explicitely' in Section 4 should be 'explicitly', and 'Wassertstein' appears in the proof of Theorem 6.23 in Section 9.3.","section":"§4 and §9.3"},{"comment":"The final proof in Appendix C is headed 'Proof of Lemma 11.7', but the statement being proved is Corollary 11.7; the cross-reference should be corrected.","section":"Appendix C"},{"comment":"The convention that 'for large enough N and n' means N > λ and n > Λ(N) is introduced in Remark 9.13, but the phrase is used earlier in Section 9.2 and in the statements of Lemmas 9.7–9.9. Moving this convention before its first use would improve readability.","section":"§9.2, Remark 9.13"}],"recommendation":"major_revision","confidential_remarks":"The paper's main proof is impressive but not self-contained: the central Wasserstein argument is conditioned on external asymptotics from [10], and the rate theorem depends on an omitted lemma. If the prior results in [10] are published and exact, the omissions may be acceptable after the authors clearly label imported results and verify each cited lemma; if not, the paper would need to include those proofs. The same-author dependence should be made explicit in the introduction, since readers will need to check whether the quoted lemmas already appear in published form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this paper proves the distributional limit theorem conjectured in [10] for directed polymers on diamond hierarchical graphs in the marginal b=s case. The limit law is characterized by its mean, variance, centered moments, and a recursive distributional identity, and the same limit is obtained for the site-disorder model. The proof goes through a Wasserstein-2 contraction argument driven by a perturbative Stein's method. That is a genuine technical step, because the usual Wiener chaos expansion does not exist in this marginal regime.\n\nThe paper does several things well. The array framework (Q-pyramidic arrays) is clean and leads to a mild universality statement: any minimally regular array with the right variance and vanishing fourth moment converges to the same limit. The proof of the main theorem is detailed, with the core bounds in Lemmas 9.7–9.9. The site-disorder reduction in Section 14 is a nice piece of work, reducing vertex disorder to edge disorder after integrating out low generations.\n\nThe soft spots are real but localized. The variance function R_b(r) and higher moment functions R_b^(m)(r) are imported from [10] as Lemma 2.3 and Theorem 2.4 without proof. These are load-bearing: the r→−∞ asymptotics enter the Wasserstein contraction in Proposition 9.1 and the approximation lemmas. If those asymptotics were false, the argument would collapse. This reliance is legitimate when the prior paper is accepted, but it makes the current paper not self-contained in a central way. Proposition 6.14 is also asserted without proof (deferred to [10]), and Lemma 13.5 is described as a lengthy near-repetition and not proved. These omissions are not internal contradictions, and I do not think they undermine the main theorem, but they should be supplied or precisely referenced before publication.\n\nWho this is for: people working on weak-disorder scaling limits, hierarchical polymer models, and marginally relevant disorder. The paper deserves a serious referee who can check the Wasserstein estimates against [10]. My recommendation: send it to peer review, with the requirement that the author either proves the imported lemmas or gives exact pointers to where each one is proved, especially Proposition 6.14 and Lemma 13.5.","headline":"Proves the conjectured critical weak-disorder limit for diamond-graph polymers via a perturbative Stein method; the proof leans on unproved lemmas from the author's earlier paper, which is the main thing to check.","tokens_in":86583,"tokens_out":1883,"would_cite":true,"duration_ms":23763,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","82B44","60F05"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves the conjectured distributional limit theorem for partition functions of directed polymers on diamond hierarchical graphs in the $b=s$ critical case: the limit laws are unique, universal in the disorder up to low-order…","keywords":["directed polymers in random environment","diamond hierarchical graphs","marginal relevance","critical weak-disorder limit","partition function","distributional limit theorem","Wasserstein distance","Stein's method"],"falsifier":"Simulate the partition-function recursion (2.2) for $b=s=2$ and $b=s=3$ with $\\beta_{n,r}$ given by (2.5) for a fixed $r$, using two different disorder distributions with matching third and fourth cumulants; if the empirical fourth centered moments fail to approach $R_b^{(4)}(r)$ at the rate allowed by Theorem 7.3, or if the two distributions yield different limiting laws, the central claim is false.","tokens_in":85576,"feed_emoji":"🎲","tokens_out":13881,"duration_ms":137076,"temperature":0.7,"pith_summary":"This paper proves a conjecture from an earlier study of directed polymers on diamond hierarchical graphs: in the marginally relevant case with branching parameter $b$ equal to segmenting parameter $s$, the partition functions $W_n^\\omega(\\beta_{n,r})$ converge in distribution as $n\\to\\infty$ to a unique family of limit laws $\\mathcal{L}_r$, provided the inverse temperature is scaled as $\\kappa_b/\\sqrt{n}$ with specific third- and fourth-cumulant corrections. The limit is universal across disorder distributions sharing the same low-order cumulants and is uniquely characterized by mean one, variance $R_b(r)$, centered moments $R_b^{(m)}(r)$, and the recursive identity $W_{r+1}=b^{-1}\\sum_{i=1}^b\\prod_{j=1}^b W_r^{(i,j)}$. It matters because the $b=s$ model is the hierarchical analogue of the marginally relevant $(2+1)$-dimensional polymer, for which no Wiener-chaos expansion exists and uniqueness of the critical limit is still open. The proof works instead through Wasserstein distance and a perturbative generalization of Stein's method, and it also yields the same limit law for the site-disorder model.","feed_headline":"Critical disorder yields a unique limit law for polymer partition functions","feed_subtitle":"Diamond-graph polymers at the critical scaling converge in distribution, confirming a conjectured moment limit.","key_machinery":"The central object is the renormalization map $Q$ on edge-labeled arrays: with $\\{X_h\\}$ the centered disorder weights $e^{\\beta\\omega_h}/\\mathbb E[e^{\\beta\\omega_h}]-1$, the partition function is $W_n^\\omega(\\beta)=1+Q^n\\{X_h\\}$. The linearization $L$ preserves variance and $E=Q-L$ produces uncorrelated error terms, so repeated application of $Q$ builds a pyramid of arrays whose layer moments converge to the functions $R^{(m)}(r-k)$. The variance recursion is driven by $M_{b,b}(x)=b^{-1}[(1+x)^b-1]$, whose fixed point at 0 is marginally repelling precisely when $b=s$; the fine-tuned scaling is chosen so that the initial variance matches the required asymptotics and converges to the function $R_b(r)$ with the shift property $M_{b,b}(R_b(r))=R_b(r+1)$. Convergence in law is obtained by comparing $Q^N$ applied to approximating arrays through a contractive $L^2$ bound, with Gaussian approximations at two intermediate generational scales controlled by a two-variable perturbative Stein equation and the zero-bias transformation.","core_discovery":"The central claim is Theorem 2.7: for each $b\\in\\{2,3,\\ldots\\}$ and each $r\\in\\mathbb R$, the edge-disorder partition functions $W_n^\\omega(\\beta_{n,r})$ converge in distribution to a limit law $\\mathcal{L}_r$ uniquely determined by mean 1, variance $R_b(r)$, centered moments $R_b^{(m)}(r)$, the Gaussian behavior $\\sqrt{-r}(W_r-1)\\Rightarrow N(0,\\kappa_b^2)$ as $r\\to-\\infty$, and the recursion $W_{r+1}=b^{-1}\\sum_{i=1}^b\\prod_{j=1}^b W_r^{(i,j)}$. The same limit family arises for the vertex-disorder model under the inverse-temperature scaling $\\hat\\kappa_b/n$. As $r$ runs from $-\\infty$ to $+\\infty$, the family passes from Gaussian fluctuations around 1 to concentration near 0, so the critical scaling captures the transition from weak to strong disorder within the limit.","pith_inferences":["Likely transferable: the Wasserstein-plus-Stein template could be adapted to other marginally relevant hierarchical disordered systems and might provide a route to proving uniqueness of the critical distributional limit for the $(2+1)$-dimensional rectangular-lattice polymer, where only subsequential limits are currently known.","The infinite $Q$-pyramidic array of Theorem 6.16 can be read as a renormalization fixed point indexed by $r$; a natural next step would be to use it to construct a continuum random measure on the diamond fractal and study the log-Hausdorff dimension of typical path intersections.","A direct numerical test of the rate bound is feasible: simulate the recursion (2.2) for $b=2,3$ and check whether the Wasserstein-2 error to the limit law decays like $n^{-\\upsilon}$ for $\\upsilon<\\alpha/9$, as predicted by Theorem 7.3.","Sharpening the quoted asymptotics of $R_b(r)$ would likely yield explicit finite-$n$ corrections to the limit laws and a more quantitative connection to lognormal-type fluctuations at criticality."],"forward_implications":["The $b=s$ critical limit exists as a genuine distributional limit, not merely a moment limit; the super-factorial growth of $R_b^{(m)}(r)$ is consistent with a unique law.","Universality in the disorder: any centered variance-one disorder with finite exponential moments and fixed third and fourth cumulants produces the same family $\\mathcal{L}_r$ under the same scaling, with the cumulant corrections in the inverse-temperature expansion exactly accounting for the dependence on those moments.","The vertex-disorder model is in the same universality class: its partition functions converge, under $\\hat\\beta_{n,r}\\sim\\hat\\kappa_b/n$, to the same limiting variables $W_r$.","The critical scaling resolves the weak-to-strong disorder transition: $\\sqrt{-r}(W_r-1)$ converges to a centered normal with variance $\\kappa_b^2$ as $r\\to-\\infty$, while $W_r$ concentrates near zero as $r\\to\\infty$.","The proof supplies a template for critical weak-disorder limits without a Wiener-chaos expansion: hierarchical symmetry plus Wasserstein-2 contraction plus perturbative Stein's method."],"supporting_citations":[{"why":"Supplies the variance function $R_b(r)$, the limiting centered moments $R_b^{(m)}(r)$, and the conjecture that Theorem 2.7 completes; the paper quotes Lemma 2.3 and Theorem 2.4 from it without proof.","marker":"[10]"},{"why":"Provides the $b