{"id":"44beb995-e4cc-405e-8602-63acad7171bd","arxiv_id":"2608.12270","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For the critical 2D stochastic heat flow, the minimal number of ε-balls carrying almost all mass is ε^{-2}(log 1/ε)^{-1/2+o(1)}, with the complementary sparsity bound holding simultaneously.","lead":"The authors prove that the critical two-dimensional stochastic heat flow, a random measure obtained as the scaling limit of directed polymers, is supported on about ε^{-2}(log 1/ε)^{-1/2} balls of radius ε at every scale. This sharp logarithmic fractal statement gives the first precise account of how the mass of this canonical random measure concentrates on sparse, high peaks.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the proof of Theorem 1.1 is internally coherent, and the residual gap is confined to the abstract's per-ball mass phrasing rather than the theorem's mathematical content.","rationale":"The reader's conditional verdict is reasonable, but the stated weakest assumption—uniformity of the imports (62) and (63) near the terminal index—is not where the paper is most exposed, because all bulk indices are separated from the terminal scale by a diverging power of log ε^{-1}. A careful check of the small-total-length argument in Proposition 5.1 and the Radon–Nikodym bound in Proposition 11.1 shows that the smallness condition (300) is satisfied for the relevant sets once one uses N_{ε,b} ≍ log ε^{-1}; similarly, the prefactor estimates in Proposition 10.3 are crude but adequate. The genuine discrepancy is the abstract's per-ball mass statement, which goes beyond Theorem 1.1; this warrants a wording qualification but does not affect the validity of the main theorem. I therefore find no load-bearing objection and recommend no verdict change.","tokens_in":68644,"tokens_out":47362,"duration_ms":399387,"concrete_test":"Independently verify the imported uniform moment bounds (62) and (63) from [35] on the bulk window [1, N − (log ε^{-1})^η], especially the higher-moment bound E|W_i|^p ≤ C_p (N−i)^{-p/2} for i within log-precision of the cutoff; if these fail uniformly as ε → 0, the large-deviation rate (52) and hence the covering exponent would be unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I find no load-bearing mathematical flaw in the central argument. The reader's identified weakest assumption, the imported uniform moment bounds (62) and (63), is in fact safe for the indices where it is used: the bulk cutoff M_{ε,η} = N_{ε,b} − (log ε^{-1})^η lies at distance ℓ_{ε,η} → ∞ from the terminal index N_{ε,b} − c*, so the moment bounds apply uniformly as ε → 0. The apparent difficulty in Proposition 11.1's smallness hypothesis (300) also resolves on correct bookkeeping: since N_{ε,b} ≍ log ε^{-1}, the condition requires |U|·Δ_U = o(1/log log ε^{-1}), which holds for the sets |U| = O(log log ε^{-1}) that arise in Proposition 5.1. The Berry–Esseen prefactor estimate in Proposition 10.3 is crude but sufficient: the stated error (log ε^{-1})^{-η/2} is indeed negligible relative to the Gaussian prefactor (log N_{ε,b})^{-1/2} ≍ (log log ε^{-1})^{-1/2}. The mass concentration, covering, and sandwiching arguments in Sections 8 and 9 are consistent, and the ε-dependent parameter transfer in Corollary 8.4 is supported by the stated monotonicity and variance estimates. The only substantive gap between the paper's claims and its formal results is the abstract's assertion that the covering balls each contain ε^2 (log ε^{-1})^{1/2+o(1)} mass: Theorem 1.1 guarantees each covering ball has mass at least ε^2 (log ε^{-1})^{1/2−o(1)}, and the lower bound (ii) prevents too few balls, but an individual ball mass upper bound is not proved. This is a presentation issue, not a threat to Theorem 1.1.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a sharp logarithmic intermittency statement for the critical 2D stochastic heat flow. In Theorem 1.1 it shows that, almost surely, for every small ε, the point-to-plane SHF in [0,1]^2 can be covered up to vanishing mass by at most ε^{-2}(log ε^{-1})^{-1/2+ξ} balls of radius ε, and conversely that any union of ε^{-2}(log ε^{-1})^{-1/2-ξ} such balls carries negligible mass. The proof proceeds by decomposing the time interval into geometric blocks, comparing the SHF with a decoupled product of single-block partition functions, proving a sharp upper-tail large-deviations estimate for the bulk product (Proposition 2.1), establishing a conditional decoupling bound (Proposition 2.2), and then deriving a mass-concentration estimate (Proposition 2.3) that is transferred from the sparse sequence r_m=e^{-e^m} to all scales by sandwiching. The heart of the argument is the large-deviations analysis of the decoupled product in Section 10 and the Radon-Nikodym stability estimate in Section 11.","tokens_in":69008,"tokens_out":11639,"duration_ms":112248,"significance":"If correct, this is a substantial quantitative advance: it gives the first sharp covering-number estimate for the support of the critical SHF, exposing the logarithmic correction beyond ordinary Minkowski dimension, and it introduces reusable tools (conditional-density estimates under large-deviation conditioning, approximate log-Laplace expansions, and a projected Radon-Nikodym stability estimate). The proof is internally structured and does not fit parameters: the log-normal and Gaussian heuristics are motivational only, and the main imports from [35], [45], and [46] are prior results with independent proofs. The central theorem is stated precisely and is falsifiable. The issues I identify do not appear to threaten the core derivation, but one uniformity gap in the proof of the lower covering bound needs to be repaired before the theorem is fully proved as stated.","major_comments":[{"comment":"The proof of part (ii) applies the Borel-Cantelli lemma to a fixed sequence of coverings C_n, but the theorem requires control of the supremum over all such coverings. For each deterministic choice of C_n the argument yields Z(C_n∩Q)→0 almost surely along that fixed sequence; this does not rule out rare adversarial coverings at each scale. The bound preceding (206) is uniform in C_n, so the gap is repairable: if a covering with Z(C_n∩Q)>δ exists, its r_n/10-neighborhood can be replaced by a union of O(m_n) rational-grid balls, and the resulting deterministic union of radius O(r_n) balls has the same small cardinality; the low-density part of the integral is then < c_2δ/2 for large n, forcing the global high-density integral in (195) to exceed c_2δ/2, an event whose probability is summable by (195). As written, however, the uniformity over all coverings is not justified, and this is load-bearing for Theorem 1.1(ii).","section":"Section 9, Step 3 (around Eq. (206))"},{"comment":"The abstract states that the SHF mass is concentrated on balls 'each containing ε^2 log^{1/2+o(1)}(1/ε) mass.' Theorem 1.1 and its proof only establish that the total mass outside a family of at most ε^{-2}(log ε^{-1})^{-1/2+ξ} balls is o(1); this gives a lower bound for the average mass per covering ball of order ε^2(log ε^{-1})^{1/2-o(1)}, but no upper bound on the mass of an individual covering ball is proved. The formal theorem also concerns Q=[0,1]^2, while the abstract says 'in any domain.' These two phrases should be reformulated so that the advertised claims match the theorem and its proof.","section":"Abstract and Remark 1.3"}],"minor_comments":[{"comment":"The extension to general ε is phrased as an argument along an arbitrary decreasing sequence ε_n; for the 'simultaneously for all ε' formulation it would be clearer to state that for each ε one takes m=m(ε) with r_{m+1}<ε≤r_m, so the same estimates give the simultaneous statement directly.","section":"Section 9, Step 4"},{"comment":"The term (log ε^{-1})^{-η/2+o(1)} is used as an additive error in a logarithmic ratio; the o(1) in the exponent should be made unambiguous, since the final uniformity in |s|≤A is important for the application in Section 11.","section":"Section 10.2, Eq. (282)"},{"comment":"The use of Lemma B.1 and Corollary B.3 to justify that Y(eI,eJ) is measurable with respect to H_{eK} is correct but terse; a sentence explaining why the Radon-Nikodym factor depends only on the bulk variables would improve readability.","section":"Section 7.4, Eq. (137)"},{"comment":"The notation bG^tail_{eI} is typographically heavy and easily confused with the tail product G_{η;2}; renaming it, for instance G^{tail,¬}_{eI}, would reduce the notational load in Sections 5 and 7.","section":"Notation around (73)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is technically dense and the core large-deviations/decoupling argument appears sound. My main concern is the missing uniformity step in the proof of Theorem 1.1(ii) in Section 9; it is local and repairable, but it is genuinely needed for the stated supremum over all coverings. The abstract also overstates what is proved about individual ball masses. The proof depends heavily on several recent preprints ([35], [45], [46], [23]); the editor may wish to confirm that these references are in stable form before acceptance. No issue of circularity or parameter-fitting arises."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper. First, it delivers the first sharp logarithmic intermittency result for the critical 2D SHF: the covering number is ε^{-2}(log 1/ε)^{-1/2+o(1)} and the mass is sparse in exactly that sense. Second, despite the length and the heavy machinery, the proof holds up. I read the reader's concern about the imported moment bounds (62)-(63), and the stress-test note is right: the bulk cutoff M_{ε,η} is far enough from the terminal index for those bounds to apply uniformly. The apparent gap in Proposition 11.1's smallness hypothesis also resolves on correct bookkeeping. The central argument is coherent.\n\nWhat is genuinely new: the two-sided covering and sparsity exponents, and the conditional large-deviations machinery in Sections 10 and 11. That is not an application of an existing tool; the paper develops a tilting argument for products of near-one factors, with sharp MGF expansions and a Berry-Esseen step under the tilted measure. The Brownian-conditioning intuition in the introduction is a useful guide, and the proof follows it honestly. The intermediate propositions (2.1, 2.2, 2.3) are stated precisely, the error terms are explicit, and the Borel-Cantelli steps in Section 9 are standard and sound.\n\nThe soft spots are real but not load-bearing. The abstract claims that each covering ball contains ε^2(log 1/ε)^{1/2+o(1)} mass. Theorem 1.1 does not prove an upper bound per ball; it proves the lower bound and the global sparsity. That is an overstatement in the abstract, not a flaw in the mathematics. The paper also imports several black-box estimates from [35], [45], [46] and [51]. Those are independent prior works, and the reliance is clearly flagged, but the uniform moment asymptotics carry a lot of weight. On reading, the index separation made me comfortable that the imports are safe where used. The ε-dependent parameter transfer is handled via remarks (7.10, 8.3, 11.2) rather than fully expanded proofs; acceptable for a first version, but the final version should make those remarks more explicit.\n\nWho this is for: probabilists working on the SHF, directed polymers, or intermittency in random measures. It is a serious technical paper, not a light read, but it answers two natural questions that were explicitly open. I would cite it if I worked in this area. It deserves a serious referee—send it to peer review. The referee should check the large-deviations estimates carefully and ask the authors to fix the abstract's per-ball claim, but the main theorem is believable and well-supported.","headline":"Sharp logarithmic intermittency for critical 2D SHF, proved with substantial new machinery; the abstract overstates one per-ball claim, but the main theorem is sound and deserves peer review.","tokens_in":69574,"tokens_out":2105,"would_cite":true,"duration_ms":19150,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H15","60F10","82B44","60J65"],"pacs":[],"model":"deepseek-v4-flash","headline":"The critical 2D stochastic heat flow is logarithmically intermittent.","keywords":["critical 2D stochastic heat flow","directed polymer","intermittency","large deviations","logarithmic fractal","covering number","mass concentration","random measures"],"falsifier":"Evaluate, analytically or numerically, the second moment of $G_i$ for indices $i$ within $(\\log\\varepsilon^{-1})^\\eta$ of the terminal scale: if $\\sup_{\\varepsilon,i}(N_{\\varepsilon,b}-i)\\,\\mathbb{E}(G_i-1)^2$ fails to converge to $1$ as $b\\to0$, or if the high-moment bound $\\mathbb{E}|G_i-1|^p\\le C_p(N_{\\varepsilon,b}-i)^{-p/2}$ fails there, then the upper-tail exponent and the covering-number exponent collapse. A purely numerical falsifier would be to simulate the critical lattice directed polymer at inverse temperature $\\beta_N$ satisfying (5) and count the $\\varepsilon$-balls needed to contain almost all mass: the count must be $\\varepsilon^{-2}(\\log\\varepsilon^{-1})^{-1/2+o(1)}$, and a different logarithmic exponent would disprove Theorem 1.1.","tokens_in":68426,"feed_emoji":"🌊","tokens_out":9655,"duration_ms":82713,"temperature":0.7,"pith_summary":"The paper establishes that the critical two-dimensional stochastic heat flow (SHF) — the random measure obtained as the scaling limit of $2+1$-dimensional directed polymers at their critical intermediate-disorder scaling — is logarithmically intermittent. Almost surely, for every sufficiently small $\\varepsilon>0$, all but a vanishing fraction of the flow's mass in $[0,1]^2$ is carried by at most $\\varepsilon^{-2}(\\log\\varepsilon^{-1})^{-1/2+\\xi}$ balls of radius $\\varepsilon$, each carrying roughly $\\varepsilon^2(\\log\\varepsilon^{-1})^{1/2}$ mass; conversely, any union of $\\varepsilon^{-2}(\\log\\varepsilon^{-1})^{-1/2-\\xi}$ radius-$\\varepsilon$ balls has vanishing mass. Hence both the minimal covering number and the maximum sparse mass are exactly $\\varepsilon^{-2}(\\log\\varepsilon^{-1})^{-1/2+o(1)}$. This refines the earlier almost-sure singularity of the SHF into a sharp logarithmic-fractal statement, and the proof supplies a large-deviation machinery for the block decomposition of the flow, including a conditioning estimate showing that conditioning on a large upper tail shifts means but barely changes the law projected onto a few coordinates.","feed_headline":"Critical 2D SHF mass is logarithmically sparse","feed_subtitle":"At radius ε, almost all SHF mass lies on ε^{-2}(log 1/ε)^{-1/2+o(1)} balls, which is the sharp covering number.","key_machinery":"The central object is the multiplicative block decomposition $G_i=p(t_{i-1})\\blacktriangleleft Z_i\\blacktriangleright 1$, with $t_i=\\varepsilon^2b^{-2i}$ and $N_{\\varepsilon,b}\\asymp\\log\\varepsilon^{-1}$ blocks; the $G_i$ are independent and in the quasi-critical bulk each is concentrated near $1$, with $\\mathbb{E}(G_i-1)^2\\sim(N_{\\varepsilon,b}-i)^{-1}$ and matching high-moment bounds. Because of this concentration, $\\log G_i$ is close to $W_i-W_i^2/2$, turning products into tilted Gaussian-like sums. The machinery consists of three interlocking estimates: sharp upper-tail large deviations for the bulk product (Proposition 2.1, via moment-generating-function expansions and a Berry–Esseen bound under exponential tilt), a conditional-decoupling estimate comparing the true partition function to the decoupled product (Proposition 2.2), and a Radon–Nikodym bound showing that conditioning on the upper tail shifts the mean of each coordinate by $\\alpha/(N_{\\varepsilon,b}-i)$ but leaves the joint law on any small coordinate set nearly unchanged (Propositions 7.7 and 11.1). The mass-concentration and covering-number conclusions follow by first-moment arguments on these level sets.","core_discovery":"On the paper's own terms, the discovery is the exact logarithmic sparsity of the critical two-dimensional SHF. Theorem 1.1 states that for every fixed $\\xi\\in(0,1/10)$, almost surely, for all small $\\varepsilon$, there is a collection $\\mathcal{B}_\\varepsilon$ of radius-$\\varepsilon$ balls with $|\\mathcal{B}_\\varepsilon|\\le \\varepsilon^{-2}(\\log\\varepsilon^{-1})^{-1/2+\\xi}$ such that the SHF mass of $[0,1]^2\\setminus\\bigcup_{B\\in\\mathcal{B}_\\varepsilon}B$ is $o_\\varepsilon(1)$; and conversely every union of at most $\\varepsilon^{-2}(\\log\\varepsilon^{-1})^{-1/2-\\xi}$ such balls has mass $o_\\varepsilon(1)$. Since the flow gives positive mass to every nonempty open set, the random measure is not supported on a finite set: it is carried by a dense but logarithmically sparse collection of tiny peaks, each of height $\\varepsilon^2(\\log\\varepsilon^{-1})^{1/2+o(1)}$ relative to Lebesgue measure.","pith_inferences":["A natural next step the paper leaves open is the exact Hausdorff measure with gauge $r^2(\\log r^{-1})^{1/2}$; the two-sided covering estimates here are exactly the scale at which such a gauge statement should be provable.","The conditioning mechanism — a Brownian large-endpoint event shifts means without changing increment laws — is generic for log-Normal-like cascades, so the same large-deviation-plus-Radon–Nikodym strategy may transfer to other critical multiplicative measures such as critical Gaussian multiplicative chaos.","The paper only treats the upper tail through the decoupled proxy; extending the sharp large-deviation estimate directly to the true partition function would yield extreme-value statistics for the SHF maxima, for instance a prediction for the law of the maximum over $\\varepsilon$-balls."],"forward_implications":["The minimal number of $\\varepsilon$-balls needed to capture all but a vanishing fraction of the point-to-plane SHF mass in a bounded domain is exactly $\\varepsilon^{-2}(\\log\\varepsilon^{-1})^{-1/2+o(1)}$, so the measure's support has Minkowski dimension $2$ but a precise logarithmic correction.","Any $\\varepsilon$-ball in the carrying collection carries mass $\\varepsilon^2(\\log\\varepsilon^{-1})^{1/2+o(1)}$, while a typical $\\varepsilon$-ball carries only $\\varepsilon^2(\\log\\varepsilon^{-1})^{-1/2}$; the mean is therefore dominated by rare peaks whose probability is $(\\log\\varepsilon^{-1})^{-1/2+o(1)}$.","The same comparison between the true flow and its decoupled product implies that the small-ball averaged SHF is, at the level of its first moment, indistinguishable from a log-Normal variable whose variance is $\\log\\log\\varepsilon^{-1}$, confirming the heuristic that the measure is almost a function.","Because the result holds simultaneously for all small $\\varepsilon$ almost surely, it gives a sharp two-sided statement, not just an upper envelope: no sparse set with slightly fewer balls can capture a non-vanishing fraction of the mass.","The proof's estimates are stable under $\\varepsilon$-dependent choices of the block ratio $b$ and coupling $\\vartheta$, so the logarithmic covering statement transfers from the auxiliary scale-$\\varepsilon$ object to the original spatial $[0,1]$ flow."],"supporting_citations":[{"why":"Constructs the critical 2D SHF as the scaling limit of the directed polymer partition function; this is the object whose logarithmic intermittency is characterized.","marker":"[14]"},{"why":"Supplies the log-log Gaussian fluctuations and, critically, the uniform moment asymptotics (62)-(63) for the single-scale factors $G_i$ that drive the block concentration and large-deviation estimates.","marker":"[35]"},{"why":"Established almost-sure singularity of the SHF and the quasi-critical Edwards-Wilkinson behavior used to seed the Gaussian approximation of the blocks.","marker":"[17]"},{"why":"Provides the shrinking-ball moment asymptotics used for terminal blocks and for the short-interval moment bounds in Lemmas 7.2 and 7.3.","marker":"[45]"},{"why":"Strict local positivity of the SHF ensures the block variables $G_i$ and the point-to-line mass are positive almost surely, so $\\log G_i$ is well defined throughout the large-deviation analysis.","marker":"[46]"},{"why":"Defines the convolution product of singular SHF kernels used in the expansion of the full partition function around the decoupled proxy.","marker":"[23]"}],"fun_headline_variants":["2D critical SHF mass is log-sparse, sharp","Sharp log-sparse covering for critical 2D SHF","Critical 2D SHF: mass on log-sparse balls","Logarithmic sparsity of 2D SHF made sharp"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire argument imports from [35] the uniform moment bounds that each quasi-critical block variable $G_i$ is concentrated near $1$ with second moment $\\asymp(N_{\\varepsilon,b}-i)^{-1}$ and all higher moments matching that decay, uniformly for every bulk index; if these bounds failed near the final block $M_{\\varepsilon,\\eta}$, the sharp large-deviation rate and therefore the covering exponent in Theorem 1.1 would no longer be established.","fun_headline_variants_meta":{"raw":{"variants":["2D critical SHF mass is log-sparse, sharp","Sharp log-sparse covering for critical 2D SHF","Critical 2D SHF: mass on log-sparse balls","Logarithmic sparsity of 2D SHF made sharp"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00026,"raw_usage":{"total_tokens":1694,"prompt_tokens":1157,"completion_tokens":537,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":773,"completion_tokens_details":{"reasoning_tokens":464}},"tokens_in":773,"tokens_out":537,"duration_ms":5633,"temperature":1.0,"reasoning_tokens":464,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:11:09.545206+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate, analytically or numerically, the second moment of $G_i$ for indices $i$ within $(\\log\\varepsilon^{-1})^\\eta$ of the terminal scale: if $\\sup_{\\varepsilon,i}(N_{\\varepsilon,b}-i)\\,\\mathbb{E}(G_i-1)^2$ fails to converge to $1$ as $b\\to0$, or if the high-moment bound $\\mathbb{E}|G_i-1|^p\\le C_p(N_{\\varepsilon,b}-i)^{-p/2}$ fails there, then the upper-tail exponent and the covering-number exponent collapse. A purely numerical falsifier would be to simulate the critical lattice directed polymer at inverse temperature $\\beta_N$ satisfying (5) and count the $\\varepsilon$-balls needed to contain almost all mass: the count must be $\\varepsilon^{-2}(\\log\\varepsilon^{-1})^{-1/2+o(1)}$, and a different logarithmic exponent would disprove Theorem 1.1.","supporting_citations":[{"cited_title":"On the moments of the mass of shrinking balls under the critical 2d stochastic heat flow.Communications in Mathematical Physics, 407(6):132, 2026","cited_arxiv_id":null,"evidence_quote":"Provides the shrinking-ball moment asymptotics used for terminal blocks and for the short-interval moment bounds in Lemmas 7.2 and 7.3."},{"cited_title":"Continuum polymer measures corresponding to the critical 2d stochastic heat flow.Communications in Mathematical Physics, 407(7):152, 2026","cited_arxiv_id":null,"evidence_quote":"Defines the convolution product of singular SHF kernels used in the expansion of the full partition function around the decoupled proxy."}],"review_version":1}