{"id":"e76e1bfe-ca6f-4f85-9b35-4fc10445c7d7","arxiv_id":"2501.12172","paper_version":2,"verdict":"REJECT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"A quadratic backward stochastic differential equation is used to define and prove weak convergence of an ultraviolet-regulated 2D sine-Gordon measure that is absolutely continuous with respect to the Gaussian free field for beta-squared below 2.","lead":"This paper constructs a 2D sine-Gordon quantum field measure on bounded domains using backward stochastic differential equations, for inverse temperatures beta-squared below 2 and arbitrary coupling. It also connects the construction to XOR-Ising spin correlations and two-dimensional log-gas charge distributions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (3.4) rests on the false diagonal identity ∫f(x)^k dx = ∫f(x,y)^k 1_{x=y} dxdy; the integral actually diverges like ε^{-β²/2}, so the uniform L∞/BMO bounds behind Theorem 3.8 are unsupported.","rationale":"The reader's weakest_assumption identifies the same load-bearing error: the diagonal identity in (3.3)–(3.4) is false, and the uniform L∞ bound on the Wick-ordered cosine terminals is not available. I independently traced the use of this bound through Lemma 3.1, Proposition 3.7, and Theorem 3.8: it is exactly what produces the BMO control on the BSDE integrands and the uniform integrability of the stochastic exponentials Γ_ε. This is an internal inconsistency in the proof, not a disagreement with the known subcritical sine-Gordon construction; a repair may be possible via L^p or L² estimates for the terminal conditions of imaginary multiplicative chaos, but that repair is not present in the manuscript. Therefore the reader's REJECT verdict is appropriate, and no verdict adjustment is needed.","tokens_in":71618,"tokens_out":5239,"duration_ms":57536,"concrete_test":"Take Λ to be the unit disk and a fixed C_c^∞ mollifier. Compute I(ε) = ∫_{Λ_ε} exp((β²/2)A^ε_Λ(x)) dx for β² = 1. If I(ε)·ε^{1/2} tends to |Λ|·const as ε→0, the uniform bound (3.4) is refuted. Then check whether Lemma 3.1's BMO estimate (3.6) can be re-derived using only the correct L² moment bound E|([[cos]],φ)|² < ∞, which follows from ∫∫ e^{β²A^ε(x,y)} dxdy ≈ ∫∫ |x−y|^{-β²} dxdy < ∞. If the L²-only argument fails, the proof of Proposition 3.7 and Theorem 3.8 is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The critical failure is in the estimate (3.3)–(3.4). The authors assert that ∫_{Λ_ε} [A^ε_Λ(x)]^k dx equals ∫_{Λ_ε×Λ_ε} [A^ε_Λ(x,y)]^k 1_{x=y}(x,y) dxdy, then bound this by ∫_{Λ×Λ} [A_Λ(x,y)]^k dxdy. The displayed equality is measure-theoretically false: {x=y} has Lebesgue measure zero, so the middle integral is zero, not the left side. The intended conclusion also fails. For x away from ∂Λ, A^ε_Λ(x) = (ρ_ε ∗ ρ_ε ∗ A_Λ)(x,x) = log(1/ε) + O(1), so ∫_{Λ_ε} exp((β²/2)A^ε_Λ(x)) dx ∼ |Λ| e^{O(1)} ε^{-β²/2}, which diverges for every β>0. Thus the claimed uniform L∞ bound on the terminal variables ([[cos(βφ^ε_1)]], φ) in (3.3) is false. That bound is used in Lemma 3.1 to obtain the BMO estimate (3.6); the same sup_ε ||ξ_ε||_{L∞} < ∞ is invoked in Proposition 3.7 for the Vitali convergence argument and in Theorem 3.8 for the reverse Hölder bounds on Γ_ε and for E[Γ]=1. Without it, the paper does not prove convergence of the quadratic BSDEs to (3.22), nor the absolute continuity of the limiting sine-Gordon measure. The later applications in Section 4 inherit the same unsupported uniform estimates.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a new construction of the two-dimensional sine-Gordon measure on bounded domains, for inverse temperature β² ∈ [0,2), using quadratic backward stochastic differential equations driven by a cylindrical Wiener process. The terminal condition of the approximating BSDE is the Wick-ordered cosine of the mollified Gaussian free field integrated against a test function, and the main result, Theorem 3.8, asserts weak convergence of the approximate sine-Gordon measures to a measure μ_SG^χ defined by μ_SG^χ(A) = E[Γ(χ) 1_A], where Γ(χ) is the stochastic exponential of α times the martingale part of a limiting quadratic BSDE. The paper also draws applications to partition functions, the critical planar XOR-Ising model, and the sine-Gordon representation of two-dimensional log-gases. The central analytical claim is a uniform L∞ bound, in the ultraviolet cutoff, on the Wick-ordered cosine interaction; the convergence proof rests on this bound.","tokens_in":71955,"tokens_out":3758,"duration_ms":40736,"significance":"If the main theorem were valid, the paper would offer a genuinely new BSDE-based perspective on the sine-Gordon model and a conceptually clean proof of absolute continuity with respect to the Gaussian free field in the subcritical regime, together with nontrivial connections to imaginary multiplicative chaos and to the XOR-Ising model. The paper is clearly organized and engages seriously with the BSDE and imaginary-chaos literature. However, the central uniform bound is false as stated, and the main convergence result is therefore not established. Since the applications in Section 4 inherit the same unsupported estimates, the current manuscript does not deliver a proof of its advertised claims.","major_comments":[{"comment":"The estimate (3.4) is false. The displayed equality ∫_{Λ_ε} [A^ε_Λ(x)]^k dx = ∫_{Λ_ε×Λ_ε} [A^ε_Λ(x,y)]^k 1_{x=y}(x,y) dxdy is measure-theoretically incorrect: the set {x=y} has Lebesgue measure zero, so the right-hand side is zero for k≥1 rather than equal to the left-hand side. The subsequent bound by ∫_{Λ×Λ} [A_Λ(x,y)]^k dxdy therefore does not control the left-hand side. In fact, for x away from the boundary of Λ, A^ε_Λ(x) = log(1/ε) + O(1), so ∫_{Λ_ε} exp((β²/2)A^ε_Λ(x)) dx ∼ |Λ| e^{O(1)} ε^{-β²/2}, which diverges for every β>0. Consequently the asserted uniform L∞ bound on ([[cos(βφ_1^ε)]], χ) in (3.3) is not justified and is false as stated.","section":"Section 3.1, Eq. (3.4)"},{"comment":"The BMO estimate (3.6) in Lemma 3.1 depends directly on the uniform boundedness of the terminal variables, which in turn relies on (3.3)–(3.4). Since that estimate fails, the conclusion that Z^{ε,λF}·A is a BMO martingale with the stated bound is unsupported. The BMO property is also used in the uniqueness argument via the Girsanov-type Lemma A.4 of [32], so the well-posedness statement for the approximating BSDE is not established as proved.","section":"Lemma 3.1"},{"comment":"The convergence of the approximating BSDEs to the limiting quadratic BSDE (3.22) is not proved, because the proof repeatedly uses the false uniform bound sup_ε ||ξ_ε||_{L∞} < ∞. In Proposition 3.7, this bound is invoked to justify uniform integrability and dominated convergence for the difference of terminal conditions; in Theorem 3.8 it is used for the BMO/reverse-Hölder constants in (3.14), (3.24), and (3.29), and for the conclusion E[Γ(χ)]=1. Without a valid alternative control on the Wick-ordered interaction, the claimed weak convergence of μ_{SG}^{χ,ε} and the absolute continuity of μ_SG^χ with respect to the GFF law do not follow.","section":"Proposition 3.7 and Theorem 3.8"},{"comment":"The applications inherit the unsupported uniform L∞ bound. In Proposition 4.1 the identity Q_{χ,μ}^{(ε)} = Ξ_{χ,ε} and the interchange of limit and series use sup_ε ||([[cos(βφ_1^ε)]], χ)||_{L∞}, which fails by the same divergence. Theorem 4.6 invokes (3.28) and (3.4) to prove convergence of Fourier transforms and the continuity of Ψ_χ. Thus the results on the XOR-Ising model and log-gases are conditional on an invalid estimate and are not established.","section":"Section 4, Eqs. (4.5) and (4.9)"}],"minor_comments":[{"comment":"Reference [39] lists the year as 2001 for Probab. Theory Relat. Field 185; this appears to be a typo and should be corrected, since the cited work is clearly much more recent.","section":"Bibliography"},{"comment":"The MSC2020 classification line reads \"60H10; 81S20: 81T08\"; the punctuation is inconsistent and should be uniform, e.g. \"60H10; 81S20; 81T08\".","section":"Title page"},{"comment":"The notation ⟨cos(βφ_1), χ⟩_μ defines the limit through (3.18) using convergence in probability from [38], but the paper does not explicitly state the integrability of this limit needed for later expectations; this should be clarified.","section":"Section 3.2, after Eq. (3.18)"}],"recommendation":"reject","confidential_remarks":"The central problem is a single false measure-theoretic identity in Eq. (3.4), but it is genuinely load-bearing: the claimed uniform L∞ bound is not merely unproved, it is false because the mollified diagonal Green function diverges logarithmically. The main theorem, the convergence of the BSDEs, and all applications in Section 4 depend on this bound. A local correction does not appear possible without a substantially different argument, so rejection is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the quadratic BSDE framework is genuinely new and the paper is clearly written around a real idea, but the central convergence theorem rests on a false estimate. I agree with the stress-test note. Equation (3.4) uses the diagonal identity ∫ f(x)^k dx = ∫ f(x,y)^k 1_{x=y} dxdy, which is measure-theoretically wrong, and the intended uniform bound also fails because A^ε_Λ(x) diverges like log(1/ε), so ∫ exp(β²/2 A^ε) dx ~ ε^{-β²/2}. This kills the uniform L∞ bound on the Wick-ordered terminal condition, and that bound is used in Lemma 3.1, Proposition 3.7, and Theorem 3.8 for BMO control, Vitali convergence, and E[Γ]=1. Section 4 inherits the problem.\n\nWhat is actually new: applying quadratic BSDEs with a purely quadratic generator to normalize the sine-Gordon measure, and connecting the BSDE terminal data to imaginary multiplicative chaos. The Section 4 applications to partition functions, XOR-Ising correlations, and log-gas charge distributions are interesting if the foundations hold. The literature is engaged honestly, and the self-citation to [32] for a BMO lemma is not a concern; that is a parameter-free technical result.\n\nThe flaw is not a minor gap. The terminal condition is not merely unbounded in a way that a better inequality might fix; it is deterministically divergent in ε. So one cannot patch the proof by switching to a different norm in the same argument. The authors explicitly rely on uniform boundedness, and without it the quadratic BSDE well-posedness arguments do not go through. The later applications also use (3.4) directly, for example in Theorem 4.6, so they are unsupported too.\n\nWho this is for: readers interested in BSDE methods for constructive quantum field theory may find the setup worth knowing about, but the paper should not be relied upon as a construction until the convergence proof is corrected. I would not send it to a serious referee as it stands. If the authors can replace the false uniform bound with a correct estimate, or reformulate the terminal condition so that the BSDE well-posedness actually holds, the framework could become worth revisiting.","headline":"A genuinely new BSDE framework for the sine-Gordon measure, but the central convergence proof rests on a false diagonal identity, so the main theorem is unsupported.","tokens_in":72504,"tokens_out":3301,"would_cite":false,"duration_ms":38756,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H10","81S20","81T08"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper constructs the two-dimensional sine-Gordon measure on bounded simply connected domains for inverse temperatures $\\beta^2\\in[0,2)$ by showing that a family of cutoff-indexed quadratic backward stochastic differential equations…","keywords":["quadratic BSDE","sine-Gordon model","Gaussian free field","imaginary multiplicative chaos","ultraviolet renormalization","BMO martingales","XOR-Ising model","log-gases"],"falsifier":"Evaluate the left side of estimate (3.4) on the unit square for a mollified two-dimensional Dirichlet Green function: the asymptotic $A^\\varepsilon_\\Lambda(x)\\sim \\log(1/\\varepsilon)$ would make the integral diverge like a positive power of $1/\\varepsilon$ for $\\beta^2>0$, contradicting the uniform bound that supports Lemma 3.1 and Proposition 3.7.","tokens_in":71376,"feed_emoji":"🧮","tokens_out":7461,"duration_ms":77058,"temperature":0.7,"pith_summary":"The paper aims to give a new construction of the two-dimensional sine-Gordon model on bounded domains that avoids renormalization group flows and regularity structures. It introduces a family of quadratic backward stochastic differential equations indexed by an ultraviolet cutoff, with terminal conditions given by Wick-ordered cosine interactions. The main theorem proves that, as the cutoff vanishes for $\\beta^2\\in[0,2)$, these BSDEs converge to a single quadratic BSDE, and the stochastic exponential of its martingale part defines the density of the sine-Gordon measure with respect to the Gaussian free field. If correct, this yields a direct probabilistic normalization of the sine-Gordon measure, plus a route to the scaling limit of critical planar XOR-Ising correlation functions and to the weak convergence of normalized two-dimensional log-gas charge distributions.","feed_headline":"A quadratic BSDE constructs the 2D sine-Gordon measure","feed_subtitle":"The limiting measure stays absolutely continuous with respect to the Gaussian free field, with density a stochastic exponential.","key_machinery":"The central object is the quadratic BSDE $dY_t=-\\frac{\\alpha}{2}\\|Z_t\\|^2_{H^{-1}_0(\\Lambda)}\\,dt+Z_t\\,dA_t$ with terminal value given by the Wick-ordered cosine interaction tested against a test function, driven by a cylindrical Wiener process $A$ on $L^2(\\Lambda)$. The generator is purely quadratic in the control variable $Z$, which makes the equation amenable to BMO-martingale methods: the proof controls $\\|Z\\cdot A\\|_{BMO}$ uniformly in the cutoff, applies reverse H\\\"older inequalities to the stochastic exponential, and uses a linearized variational BSDE to identify the derivative of the log-Laplace transform. The limiting equation's solution $D(\\chi)$ feeds the stochastic exponential $\\Gamma(\\chi)$ that defines the sine-Gordon density.","core_discovery":"The paper claims that for any regular bounded simply connected domain $\\Lambda\\subset\\mathbb C$ and any $\\beta^2\\in[0,2)$, the approximate sine-Gordon measures with ultraviolet cutoff converge weakly to a measure $\\mu^{\\chi}_{SG}$ whose action on a Borel set $A$ is $\\mu^{\\chi}_{SG}(A)=\\mathbb E[\\Gamma(\\chi)\\mathbf 1_A]$. Here $\\Gamma(\\chi)$ is the stochastic exponential $\\exp\\big(\\alpha\\int_0^1 D_t(\\chi)\\,dA_t-\\frac{\\alpha^2}{2}\\int_0^1\\|D_t(\\chi)\\|^2_{H^{-1}_0(\\Lambda)}\\,dt\\big)$, where $(Y_t(\\chi),D_t(\\chi))$ solves the limiting quadratic BSDE with terminal condition $\\langle\\cos(\\beta A_1),\\chi\\rangle$, the real part of imaginary multiplicative chaos tested against $\\chi$. The proof establishes $\\mathbb E[\\Gamma(\\chi)]=1$, so the limiting sine-Gordon measure is absolutely continuous with respect to the law of the Gaussian free field, and the same BSDE machinery also gives a partition-function representation and the weak convergence of normalized charge distributions.","pith_inferences":["Editorial extension: the density $\\Gamma(\\chi)$ has the form of a Girsanov density for a drift change on the Gaussian free field path space, so if the convergence proof is completed the construction would imply mutual absolute continuity between the sine-Gordon measure and the Gaussian free field on the filtration generated by the cylindrical Wiener process.","Editorial extension: the same scheme could be tested on other Wick-renormalizable interactions whose mollified terminal conditions converge in probability, since the quadratic generator would remain unchanged while only the terminal condition is replaced.","Editorial extension: the test-function dependence of $\\Gamma(\\chi)$ indicates a family of normalizations rather than a single intrinsic measure; comparing the densities for different $\\chi$ would clarify whether the limiting object has a canonical localization-independent meaning."],"forward_implications":["For $\\beta^2\\in[0,2)$ and arbitrary coupling $\\alpha$, the sine-Gordon measure exists on any regular bounded simply connected domain and is absolutely continuous with respect to the Gaussian free field.","Expectations under the sine-Gordon measure reduce to Gaussian free field expectations of the form $\\mathbb E[\\Gamma(\\chi)f(A_1)]$, giving an explicit probabilistic representation of correlation functions.","The partition function of the corresponding Coulomb-type gas is represented as $\\exp(\\alpha Y_0(\\chi))$, where $Y_0(\\chi)$ is the initial value of the limiting quadratic BSDE.","At the specific parameters $\\alpha=2^{-1/2}C^2$, $\\beta=2^{-1/2}$ and with the conformal density, the partition function equals the scaling limit of the exponential moment of the critical planar XOR-Ising spin field tested against $\\chi$.","Normalized charge distributions of two-dimensional log-gases converge weakly, with the limiting characteristic function expressed through the sine-Gordon density."],"supporting_citations":[{"why":"Supplies the well-posedness and stability estimates for BSDEs with stochastic Lipschitz coefficients in Hilbert spaces, used to solve the linearized variational BSDE and to prove the Taylor expansion.","marker":"[13]"},{"why":"Provides the BMO martingale theory, including the uniform integrability of stochastic exponentials and the reverse H\\\"older inequality that control the density $\\Gamma(\\chi)$ uniformly in the cutoff.","marker":"[36]"},{"why":"Develops imaginary multiplicative chaos and the Onsager inequalities used to define the limiting terminal condition and to bound the partition functions.","marker":"[34]"},{"why":"Gives the universality result for subcritical complex Gaussian multiplicative chaos, which yields convergence in probability of the Wick-ordered cosine terminal conditions as the cutoff vanishes.","marker":"[38]"},{"why":"Provides the Green function estimates and Gaussian free field background, including the diagonal asymptotics on which the paper's uniform bound (3.4) relies.","marker":"[8]"},{"why":"Establishes the conformally invariant scaling limit of spin correlations in the planar Ising model, which the paper uses to connect its partition function to the XOR-Ising model.","marker":"[17]"}],"fun_headline_variants":["Quadratic BSDEs build the 2D sine-Gordon measure","Sine-Gordon measure via quadratic BSDEs","A BSDE route to sine-Gordon normalization","Quadratic BSDE limit gives sine-Gordon measure"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof requires the integral of the exponentially mollified diagonal Green function over the domain to stay uniformly bounded as the ultraviolet cutoff vanishes, and all subsequent uniform BMO and reverse-H\\\"older controls depend on that single estimate.","fun_headline_variants_meta":{"raw":{"variants":["Quadratic BSDEs build the 2D sine-Gordon measure","Sine-Gordon measure via quadratic BSDEs","A BSDE route to sine-Gordon normalization","Quadratic BSDE limit gives sine-Gordon measure"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000371,"raw_usage":{"total_tokens":2003,"prompt_tokens":980,"completion_tokens":1023,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":596,"completion_tokens_details":{"reasoning_tokens":958}},"tokens_in":596,"tokens_out":1023,"duration_ms":9998,"temperature":1.0,"reasoning_tokens":958,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:27:26.143131+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the left side of estimate (3.4) on the unit square for a mollified two-dimensional Dirichlet Green function: the asymptotic $A^\\varepsilon_\\Lambda(x)\\sim \\log(1/\\varepsilon)$ would make the integral diverge like a positive power of $1/\\varepsilon$ for $\\beta^2>0$, contradicting the uniform bound that supports Lemma 3.1 and Proposition 3.7.","supporting_citations":[{"cited_title":"BSDEs with stochastic Lipsc hitz condition and quadratic PDEs in hilbert spaces","cited_arxiv_id":null,"evidence_quote":"Supplies the well-posedness and stability estimates for BSDEs with stochastic Lipschitz coefficients in Hilbert spaces, used to solve the linearized variational BSDE and to prove the Taylor expansion."},{"cited_title":"Continuous Exponential Martingales and BMO , Springer-Verlag, 1994","cited_arxiv_id":null,"evidence_quote":"Provides the BMO martingale theory, including the uniform integrability of stochastic exponentials and the reverse H\\\"older inequality that control the density $\\Gamma(\\chi)$ uniformly in the cutoff."},{"cited_title":"Imaginary multiplic ative chaos: moments, regularity and connections to the Ising model","cited_arxiv_id":null,"evidence_quote":"Develops imaginary multiplicative chaos and the Onsager inequalities used to define the limiting terminal condition and to bound the partition functions."},{"cited_title":"A universality result for subcritical compl ex Gaussian multiplicative chaos","cited_arxiv_id":null,"evidence_quote":"Gives the universality result for subcritical complex Gaussian multiplicative chaos, which yields convergence in probability of the Wick-ordered cosine terminal conditions as the cutoff vanishes."},{"cited_title":"Conformal invari ance of spin correlations in the planar Ising model","cited_arxiv_id":null,"evidence_quote":"Establishes the conformally invariant scaling limit of spin correlations in the planar Ising model, which the paper uses to connect its partition function to the XOR-Ising model."}],"review_version":1}