{"id":"2596a63f-30de-47d3-9cf3-4bf868d03b67","arxiv_id":"2501.12259","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"For lattice QCD spectral reconstruction, the Wertevorrat from Nevanlinna-Pick interpolation bounds the analytic-continuation uncertainty, and in a model test this bound falls roughly exponentially with the number of data points.","lead":"This lattice QCD conference talk explains how Nevanlinna-Pick interpolation, a tool from complex analysis, can put rigorous error bars on spectral reconstructions from Euclidean correlation functions. It also reports a new numerical observation: the uncertainty bound shrinks roughly exponentially as more imaginary-frequency data points are used.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Wertevorrat's rigor rests on an asserted, unproven disk-to-disk map; for bosonic correlators the value transform is nontrivial, and the omega=0 Matsubara point maps to the boundary, so the Pick-theorem hypotheses are not self-evident.","rationale":"Reading the talk in good faith: the aim is to publicize a framework from Ref. [29] and one new empirical observation (Fig. 2). The framework is standard Nevanlinna-Pick theory, and the talk is honest that noisy data can violate Pick positivity. The main advertised value is the claim that the Wertevorrat gives a rigorous, model-independent handle on analytic-continuation uncertainty. The load-bearing condition is that the true Green function is genuinely in the Schur class after the two conformal maps. That condition is asserted in a single sentence and never demonstrated; for bosonic correlators, where the spectral density is not positive definite over the whole real axis, the needed map is nontrivial and only cited. If the map is wrong, the bound does not bound the physical quantity. The l=0 boundary issue is a small but concrete manifestation of the same unexamined hypotheses. The reader's CONDITIONAL verdict already captures this; I do not think the talk deserves REJECT because it is a short proceedings contribution, clearly labels the scaling as empirical, and explicitly flags the open statistical problem. The concern should keep the verdict at CONDITIONAL, and the proposed reproduction test would settle it.","tokens_in":9722,"tokens_out":15429,"duration_ms":169932,"concrete_test":"Implement the omitted bosonic codomain map from Ref. [29] for the Bernecker-Meyer vector correlator used in Figs. 1-2, then check: (i) |f(ζ)|<1 on a dense grid of ζ∈D; (ii) whether the l=0 Matsubara point is used and whether its domain image lies in the open disk; (iii) positive semidefiniteness of the Pick matrix (Eq. (15)) for N=60 and N=100 exact data. If (i) or (iii) fails, the Wertevorrat computed in Fig. 2 is not a bound on the physical R(s,ε). If l=0 lies on the boundary, rerun the scaling excluding it and with a small additive noise drawn from a realistic covariance; if percent precision at N=60 is lost or the Pick condition breaks, the paper's practical claim is not yet supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the Wertevorrat Δ_N(ζ) rigorously contains all possible analytic continuations holds only if, after the domain Cayley transform and the codomain map of Ref. [29], the physical Green function is a Schur function f:D→D. Section 4 asserts this end result but does not state or derive the bosonic codomain map. The step is not a formality: the retarded bosonic Green function is not Herglotz on the full upper half-plane (its spectral density is odd), so the simple value-space Cayley map used for fermions does not put the data in D, and one must rely on an unstated construction whose Schur property is unverified here. If that construction fails, Pick's theorem constrains only a transformed object, and Corollary 1 does not bound the true spectral reconstruction. A second, independent gap is that Eq. (11) writes {ζ_l}∈D, but for the l=0 Matsubara frequency C(0)=-1 lies on the boundary of D, so Theorem 2's open-disk hypothesis is violated unless that point is excluded; the talk does not say so. The paper honestly concedes that statistical errors can make the Pick matrix indefinite, so both the rigorous bound and the exponential scaling in Fig. 2 are established only for noiseless data satisfying all hypotheses. The new percent/permille claims inherit every one of those conditions, so the central claim is conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This proceedings contribution from LATTICE summarizes a new perspective on the lattice-QCD spectral reconstruction inverse problem, arguing that smeared spectral functions are the practically and conceptually right target (Sections 2–3), and then presents the Nevanlinna–Pick (NP) interpolation framework as a way to bound the systematic uncertainty of analytic continuation from a finite set of Euclidean data (Section 4). The central technical object is the Wertevorrat, the disk of all possible analytic continuations through given interpolation data, whose radius and center are given in Corollary 1. The talk's new contribution is an empirical observation, shown in Fig. 2 for a Bernecker–Meyer model of the R-ratio, that the fractional uncertainty of the Wertevorrat decreases roughly exponentially with the number of interpolation points, reaching percent and permille precision at 60 and 100 points for one fixed extrapolation point.","tokens_in":9975,"tokens_out":5941,"duration_ms":60536,"significance":"If the hypotheses of the NP framework are satisfied by lattice-QCD Green functions, the Wertevorrat is a parameter-free, rigorous bound on the systematic uncertainty of analytic continuation from finitely many Euclidean data points, with no model assumptions and no ad hoc regularization. This would be a genuinely useful tool for inclusive hadronic quantities, where the full spectral function is needed and smeared observables are the natural interface with experiment. The paper is honest about an important limitation: statistical uncertainties can violate the Pick condition, and the paper correctly describes the treatment of such indefinite Pick matrices as an open question. It also cites and connects to the relevant literature, including the HLT method, Bayesian/GP reinterpretations, and prior NP work in condensed matter. The new exponential-scaling observation is potentially important for planning calculations, but in this manuscript it is purely empirical and rests on a narrow set of tests.","major_comments":[{"comment":"The statement that the mapped Euclidean data satisfy {ζ_l} ∈ D is false for bosonic Matsubara frequencies: for ω_0 = 0, ζ_0 = C(0) = -1 lies on the boundary of D rather than in the open unit disk. Theorem 2, as stated, requires interpolation points in the open disk, so Corollary 1 cannot be applied to the bosonic example in Fig. 1 unless that point is excluded or a separate boundary-limiting argument is supplied. The manuscript should state explicitly how the l = 0 Matsubara point is handled.","section":"Section 4, Eq. (11)"},{"comment":"The claimed rigor of the Wertevorrat bound depends on the physical Green function becoming a Schur function—a bounded analytic map from D to D—after the Cayley transform of the domain and the codomain map of Ref. [29]. This is asserted but not established in the manuscript: the bosonic retarded Green function is not Herglotz on the entire upper half-plane, so the codomain construction is nontrivial. If that map does not produce a Schur function, Pick's theorem constrains only a transformed object and Corollary 1 does not bound the true spectral reconstruction. Please either give the codomain map explicitly or clearly state that the bound is conditional on the construction in Ref. [29].","section":"Section 4, paragraph after Eq. (10)"},{"comment":"The new scaling result is presented as 'roughly exponential' but no derivation or robustness checks are given. The figure uses a single model (Bernecker–Meyer), a single extrapolation point (s = 0.65 GeV², ε = 0.1), and, as in Fig. 1, exact Euclidean data. The statement that '60 and 100 points suffice to determine R(s, ε) with percent and permille precision' is therefore a model- and point-specific empirical observation, not an established general property. The manuscript should either provide a derivation of the scaling or add checks that vary the spectral model, the smearing ε, and the target energy, and should qualify the caption accordingly.","section":"Section 4, Fig. 2 and preceding bullet"}],"minor_comments":[{"comment":"The phrase 'Poisson kernel of Eq. (6a)' is a typo: Eq. (6a) defines the Gaussian kernel, while the Poisson kernel is Eq. (6b).","section":"Section 4, text near Eq. (10)"},{"comment":"The sentence 'One might image that a suitably smooth function function could, in fact, be well determined' contains a duplicated word 'function'.","section":"Section 2, first paragraph"},{"comment":"For clarity, the manuscript should specify that only nonnegative Matsubara frequencies l ≥ 0 are used in the interpolation, since negative frequencies lie on the negative imaginary axis, which is outside the upper half-plane where the retarded Green function is analytic and on which the Cayley transform is defined.","section":"Section 4, Eq. (9) and Eq. (11)"}],"recommendation":"major_revision","confidential_remarks":"This is a conference-proceedings writeup whose main new claim—the exponential scaling of the Wertevorrat with the number of points—is empirical and not yet supported by derivation or robustness checks. The boundary-point issue and the unverified Schur condition are technical gaps in the presentation of the central bound, but they appear fixable in a revision by adding explicit hypotheses and caveats. The reliance on the author's own Ref. [29] is substantial but is not a circularity problem, as that reference provides the codomain construction. For the journal's standard, I see the appropriate recommendation as major_revision; if the venue is a proceedings series that explicitly tolerates conditional statements with deferred proofs, a carefully qualified minor revision might be acceptable, but as written the claims outrun the demonstrated hypotheses."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a proceedings talk that re-presents the Wertevorrat method from the author's earlier PRD paper and adds one genuinely new observation—that the bound shrinks roughly exponentially with the number of interpolation points, reaching percent and permille precision at 60 and 100 points for the Bernecker–Meyer R-ratio model. If that scaling survives contact with noisy, truncated, finite-volume data, it is a practically useful result for spectral reconstruction.\n\nThe talk does several things well. The Nevanlinna–Pick theorem is standard, quoted correctly, and the recursive Schur-algorithm construction is summarized accurately. The worked R-ratio example is concrete and the connection between analytic continuation and Poisson smearing is nicely explained. The author is also honest: the exponential scaling is explicitly flagged as an empirical observation, and the paper concedes that statistical errors can break the Pick-matrix positivity condition.\n\nThe soft spots are real but proportionate. The reduction of the physical Green function to a Schur function (disk-to-disk analytic map) is asserted rather than derived. For bosonic correlators that step is nontrivial—the retarded Green function is not Herglotz on the full upper half-plane—and the talk just points to Ref. [29]. That is acceptable for a proceedings summary if the derivation exists, but it leaves the reader unable to check the key hypothesis. Second, Eq. (11) says the interpolation points {ζ_l} lie in D, but the l=0 Matsubara frequency maps to the boundary of the unit disk, so the open-disk hypothesis of Pick's theorem is violated unless that point is explicitly excluded. The talk does not say so. Neither gap is fatal, but both should be stated cleanly.\n\nThe new scaling result is based on a single noiseless model, with no code, data, or robustness checks. That is a limitation, but the author does not overclaim: the figures are presented as an example, not a theorem.\n\nFor a proceedings paper, this is solid work. It is not a full research article, but it is a fair, readable summary of a mathematically grounded method with one new quantitative observation. I would send it to referees, with the expectation that they ask for a clarification of the boundary-point issue and a more explicit statement about the bosonic codomain map assumptions.","headline":"A clean proceedings talk: the Wertevorrat method from Ref. [29] plus one genuinely new empirical scaling observation; honest about limits, though it glosses over two technical hypotheses that matter.","tokens_in":10515,"tokens_out":2182,"would_cite":true,"duration_ms":23610,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["12.38.Gc"],"model":"deepseek-v4-flash","headline":"Nevanlinna–Pick interpolation bounds every possible analytic continuation of lattice-QCD Euclidean data in an explicit disk—the Wertevorrat—whose size falls roughly exponentially with the number of data points.","keywords":["lattice QCD","spectral reconstruction","analytic continuation","Nevanlinna–Pick interpolation","Wertevorrat","smeared spectral functions","inverse Laplace transform","R-ratio"],"falsifier":"Take a known spectral model (e.g., the R-ratio model used in the paper's numerical tests), generate exact Euclidean data, apply the paper's conformal maps, and evaluate the mapped Green function on a dense grid in the unit disk: if any point has modulus greater than 1, the boundedness premise fails and the Wertevorrat is not guaranteed to contain the true smeared spectral function. On the data side, realistic noisy lattice data that yield an indefinite Pick matrix would likewise show the theorem's condition is not met.","tokens_in":9480,"feed_emoji":"🎯","tokens_out":18175,"duration_ms":164092,"temperature":0.7,"pith_summary":"Lattice QCD computes Euclidean correlation functions at a finite set of times, but many inclusive observables—the R-ratio for $e^+e^-\\to$ hadrons, inclusive $\\tau$ and $B$ decays, neutrino-nucleus scattering, transport coefficients—require the full spectral function, which is an inverse Laplace transform away and notoriously ill-posed. The paper argues that this inverse problem becomes tractable when viewed as analytic continuation: evaluating the momentum-space Green function at $z=\\omega+i\\epsilon$ is exactly a Poisson-smoothed spectral function, and the known analytic structure (singularities only on the real axis) can be exploited. After conformally mapping both the frequency plane and the Green-function values onto the unit disk, Nevanlinna–Pick interpolation gives, in the Wertevorrat, a disk that provably contains every analytic continuation consistent with the finite data, with no model assumptions. The new result in this talk is empirical: for a realistic R-ratio model, the radius of that disk decreases roughly exponentially with the number of Euclidean points, reaching percent precision at 60 points and permille at 100 for a fixed smearing.","feed_headline":"Rigorous error disks shrink exponentially with lattice-QCD data points","feed_subtitle":"With 60 Euclidean time points, the R-ratio bound hits one percent; with 100, one per mille.","key_machinery":"The load-bearing mechanism is the reduction of the inverse problem to Nevanlinna–Pick interpolation. A Cayley transform maps the upper half plane to the unit disk and a second conformal map sends the Green function's values into the disk, so the physical analytic continuation becomes an analytic function $f:\\mathbb{D}\\to\\mathbb{D}$ (a Schur function) interpolating the transformed data $(\\zeta_\\ell,w_\\ell)$. Nevanlinna's theorem, built by Schur's algorithm with Blaschke factors, rewrites every interpolant as a fractional-linear expression in one arbitrary Schur function $f_N$; the set of values that expression can attain at a fixed $\\zeta$ is the Wertevorrat, a disk whose center and radius are explicit rational functions of the data. Pick's theorem supplies the existence condition: the interpolation problem has a solution if and only if the Pick matrix $(1-w_i\\bar{w}_j)/(1-\\zeta_i\\bar{\\zeta}_j)$ is positive semidefinite. The key point is that no model for the spectral function enters: only the data and the analyticity structure (singularities on the real line) determine the bounding disk.","core_discovery":"The central claim is that the systematic uncertainty of analytic continuation from a finite set of Euclidean points can be completely characterized, not just estimated. Given values $G(i\\omega_\\ell)$ at equally spaced points on the imaginary axis, the paper treats the Green function as an analytic function and applies the Cayley transform to both domain and codomain, so the problem becomes: find an analytic $f:\\mathbb{D}\\to\\mathbb{D}$ with $f(\\zeta_\\ell)=w_\\ell$. Nevanlinna's theorem says all solutions are parametrized by one arbitrary Schur function $f_N$ through explicit Nevanlinna coefficients, and the Wertevorrat—the set of all values $f(\\zeta)$ can take at any target point—is a disk with computable center and radius. The paper claims this disk rigorously contains all possible analytic continuations and therefore provides a complete, model-independent bound on the smeared spectral function obtained at $z=\\omega+i\\epsilon$; the total width of its imaginary part after mapping back to the physical plane is the uncertainty statement. It further reports the new empirical observation that, for the R-ratio model, the fractional uncertainty falls roughly exponentially with the number of interpolation points, so 60 and 100 points give percent and permille precision respectively for a smearing $\\epsilon=0.1$ near the $\\rho$ peak.","pith_inferences":["The exponential scaling is demonstrated on one R-ratio model; if it persists for spectral functions with sharp thresholds, the practical lesson generalizes—the number of points, not the statistical noise, sets the achievable resolution—an inference the paper does not make.","Pick-matrix failure under statistical noise, which the paper flags as an open question, suggests a natural preprocessing step: project noisy Euclidean data onto the nearest set that makes the Pick matrix positive semidefinite, and only then compute the Wertevorrat; the paper points at this problem but does not solve it.","The same disk bound could be read as a diagnostic of the analyticity assumptions themselves: if a lattice correlator cannot be made Schur-bounded after the conformal maps at any reasonable smearing, that would indicate a missing singularity or a finite-volume artifact, turning the method into a consistency test for the Euclidean data.","Used together, the localized-average point estimate and the Wertevorrat envelope would give both a best reconstruction and a rigorous band around it; the paper presents them as alternatives, and the synthesis is left implicit."],"forward_implications":["For inclusive observables such as the R-ratio, the method turns 'how many Euclidean times do we need?' from a guess into a computable question: the Wertevorrat gives the systematic error directly, and 60–100 well-chosen points reach percent-to-permille smeared reconstructions in the tested model.","Because the bound relies only on analytic structure, the same machinery applies to $\\tau$ decays, inclusive $B$-meson semileptonic decays, neutrino-nucleus structure functions, and transport coefficients whenever the Euclidean correlator's singularities lie on the real axis.","The arbitrary Schur function $f_N$ in Nevanlinna's theorem is a bookkeeping device for missing information: any future constraint on $f_N$, physical or algorithmic, can only shrink the Wertevorrat, so additional input translates directly into a smaller rigorous uncertainty band.","The exponential point-count scaling gives a practical roadmap for lattice calculations: determine how many points are needed for a target smearing and precision first, then optimize the correlator computation accordingly rather than maximizing statistics blindly.","The Wertevorrat's growth as $\\epsilon\\to 0$ quantifies the cost of going differential: uncertainty grows roughly exponentially as the smearing is removed, making the trade-off between energy resolution and controlled systematic error explicit."],"supporting_citations":[{"why":"Introduces the Nevanlinna–Pick conformal-mapping setup for hadronic structure and applies the Wertevorrat to the R-ratio; the present talk extends it with the point-count scaling study.","marker":"[29]"},{"why":"Establishes Nevanlinna analytic continuation for Green functions, supplies the disk-to-disk mapping idea, and notes that statistical noise can break the Pick condition.","marker":"[38]"},{"why":"Pick's theorem, the positive-semidefinite existence condition that the whole bounded-interpolation framework depends on.","marker":"[34]"},{"why":"Provides the recursive construction of Nevanlinna coefficients used to evaluate the Wertevorrat at any target point.","marker":"[35]"},{"why":"Supplies the R(s) parameterization used for the numerical reconstruction and scaling plots.","marker":"[49]"},{"why":"The previous spectral-density extraction method based on optimized smearing kernels whose stability analysis provides the comparative baseline for the new bound.","marker":"[20]"},{"why":"Introduces smeared spectral functions and the ordered large-volume/smearing limit that justify computing smeared inclusive observables.","marker":"[19]"},{"why":"A recent lattice calculation of the smeared R-ratio that motivates the question of how narrow the smearing can be and the need for controlled systematic bounds.","marker":"[2]"}],"fun_headline_variants":["Rigorous error disks shrink exponentially with lattice data","Complex analysis gives tight rigorous bounds for spectral functions","60 points to 1% error: new bound for lattice QCD R-ratio","Nevanlinna method sets rigorous uncertainty disks for spectral data","Exponential error reduction in lattice QCD spectral reconstruction"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on the assumption that, after the conformal maps, the QCD Green function is a bounded analytic function from the unit disk to itself and that the Euclidean data satisfy the Pick positivity condition; if either fails, the Wertevorrat may not contain the true analytic continuation.","fun_headline_variants_meta":{"raw":{"variants":["Rigorous error disks shrink exponentially with lattice data","Complex analysis gives tight rigorous bounds for spectral functions","60 points to 1% error: new bound for lattice QCD R-ratio","Nevanlinna method sets rigorous uncertainty disks for spectral data","Exponential error reduction in lattice QCD spectral reconstruction"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000344,"raw_usage":{"total_tokens":1846,"prompt_tokens":857,"completion_tokens":989,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":473,"completion_tokens_details":{"reasoning_tokens":904}},"tokens_in":473,"tokens_out":989,"duration_ms":9958,"temperature":1.0,"reasoning_tokens":904,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:20:28.427953+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a known spectral model (e.g., the R-ratio model used in the paper's numerical tests), generate exact Euclidean data, apply the paper's conformal maps, and evaluate the mapped Green function on a dense grid in the unit disk: if any point has modulus greater than 1, the boundedness premise fails and the Wertevorrat is not guaranteed to contain the true smeared spectral function. On the data side, realistic noisy lattice data that yield an indefinite Pick matrix would likewise show the theorem's condition is not met.","supporting_citations":[{"cited_title":"Fei, C.-N","cited_arxiv_id":null,"evidence_quote":"Establishes Nevanlinna analytic continuation for Green functions, supplies the disk-to-disk mapping idea, and notes that statistical noise can break the Pick condition."},{"cited_title":"Pick,Über die Beschränkungen analytischer Funktionen, welche durch vorgegebene Funktionswerte bewirkt werden, Math","cited_arxiv_id":null,"evidence_quote":"Pick's theorem, the positive-semidefinite existence condition that the whole bounded-interpolation framework depends on."}],"review_version":1}