{"id":"67881321-5164-41f0-af8e-489eee8e61ec","arxiv_id":"2607.22863","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Finite-depth parameterized quantum circuits have an emergent critical correlation length that grows with depth as D^kappa, with fitted exponents from about 1 to 3 across eight Ising-model ansatze, giving a common benchmark for variational efficiency.","lead":"This paper assigns each finite-depth variational quantum circuit an emergent correlation length that grows algebraically with depth, and measures a scaling exponent that ranks how efficiently different circuit architectures reproduce long-distance correlations of a critical quantum state.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Algebraic-growth assumption for the exponential-kernel ansätze is never tested against ξ_D ∼ e^{cD}; the reported κ≈3 may be a finite-window artifact.","rationale":"The paper is careful and transparent: it uses exact free-fermion numerics, reports two independent extraction routes, and includes explicit caveats (Sec. IVC2) about power-law ansätze leaving the Ising universality class. Those caveats are the reader's identified weak point, and they are real, but the paper itself already flags them, so they do not by themselves change the conditional verdict. The concern I find most load-bearing is more fundamental and less explicitly tested: the framework's defining object, ξ_D = A D^κ (Eq. (3)), is assumed rather than established. 'Compatible with algebraic growth' is a weak statement when the longest usable depth window for the headline EXP combined ansatz is D=2–5 (Table I). The exponential kernel f_λ(r)=λ^{r−1} has a natural length scale, λ is optimized layer-by-layer, and the approximation-theory explanation offered in the paper (Beylkin–Monzón) is precisely a mechanism by which a few exponentially spaced scales reproduce correlations over exponentially large distances. Over four depths, ξ_D ∼ e^{cD} is difficult to distinguish from D^3 with a log-log fit; the small jackknife errors only measure stability within the chosen model, not the validity of the algebraic model. If EXP growth is exponential, the universal benchmark fails for the ansatz that the abstract highlights most. The proposed test—extended depths/system sizes and explicit model comparison against exponential and mixed forms—is within reach because the numerics are exact and free-fermion. Should the test reject algebraic growth, the right outcome would be to restrict the framework's claim to the observed window rather than claim a universal κ; should it confirm algebraic growth, the conditional acceptance is justified.","tokens_in":32827,"tokens_out":11414,"duration_ms":114940,"concrete_test":"Recompute ξ_D(D) for EXP combined (Eq. (32)) at L≥2048 for D=1,...,12 and compare fits ξ_D = A D^κ, ξ_D = B e^{cD}, and ξ_D = A D^κ e^{cD} by AIC or weighted residuals on log ξ_D. Also repeat the κ_C, κ_E extraction with a prespecified bulk-energy filter tolerance. If the exponential or mixed model is preferred, or if κ changes by more than ~0.5 when D=6–8 are added, then the reported κ≈3 is not a universal algebraic figure of merit.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Eq. (3), Table I) requires ξ_D = Aξ D^κ. The paper only states that all ansätze are 'compatible with algebraic growth' in the accessible window; it does not discriminate against non-algebraic growth. For the EXP ansätze (Eqs. (31)–(32)) this is the most consequential gap. Their generator kernel f_λ(r)=λ^{r−1} carries an intrinsic length s=−1/lnλ, and λ_d is an independent variational parameter in every layer (Fig. 5). A product of fermionic Gaussian layers with geometrically spaced decay constants naturally mimics a scale-invariant correlator up to a cutoff that grows exponentially in D — precisely the exponential-sum mechanism (Beylkin–Monzón) the paper invokes to explain the large exponents. In log-log coordinates an exponential ξ_D = B e^{cD} is nearly linear on D=2–5, the four depths that survive the finite-depth filter for EXP combined (Table I), so it can masquerade as D^3 with a tiny jackknife error. If the true growth is exponential, κ is not a well-defined asymptotic exponent and the paper's headline separation between EXP and POW collapses. The energy-based κ_E is not an independent safeguard: Eq. (37) also presupposes ξ_D = A D^κ, feeding the same functional form into the collapse.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a finite-resource scaling framework for parameterized quantum evolutions at criticality, in which each variational architecture is assigned an emergent correlation length ξ_D and an efficiency exponent κ defined by ξ_D = A_ξ D^κ. The framework is applied to the critical transverse-field Ising model using exact fermionic-Gaussian numerics for a range of ansätze: separable and combined layers with nearest-neighbor, exponential, and power-law generators. The authors report that all ansätze are compatible with algebraic growth in the accessible depth window, with fitted exponents ranging from κ ≈ 1 to κ ≈ 3, that exponential kernels give the largest exponents, and that combined layers generally outperform separable HVA layers. A quasiparticle-occupation analysis is used to interpret ξ_D as an infrared resolution scale. The paper is careful about many of its own limitations, but the central quantitative claims rest on fits over very narrow depth windows and on functional-form assumptions that are not independently tested.","tokens_in":33084,"tokens_out":4934,"duration_ms":49557,"significance":"If the central claim were established, the framework would provide a useful, architecture-independent benchmark for critical-state preparation and a practical guide for designing layered variational ansätze. The paper has clear strengths: the numerics are exact within the fermionic-Gaussian representation, the two extraction routes (correlations and energy) agree for the TFIM HVA, and the quasiparticle interpretation is physically appealing and testable. However, the stress-test concern that the headline EXP exponents may be a finite-window artifact of fitting D^κ to data that are also consistent with exponential growth is real and lands. The algebraic-growth hypothesis is asserted rather than discriminated from alternatives, and the energy-based determination is not fully independent of the correlation-based one because it assumes the same functional form and borrows the amplitude A_ξ. For a paper whose main claim is a universal scaling law with architecture-dependent exponents, these are load-bearing issues.","major_comments":[{"comment":"The central claim of algebraic growth, ξ_D = A_ξ D^κ, is never tested against non-algebraic alternatives. For the EXP combined ansatz the energy fit uses only depths D = 2–5 (n≥3_D = 4), and the correlation-length fit covers a similarly narrow window. Over four depths an exponential growth ξ_D ∝ exp(cD) is nearly linear in log-log coordinates and can masquerade as D^κ with a small jackknife error, so the reported κ = 3.07 may be a finite-window artifact. Because the headline separation between the exponential and power-law families rests on these values, the authors should add a quantitative model-selection step, for example fitting ξ_D to D^κ, exp(cD), and a stretched form, with reported residuals or an information criterion, and should extend the accessible range so that at least six to eight depths satisfy L ≫ ξ_D. The statement in the conclusions that the largest exponents 'should be confirmed at larger system sizes' does not resolve the issue for the present claims.","section":"Table I and Section IVC2, EXP rows"},{"comment":"The correlation-length extraction defines ξ_D through the assumed exponential envelope exp(−r/ξ_D), applied to binned maxima over r/L ∈ [5%, 40%], and the deepest TFIM HVA point is excluded because it 'falls outside the observed scaling trend.' This means that Eq. (3), ξ_D = A_ξ D^κ, is not an independent observation but an inference from the same exponential model used to define ξ_D, and the exclusion rule is not specified in advance. I ask the authors to diagnose the envelope shape without assuming it, for example by computing the local logarithmic slope of the chord-corrected correlator as a function of r, and to report the sensitivity of κ_C to the binning window, to the fitting range, and to inclusion or exclusion of the deepest circuit. A transparent, prespecified criterion for dropping depths is needed.","section":"Section IVC1, Eq. (C9), and Fig. 2"},{"comment":"The energy-based exponent κ_E is not an independent determination of κ. The fit assumes ξ_D = A_ξ D^κ inside Eq. (37), and the physical constraint on the logarithmic-cutoff constant Q is imposed using A_ξ and D_min taken from the correlation-length analysis. The agreement between κ_C and κ_E therefore partly reflects a shared algebraic hypothesis and shared input data rather than a confirmation from two fully independent observables. This is especially consequential for the EXP ansätze, where only four depths enter the energy fit. The authors should quantify how much κ_E changes when the constraint on Q is relaxed or when A_ξ is varied within its uncertainty, and should present the energy analysis as a consistency check conditional on the same scaling hypothesis rather than as an independent route to κ.","section":"Section IIB, Eqs. (36)–(37), and Appendix C2b"},{"comment":"The paper correctly warns that sufficiently slowly decaying power-law couplings may drive the ansatz outside the short-range Ising universality class, in which case the Δ_ε = 1 energy scaling used in Eq. (37) is not valid. Nevertheless, Table I reports κ_E for POW HVA and POW combined on the same footing as the other ansätze, and the abstract's conclusion that power-law interactions remain close to nearest-neighbor behavior draws on those energy exponents. The authors should either restrict the energy analysis to the range of α where the Ising CFT description is justified, or provide an explicit validation of the Δ_ε = 1 assumption, for example by checking the L/ξ_D collapse of the bulk energy density with an independently determined ξ_D that does not presuppose the same scaling form.","section":"Section IVC2, POW caveat and Table I"}],"minor_comments":[{"comment":"Both POW HVA and POW combined are labeled with Eq. (31) (and similarly Eq. (32) for the two EXP rows); please append the kernel family to the equation label or state it explicitly in the ansatz column to avoid ambiguity.","section":"Table I"},{"comment":"The criterion behind the statement that 'all ansätze are compatible with algebraic growth' is not quantitative; please specify a threshold or goodness-of-fit measure used for compatibility so that the claim is reproducible.","section":"Section IVC2"},{"comment":"The caption of Fig. 2(b) reports the fit as (0.786 ± 0.028) D^(1.03 ± 0.027), while the text quotes κ_C = 1.03 ± 0.03; the non-universal amplitude A_ξ should be defined and reported consistently, since Appendix C2b uses A_ξ to constrain Q.","section":"Fig. 2 and Eq. (34)"},{"comment":"The variable-projection description should state explicitly that the least-squares weights are w ∝ y^(−2) and that the jackknife resampling deletes all system sizes belonging to one depth at a time; this is already implied but should be stated in the main fitting protocol.","section":"Appendix C2b"},{"comment":"Several log-axis tick labels are rendered ambiguously (e.g., '100' and '100.5' can be read as powers of ten or as decimal numbers); please use consistent exponent notation such as 10^0, 10^0.5.","section":"Figures 6–9"}],"recommendation":"major_revision","confidential_remarks":"The framework is attractive and the exact numerics are a real asset, but the central quantitative claim, especially the κ ≈ 3 result for the exponential ansätze, is currently supported by only four depths and by a fitting procedure that assumes the algebraic form it seeks to establish. I would like to see a model-selection test against exponential growth and an independence check for the energy route before this paper can make its advertised universal-scaling claim. If those tests cannot be performed at larger depth, the paper should be reframed as a finite-window comparative study. The authors are commendably candid about several of these limitations, but the abstract and Table I currently state the conclusions more strongly than the evidence supports."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The paper imports tensor-network finite-resource scaling to parameterized quantum evolutions and uses it to compare eight ansätze for the critical TFIM. That is a genuinely useful move, and the exact free-fermion numerics make the comparisons crisp. But the headline quantitative claim, that the exponential-kernel ansätze reach κ≈3, rests on fits over four or five depths in a log-log plot, and the paper never tests the algebraic-growth hypothesis against exponential growth. The stress-test note is right: on D=2-5 a ξ_D ~ e^{cD} can masquerade as D^3. The authors are honest that these are window-dependent estimates, but the abstract and Table I still treat κ as the architecture's figure of merit. If EXP actually grows exponentially, the EXP-versus-POW separation in κ collapses; they would need to be compared on a different footing.\n\nWhat the paper does well: it is transparent. It gives exact Gaussian-state numerics, describes the filtering criteria, reports jackknife rather than naive fit errors, and explicitly flags that the largest exponents need larger systems. The two extraction routes (correlation envelope and energy collapse) agree for the nearest-neighbor HVA, and the quasiparticle analysis showing an unresolved window around the gap-closing modes is a nice microscopic picture. The framework itself, ξ_D ~ D^κ as a common benchmark, is a good idea even if all the fitted values turn out to be non-universal.\n\nSoft spots, in proportion. The biggest is the missing model comparison: an exponential fit to the same ξ_D data would settle whether κ is meaningful. That should be a mandatory revision, not a suggestion. Minor: the correlation envelope is defined by binned maxima over r/L in [5%,40%], an ad hoc but clearly stated choice; the energy fit reuses A_ξ from the correlation fit, so the two routes are not fully independent; no code or data is released, which matters for a paper whose claims are fits over narrow windows.\n\nBottom line: this deserves a serious referee. The framework and the qualitative ordering (exponential kernels better than power-law, combined better than separable) are plausible and worth publishing, but not before the authors test the growth form and release the data. I would accept for peer review with a request for major revision; the math is sound, the execution is honest, and the central caveat is fixable in a way that would make the paper much stronger.","headline":"A useful framework with a load-bearing fitting caveat: the κ≈3 for exponential ansätze needs an explicit test against exponential growth before the benchmark claim is solid.","tokens_in":33614,"tokens_out":2906,"would_cite":true,"duration_ms":34476,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Every parameterized quantum evolution has an emergent correlation length $\\xi_D \\propto D^\\kappa$, with $\\kappa$ from about 1 to about 3, yielding a universal benchmark for critical-state preparation.","keywords":["parameterized quantum evolutions","emergent correlation length","finite-resource scaling","critical ground states","transverse-field Ising model","variational quantum ansatz","long-range interactions","conformal perturbation theory"],"falsifier":"Repeat the correlation- and energy-based scaling extractions for the exponential-combined ansatz at system sizes $L=2048$ and $L=4096$ over depths $D$ up to 12. If $\\kappa_C$ and $\\kappa_E$ cease to agree within their error bars, or if $\\xi_D$ stops growing as a pure power law in $D$, the single-emergent-length claim is falsified.","tokens_in":1930,"feed_emoji":"⚛️","tokens_out":3636,"duration_ms":137968,"temperature":0.7,"pith_summary":"Parameterized quantum evolutions — layered unitary circuits — become benchmarkable resources. The paper assigns each ansatz an emergent correlation length $\\xi_D$, the longest distance over which it faithfully captures critical correlations, and shows it grows algebraically with circuit depth, $\\xi_D \\propto D^\\kappa$. For the critical transverse-field Ising chain, fitted exponents range from $\\kappa\\simeq1$ for nearest-neighbor and power-law generators to $\\kappa\\simeq3$ for exponential-range generators, so exponential kernels convert depth into long-distance correlations most efficiently. A quasiparticle analysis shows finite depth leaves an unresolved momentum window of width $\\xi_D^{-1}$. The framework provides a universal finite-resource scaling benchmark for critical-state preparation.","feed_headline":"Exponential-range circuits reach κ≈3; power-law stay near 1","feed_subtitle":"A single scaling exponent now benchmarks how efficiently finite-depth circuits capture critical correlations.","key_machinery":"The central object is the emergent correlation length $\\xi_D$, operationally defined from the exponential envelope of the chord-corrected spin correlator: $C_{XX}(r)\\,d_{\\mathrm{chord}}(r,L)^{1/4} \\propto \\exp(-r/\\xi_D)$. The argument rests on the RG hypothesis that the optimized finite-depth state is the critical target theory perturbed by the leading Z2-even operator ($\\Delta_\\epsilon=1$), which fixes $\\nu=1$ and yields the energy-error scaling $L\\,\\delta E_{\\mathrm{abs}}=P L^2 D^{-2\\kappa}(\\kappa\\ln D+Q)$. In the fermionic quasiparticle basis, occupation errors collapse with the same $\\xi_D$, making the infrared cutoff visible on each mode.","core_discovery":"The central discovery is that a finite-depth parameterized quantum evolution targeting a critical ground state is governed by a single emergent length scale: the optimized state behaves at long distances as the critical transverse-field Ising theory perturbed by the only symmetry-allowed relevant operator, the Z2-even energy operator of scaling dimension $\\Delta_\\epsilon=1$. This perturbation opens a correlation length $\\xi_D$ that cuts off critical correlations, and the algebraic growth law $\\xi_D=A_\\xi D^\\kappa$ holds for every ansatz studied. Two independent extractions — the exponential envelope of the chord-corrected spin correlator and the conformal-perturbation scaling of the energy error — give consistent $\\kappa$ values, from about 1 to about 3. Quasiparticle occupations collapse as a function of $|k-N|\\xi_D/L$, showing that finite depth leaves an unresolved window of width $\\xi_D^{-1}$ at the Fermi surface. The paper thereby establishes $\\kappa$ as a common, ansatz-independent figure of merit for converting depth into long-range correlations.","pith_inferences":["The same two-observable protocol could produce a catalog of $\\kappa$ for interacting and higher-dimensional critical models, turning finite-resource scaling into a practical benchmark for hardware-native ansätze.","The exponential-kernel advantage suggests a design heuristic: variational generators should expose several independently tunable decay scales, as exponential kernels do, rather than a single scale-free profile.","The unresolved-window picture predicts that circuit resources are best spent on degrees of freedom near the gap-closing point, a testable hypothesis for fixed-depth optimization budgets.","An analytic link between $\\kappa$ and the generator algebra, such as the dynamical Lie algebra rank and quantum Fisher information conditioning, could predict exponents before optimization; the paper lists this as a future direction."],"forward_implications":["Architectures with larger $\\kappa$ need polynomially fewer layers to reach a given correlation length, making $\\kappa$ a practical benchmark for critical-state preparation.","Exponential-range generators are the most depth-efficient family studied, while power-law and nearest-neighbor generators sit near $\\kappa\\approx1$, so long-range connectivity alone does not create an advantage.","Combining non-commuting generators inside a single layer generally yields larger $\\kappa$ than separable HVA layers, making layer organization a genuine design axis alongside interaction range.","Finite depth leaves an unresolved momentum window of width $\\xi_D^{-1}$ around the gap-closing modes, so the variational problem reduces to resolving the non-analytic Fermi step of the critical ground state.","The correlation-envelope and energy-collapse extraction protocols, together with the bulk-energy finite-depth filter, transfer to any ansatz with a refinement parameter."],"supporting_citations":[{"why":"Establishes the finite-resource scaling idea from tensor networks, namely that finite variational resources induce an emergent correlation length, which this paper transfers to parameterized evolutions.","marker":"[54–64]"},{"why":"Provides the conformal perturbation theory of finite-resource variational states as critical theories with a relevant perturbation, backing the effective-Hamiltonian picture used throughout.","marker":"[68]"},{"why":"Supplies the conformal perturbation formulas for energy shifts and the logarithmic correction that appear in Eq. (37).","marker":"[70–72]"},{"why":"Gives the exact critical transverse-field Ising ground-state energy used to define the variational energy error $\\delta E_{\\mathrm{abs}}$.","marker":"[85, 86]"},{"why":"Provides the fermionic Gaussian-state formalism that makes all state evolutions, gradients, and observables numerically exact at large system sizes.","marker":"[92]"},{"why":"Motivates the explanation of why exponential kernels outperform power-law ones through independently tunable decay scales.","marker":"[93]"},{"why":"Introduces the hybrid digital–analog long-range interaction architectures that motivate the long-range generator families studied here.","marker":"[13]"},{"why":"Documents the non-universal low-energy behavior of slowly decaying power-law systems, grounding the paper's caution about power-law energy exponents.","marker":"[94–97]"}],"fun_headline_variants":["Depth-to-correlation exponent κ from 1 to 3 benchmarks critical-state circuits","Exponential-range circuits hit κ≈3; power-law stay near 1","Emergent length scale sets universal limit for finite-depth critical-state accuracy","Long-range support alone gives no scaling advantage: κ≈1 for power-law","Layer organization matters: combined generators beat separable in critical states"],"cache_read_input_tokens":35712,"weakest_assumption_plain":"The whole scaling analysis collapses if the optimized finite-depth state is not, at long distances, the critical ground state perturbed by just its energy-density operator; slowly decaying power-law couplings may break that assumption, as the paper itself notes.","fun_headline_variants_meta":{"raw":{"variants":["Depth-to-correlation exponent κ from 1 to 3 benchmarks critical-state circuits","Exponential-range circuits hit κ≈3; power-law stay near 1","Emergent length scale sets universal limit for finite-depth critical-state accuracy","Long-range support alone gives no scaling advantage: κ≈1 for power-law","Layer organization matters: combined generators beat separable in critical states"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002072,"raw_usage":{"total_tokens":8115,"prompt_tokens":1051,"completion_tokens":7064,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":667,"completion_tokens_details":{"reasoning_tokens":6968}},"tokens_in":667,"tokens_out":7064,"duration_ms":46085,"temperature":1.0,"reasoning_tokens":6968,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:28:38.166895+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the correlation- and energy-based scaling extractions for the exponential-combined ansatz at system sizes $L=2048$ and $L=4096$ over depths $D$ up to 12. If $\\kappa_C$ and $\\kappa_E$ cease to agree within their error bars, or if $\\xi_D$ stops growing as a pure power law in $D$, the single-emergent-length claim is falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the conformal perturbation theory of finite-resource variational states as critical theories with a relevant perturbation, backing the effective-Hamiltonian picture used throughout."},{"cited_title":"Argüello-Luengo, T","cited_arxiv_id":null,"evidence_quote":"Provides the fermionic Gaussian-state formalism that makes all state evolutions, gradients, and observables numerically exact at large system sizes."},{"cited_title":"Argüello-Luengo, A","cited_arxiv_id":null,"evidence_quote":"Motivates the explanation of why exponential kernels outperform power-law ones through independently tunable decay scales."}],"review_version":1}