{"id":"cfc68c30-6e2a-41d1-b2f4-ff81035ffacc","arxiv_id":"2511.02912","paper_version":5,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"SAC analytically continues a few noisy Rényi entropies to the von Neumann point (k=1), yielding stable entanglement-entropy estimates in randomized measurement experiments.","lead":"This paper develops a post-processing method, stabilized analytic continuation (SAC), that estimates the von Neumann entanglement entropy from a handful of Rényi entropy measurements. The method is designed to tolerate the statistical noise of real quantum-hardware experiments, and the authors demonstrate it on both simulated and trapped-ion data.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reliability of SAC is not established: Eq. (5) is a fixed linear functional of S_2...S_6, so states with identical low-order Rényi entropies but different S_vN receive the same estimate; no finite-data error bound or out-of-sample validation is provided.","rationale":"The reader's weakest assumption focuses on the analyticity domain of S_z(ρ). That is a legitimate concern, and the paper indeed only supports it with a conservative bound and numerical tests. However, the more load-bearing issue is the ill-posedness of the finite-data continuation itself: the noiseless estimator is a linear functional of S_2...S_6, so it cannot be exact for all spectra, and no error bound or out-of-sample validation is supplied. This does not invalidate the paper's empirical demonstrations, but it means the central claim 'reliably estimate' is supported only by a few in-sample benchmarks with tuned hyperparameters. The proposed concrete test would settle whether the method's implicit prior is strong enough by checking whether S_2...S_6 nearly determine S_vN for physically relevant spectra and whether SAC's extrapolation error remains small on that set. Since the reader already assigned CONDITIONAL and this concern reinforces that assessment without moving it to REJECT, the verdict should remain unchanged.","tokens_in":13403,"tokens_out":30095,"duration_ms":308187,"concrete_test":"For random 5-qubit reduced density matrices (d=32), compute exact S_2...S_6 and then numerically optimize the spectrum to maximize and minimize S_vN subject to fixed S_2...S_6 (e.g., via moment-constrained nonlinear optimization). If the achievable range of S_vN at fixed moments is large, no estimator using only these inputs can be reliable for all physical states. Then run the noiseless SAC formula Eq. (5) on the two extreme spectra and compare errors against the benchmark errors in Fig. 2; if the errors exceed those benchmarks, the method's reliability is not established and the variational selection is an unsupported prior.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that SAC reliably estimates S_vN from S_2...S_6. Even granting the assumed analyticity strip, the finite dataset does not determine S_vN. Equation (5) is an explicit linear combination of the six Rényi values; hence any two density matrices sharing S_2...S_6 but differing in S_vN yield the same SAC estimate, so the estimate cannot be correct for both. The pole-cancellation argument shows that the true discrepancy function has finite norm when alpha = S_vN, but it does not prove that the norm-minimizing alpha equals S_vN among all compatible analytic functions. The paper provides no error bound as a function of kmax, spectral properties, or noise level, and the free parameters (epsilon, eta, chi^2_0, w_0) are chosen by benchmarking on the same simulation data used to demonstrate accuracy, making the reported performance potentially in-sample. The reader's analyticity-domain concern is valid but secondary: it is a spectral-zero issue, whereas the deeper gap is that a six-point analytic continuation is an ill-posed inverse problem, and the variational selection rule acts as an unvalidated prior.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a stabilized analytic continuation (SAC) framework for estimating the von Neumann entropy S_vN from a finite set of integer Rényi entropies S_2,...,S_{kmax} obtained via randomized measurements. The construction builds a discrepancy function D_α(z)=(S_z−α)/(z−1), maps a presumed analyticity strip to the unit disk, and selects α by minimizing an L2 boundary norm. For noiseless data a closed-form estimator is given (Eq. 5); for noisy data, a χ²-constrained minimization is proposed. The method is benchmarked on simulated quench dynamics of a 10-qubit Néel state and applied to an existing trapped-ion dataset, with comparisons to Chebyshev and least-squares extrapolation. The paper claims noise robustness and suggests extensions to other nonlinear spectral functions such as logarithmic negativity and Rényi relative entropies.","tokens_in":13726,"tokens_out":11237,"duration_ms":120020,"significance":"If the central claim is validated, SAC would address a practical gap: randomized measurement experiments routinely estimate integer Rényi entropies, while the von Neumann entropy is the entanglement measure most often used in many-body physics and can behave differently from any fixed Rényi order. The paper's concrete contributions include the explicit closed-form estimator, the extension to correlated noise, and an open-source implementation. The numerical and experimental demonstrations are welcome. However, the central claim of reliability is not yet established. Equation (5) is a fixed linear functional of the input Rényi values, so it cannot distinguish states that share S_2,...,S_{kmax} but differ in S_vN. The variational argument does not prove that the norm-minimizing α equals S_vN, and the free parameters are tuned on the same simulation data used to demonstrate accuracy. These issues are load-bearing for the paper's main claim and require substantial additional work.","major_comments":[{"comment":"Equation (5) is a fixed linear functional of S_2,...,S_{kmax}: the matrix A and the terms 1/(i−1)−1 depend only on the chosen conformal map and data locations, not on the state. Consequently, any two density matrices with identical low-order Rényi entropies receive the same SAC estimate, no matter how different their S_vN. The numerical benchmarks (Fig. 2) and the experimental analysis (Fig. 3) use parameters ε, η, χ²0, w0 chosen by benchmarking on the same simulation data, so the reported accuracy is partly in-sample. To support the word 'reliably' in the abstract and introduction, the paper needs either a finite-data error bound under the stated analyticity assumptions or an explicit stress test in which the low-order Rényi data are held fixed while S_vN varies, showing that the method's error is controlled. As written, the estimate is an extrapolation of the input Rényi function and t","section":"Stabilized analytic continuation, Eq. (5)"},{"comment":"The pole-cancellation argument is not sufficient for the variational quantity actually computed. If α≠S_vN, the true discrepancy function D'_α has a pole on |w|=1, and its boundary norm (4) would be singular. However, Step 1 does not minimize over functions containing that boundary singularity; it minimizes over analytic functions Y_α holomorphic in the disk with finite norm that interpolate the data. This feasible set is nonempty for every α (e.g. polynomial interpolation), so δ(α) is finite for all α. The fact that the true function has an infinite norm at α≠S_vN does not imply that the norm-minimizing continuation does. No theorem or numerical experiment is provided to show that the quadratic δ(α) is minimized at α=S_vN rather than at another value. This is the central methodological gap. I would ask for a proof under the stated analyticity and norm assumptions, or at least a syntheti","section":"Stabilized analytic continuation, Steps 1–2 and Eq. (4)"},{"comment":"The paper treats ε, η, χ²0, and w0 as free parameters. The text says ε and η are chosen 'based on benchmarks on the simulation data' (after Eq. 3), χ²0 is 'typically chosen O(k_max)' (after Eq. 6), and w0 is a variational parameter in the noisy case. No sensitivity analysis is reported. Because the same simulation data are used both to set these parameters and to measure accuracy, the comparison against Chebyshev and least-squares in Fig. 2 is not fully out-of-sample. The assumed analyticity strip is supported only by a conservative perturbative bound and by 'direct numerical tests' that are not shown; for the 10-qubit quench trajectory, zeros of Tr ρ^z could enter the assumed strip. Since the exact density matrix is available in the simulation, a check of the strip width along the actual evolution would directly address this concern. Please provide a systematic parameter scan or a princ","section":"After Eq. (3) and Eq. (6): analyticity domain and parameter selection"},{"comment":"The experimental demonstration does not include a valid uncertainty estimate. The authors state that bootstrapping the 500 density matrices gives unrealistically small error bars, splitting gives grouping-dependent results, and a robust error bar requires larger N_u. As a result, the trapped-ion S_vN points in Fig. 3 are presented without error bars, despite the paper's emphasis on noise robustness. This is a significant limitation of the experimental demonstration and should be stated clearly in the main text. Ideally, the analysis should be repeated with N_u=1000 or validated on synthetic data with the same covariance structure, so that the error-bar procedure can be checked against a known answer.","section":"Application to Trapped-ion Quantum Simulator, Fig. 3"}],"minor_comments":[{"comment":"The sentence 'S_vN = lim_{k→1+} S_k, a consequence of Carlson’s theorem' is imprecise. The limit follows from the continuity/removable-singularity property of S_z at z=1; Carlson's theorem is about uniqueness of an analytic continuation from integer values under boundedness conditions. Please rephrase.","section":"Introduction, p. 2"},{"comment":"The paper chooses one of four possible L2 norms, motivated by the pole manifesting in the imaginary part. Since this choice is part of the regularization, a short discussion or numerical comparison of the four norms would help the reader assess how much the final estimate depends on this choice.","section":"Eq. (4) and footnote [45]"},{"comment":"The notation is slightly compressed: p_k is used both as the quantum expectation Tr(ρ^k) and as the U-statistic estimator \\p_k. Please distinguish the estimator explicitly (e.g. \\hat p_k) to avoid confusion.","section":"Randomized measurements, around Eq. (7)"},{"comment":"Some typographical issues in the text: 'N´ eel' should be 'Néel' in several places; 'analytical continuation' is used where 'analytic continuation' is standard. These are minor and do not affect the science.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the journal's scope and the idea is timely. The main obstacle is not that the method disagrees with consensus, but that the central identifiability issue is not addressed: Eq. (5) is a linear functional of the Rényi data, and the variational argument does not prove that the selected α recovers S_vN. If the authors can provide a convincing stress test on states with matched low-order Rényi entropies and a sensitivity analysis of the free parameters, the paper could become publishable. I would not reject outright, but the current claims outrun the evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's the paper you asked about.\n\nThe takeaway: this is a genuinely useful practical method—a closed-form SAC estimator for the von Neumann entropy from a few integer Rényi entropies (k=2..6), with a correlated-noise extension and a trapped-ion demonstration—but the reliability claim is stronger than the evidence supports. The method is a heuristic regularization, not a proven continuation, and the paper would benefit from saying so plainly.\n\nWhat's new: adapting Ciulli-Spearman SAC to the Rényi-to-von Neumann problem is a sensible move. The closed-form formula (5) is nice and likely useful. The correlated-noise extension is worked out in the appendix with a Lagrange multiplier solution. The numerical benchmarks show SAC beating Chebyshev and least-squares fitting on their quench example, and the trapped-ion application is a useful proof of concept. The authors are also upfront about the parameter choices and about the failure of bootstrapping for error bars—refreshingly honest.\n\nThe soft spots are real but not fatal. First, the estimator is a fixed linear functional of S_2...S_6. Two states with identical low-order Rényi entropies but different S_vN will receive the same estimate, so any claim of \"reliable\" estimation depends on the norm-minimization prior being right. There is no finite-data error bound, and the six-point continuation is genuinely ill-posed. Second, the free parameters—ε, η, χ²0, w0—are chosen by benchmarking on the simulation data that is then used to demonstrate accuracy. That is in-sample tuning. Third, the analyticity domain is supported by a conservative perturbative bound and numerical tests, not a proof. The branch-point issue is secondary to the identifiability gap, but still unproven. The experimental error bars are acknowledged to be unreliable, which is honest but limits the strength of that application.\n\nNone of this kills the contribution. It does mean the paper is a toolbox paper, not a theorem. It will be useful to experimentalists who want S_vN from existing Rényi data and to anyone working on analytic continuation. I'd send it to peer review, with a referee asked to scrutinize the parameter selection and the identifiability claim.\n\nRecommendation: a serious referee should engage with this paper, and the authors should be pushed to soften the reliability language and add a cross-validation or failure-mode analysis.","headline":"A useful, honest toolbox paper: SAC gives a practical route from a few noisy Rényi entropies to S_vN, but \"reliable\" is too strong a word—finite data can't pin down S_vN without an unvalidated prior.","tokens_in":14229,"tokens_out":3569,"would_cite":true,"duration_ms":39456,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.-a","03.65.Ud"],"model":"deepseek-v4-flash","headline":"Stabilized analytic continuation turns a few noisy Rényi entropies into a reliable estimate of the von Neumann entropy in quantum simulation experiments.","keywords":["von Neumann entropy","Rényi entropies","analytic continuation","randomized measurements","classical shadows","entanglement entropy","trapped-ion simulator","quantum simulation"],"falsifier":"Take a density matrix whose spectrum is known (for example, a small subsystem of the quenched trapped-ion state), compute the first branch points of S_z(ρ)= (1/(1−z)) log₂ Tr ρ^z, and check whether any branch point falls within the strip |Im z|<ε used by SAC. If one does, feed the same Rényi inputs to SAC and to an exact-diagonalization estimate of S_vN; a bias in the SAC estimate larger than its reported uncertainty would show that the pole-cancellation argument fails when the analyticity domain is narrower than assumed.","tokens_in":13280,"feed_emoji":"⚛️","tokens_out":7895,"duration_ms":77769,"temperature":0.7,"pith_summary":"The paper aims to make the von Neumann entanglement entropy measurable in practice. Randomized measurement protocols directly access Rényi entropies of integer order, and the von Neumann entropy is the k→1 limit of the same analytic family, but the continuation is usually unstable under noise. The paper claims that stabilized analytic continuation (SAC) solves this: by adding a pole at z=1, mapping the analyticity strip to the unit disk, and selecting the continuation of smallest boundary norm, the noise is tamed. Benchmarks on simulated quench data and on trapped-ion experimental data support this, and the same framework extends to other non-polynomial diagnostics such as logarithmic negativity and Rényi relative entropies. If correct, the method closes a practical gap between what randomized measurements can measure and the entanglement quantifier many-body theory most wants.","feed_headline":"Five noisy Rényi entropies can recover the von Neumann entropy","feed_subtitle":"Stabilized analytic continuation turns randomized-measurement data into S_vN, tested on trapped-ion quench experiments.","key_machinery":"SAC (stabilized analytic continuation) is the central mechanism: given analytic function values at interior points of the unit disk, it selects the holomorphic function with the smallest L² norm of the angular derivative of the imaginary part on the boundary (Eq. 4). The paper's adaptation first defines a discrepancy function with an artificial pole at the target point z=1, then maps the semi-infinite strip Re z>1, |Im z|<ε to the disk via ξ=cosh((z−1)/ε + iπ/2), w=(ξ−η i)/(ξ+η i), so that the von Neumann entropy becomes a residue whose cancellation is measured by the norm. The matrix A_ij (Eq. 14) encodes the geometry of the data locations and turns the minimization into a linear algebraic","core_discovery":"The central claim is that the von Neumann entropy S_vN(ρ) = -Tr(ρ log₂ ρ) can be estimated from a finite set of noisy Rényi entropies S_k(ρ), k=2,...,k_max, by analytic continuation. The paper constructs a discrepancy function D_α(z) = S_z(ρ)/(z−1) − α/(z−1) whose residue at z=1 is S_vN(ρ) − α. After a two-step conformal map sends the analyticity strip into the unit disk, the true value of α cancels an artificial pole on the unit circle, causing the minimum of an L² boundary norm to drop sharply; minimizing that norm over α therefore recovers S_vN. In the noiseless case this yields a closed-form expression (Eq. 5); with noise, the data are treated as a covariance ellipsoid and the same norm","pith_inferences":["A direct extension the paper leaves implicit: because SAC needs only trace moments Tr ρ^k, it can run as a post-processing add-on on classical shadow data already collected for other observables, making S_vN nearly free in existing experiments.","The same pole-placement trick should transfer to other replica-trick quantities that are limits of integer-order data; logarithmic negativity and Rényi relative entropies are natural targets, and the closed-form Eq. (5) can likely be rederived for each.","A stress test worth running is k_max=2 or 3: the noiseless formula subtracts the first data point and sums from k=3, so the method's practical boundary lies in how few orders can still stabilize the continuation.","The analyticity-strip assumption is the main risk; computing the actual branch points of S_z(ρ) for the simulated states would either validate the strip width ε used by SAC or reveal where the method must be modified."],"forward_implications":["The von Neumann entanglement entropy becomes extractable from randomized measurement data using only Rényi orders 2 through 6, avoiding full state tomography.","The same SAC routine applies to any non-polynomial function of the density matrix accessible through an analytic continuation, including logarithmic negativity and Rényi relative entropies.","The noise-resilience of the method means existing randomized measurement datasets, even those with finite and correlated statistical errors, can be reprocessed to obtain S_vN.","An explicit noiseless estimator (Eq. 5) is available, so the method is computationally cheap and can be used as a standard post-processing tool for quantum simulator experiments.","Benchmarks indicate SAC is more accurate than polynomial extrapolation, which suggests the continuation problem for entanglement entropies is tractable rather than fundamentally ill-posed."],"fun_headline_variants":["Analytic continuation recovers von Neumann entropy from noisy Rényi data","SAC method: five noisy Rényi entropies yield S_vN","Residue trick in stabilized analytic continuation extracts von Neumann entropy","Noisy Rényi entropies? SAC still gives von Neumann entropy"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The Rényi function S_z(ρ) is assumed to stay analytic on a semi-infinite strip Re z>1, |Im z|<ε that is wide enough for the conformal map, a property the paper supports with a conservative perturbative bound and numerical tests rather than a proof.","fun_headline_variants_meta":{"raw":{"variants":["Analytic continuation recovers von Neumann entropy from noisy Rényi data","SAC method: five noisy Rényi entropies yield S_vN","Residue trick in stabilized analytic continuation extracts von Neumann entropy","Noisy Rényi entropies? SAC still gives von Neumann entropy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000677,"raw_usage":{"total_tokens":2904,"prompt_tokens":719,"completion_tokens":2185,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":463,"completion_tokens_details":{"reasoning_tokens":2114}},"tokens_in":463,"tokens_out":2185,"duration_ms":16438,"temperature":1.0,"reasoning_tokens":2114,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T00:01:17.628484+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a density matrix whose spectrum is known (for example, a small subsystem of the quenched trapped-ion state), compute the first branch points of S_z(ρ)= (1/(1−z)) log₂ Tr ρ^z, and check whether any branch point falls within the strip |Im z|<ε used by SAC. If one does, feed the same Rényi inputs to SAC and to an exact-diagonalization estimate of S_vN; a bias in the SAC estimate larger than its reported uncertainty would show that the pole-cancellation argument fails when the analyticity domain is narrower than assumed.","supporting_citations":[],"review_version":1}