{"id":"dfa10cda-123d-4232-be0c-e31843fab5fb","arxiv_id":"2506.15638","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A phase-shift scrambler placed between two successive squeezings restores joint estimability of both squeezing parameters, and joint estimation outperforms stepwise estimation in this model.","lead":"Two successive squeezing operations on a light field usually combine into one effect, so the individual squeezing amounts cannot be separated. This paper shows that inserting a phase-shift scrambler between them allows both parameters to be estimated at once, with joint estimation beating the stepwise approach.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Joint-vs-stepwise advantage is supported only by an upper-bound comparison and collapses to equality at α=0, so 'always better' is not established.","rationale":"The reader's verdict is CONDITIONAL, and this stress-test agrees that the paper is plausible but has a support gap. The reader's weakest_assumption focuses on the lossless unitary model, which is a modeling limitation but not an internal flaw. The more load-bearing issue is that the headline 'joint estimation is always better than separate one' is inferred from comparing stepwise bounds to C_Q(1+T_I), which is only an upper bound on the true joint Holevo bound. A rigorous superiority claim would need the true C_H or a lower bound on it. The α=0 case is particularly telling because direct substitution with the paper's own QFIM and Uhlmann curvature yields exact equality between the optimized stepwise bound and C_Q(1+T_I), so the plotted strict inequality cannot hold there. This does not disprove a joint advantage, but it shows the presented evidence does not establish the strong 'always better' conclusion. The paper also appears to mislabel Eq. (30) as S when the displayed expression is actually C_Q, which further weakens the 'optimized phases' narrative; however, that is secondary to the comparison issue. Since the core physics may still be correct and the requested revision is to strengthen the comparison rather than to overturn the result, the conditional verdict remains appropriate.","tokens_in":12002,"tokens_out":21338,"duration_ms":176352,"concrete_test":"Compute the exact Holevo bound C_H for the pure two-parameter Gaussian model, e.g., with the semidefinite program of Albarelli et al. or the closed-form two-parameter Gaussian expression, over a grid including α=0 and small λ1, and compare with the optimized Csep from Eq. (38). If C_H = Csep at α=0, the 'always better' claim is false as stated and should be weakened to 'no worse' or restricted to α>0; if C_H < Csep, the advantage survives but the paper must replace the upper-bound comparison with a valid exact or lower-bound comparison.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that joint estimation is always better than stepwise is not actually established by the comparison in Fig. 5. The plotted joint quantity C_Q(1+T_I) is an upper bound on the Holevo bound C_H (Eq. 15), not a lower bound on it. A strict advantage claim requires C_H < C_sep, or at least an upper bound strictly below C_sep over the whole domain. But at α=0, using the paper's own QFIM and Uhlmann curvature (Q11=8, Q22=8 cosh^2(2λ1), U12=8 cosh(2λ1), so T_I=2 cosh(2λ1)/(1+cosh^2(2λ1))), one obtains exactly C_sep = C_Q(1+T_I) for all λ1. Hence the statement that C_sep is 'larger' than the joint upper bound fails in this case, and the numerical evidence does not support 'always better' for the vacuum probe; it supports only C_H ≤ C_sep. For α>0 the inequality is checked only numerically over an unspecified grid, with no analytic proof.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies the joint estimation of two successive squeezing parameters (hyperbolic phases) acting on a coherent state of a single field mode. Because the two squeezings commute in the absence of an intermediate operation, the output state depends only on the sum of the parameters and the statistical model is sloppy. The authors insert a phase-shift scrambler V=exp(-i φ a†a) between the two unitaries and compute the quantum Fisher information matrix and the Uhlmann curvature for the resulting pure Gaussian model. They claim that the phases φ=θ=π/4 minimize the sloppiness, that the joint-estimation precision bound C_Q(1+T_I) is below the optimized stepwise bounds C_sep1 and C_sep2 over the explored parameter range, and that a suitably optimized general-dyne measurement can approach the quantum Holevo bound in the large-α and large-λ1 regime. The paper contains explicit analytic expressions for the QFIM and Uhlmann curvature in Appendix A, and the comparison with stepwise and general-dyne strategies is developed in Sections IV-VI.","tokens_in":12263,"tokens_out":19470,"duration_ms":183334,"significance":"The paper addresses a timely question, namely whether an intermediate scrambling operation can remove sloppiness in a continuous-variable two-parameter estimation problem. The model is clean, the calculations are explicit and standard (pure-state QFIM and Uhlmann curvature, Gaussian Fisher information for general-dyne detection), and the general-dyne analysis gives a concrete, experimentally meaningful measurement family. If the central claims were fully established, the paper would provide a useful CV counterpart to the qubit and interferometric results in Refs. [34,39]. However, the manuscript currently contains an apparent mislabeling of the central sloppiness formula, an unproven phase-optimization claim, and a joint-versus-stepwise comparison that supports a weaker statement than the one asserted. The paper is transparent about its ideal noiseless unitary model, which is a strength because it makes the scope of the claims clear, but the headline 'always better' conclusion needs to be either proven or significantly qualified.","major_comments":[{"comment":"The quantity displayed in Eq. (30) and labeled S is not the sloppiness S=1/det[Q] defined in Eq. (19). Using the QFIM elements in Appendix A at φ=θ=π/4, Eq. (30) evaluates to C_Q=Tr[Q]/det[Q]. For example, at α=0 Eq. (30) gives C_Q=(1+cosh^2(2λ1))/(8 cosh^2(2λ1)), whereas 1/det[Q]=1/(64 cosh^2(2λ1)). The two differ by a factor of order cosh^2(2λ1). This is not a purely cosmetic typo because the subsequent discussion of the optimized sloppiness and Fig. 2 use this expression. Please correct either the label or the formula, and re-derive the statement that the optimized quantity is S rather than C_Q.","section":"Section III, Eq. (30) and Fig. 2"},{"comment":"The claim that φ=θ=π/4 minimize the sloppiness is introduced with 'Upon inspecting analytically' but no derivation is provided. This is a load-bearing step because all subsequent quantum bounds (Figs. 2-5) and the comparison with stepwise estimation use these phases. Similarly, in Section VI the statement that 'the optimization leads to' θ=0, φ=π/4 and z=exp(2λ2) for the general-dyne measurement is asserted without showing the stationarity or global-minimum conditions. A proof or an explicit derivation of both optimizations should be included; if a full proof is not available, the claims should be downgraded to 'we choose these values' and the optimality statements removed.","section":"Section III and Section VI, phase optimization"},{"comment":"The conclusion that 'joint estimation is always better than separate one' is not established by the comparison in Fig. 5. The plotted joint quantity C_Q(1+T_I) is an upper bound on the Holevo bound C_H, and an upper bound strictly below C_sep would indeed certify a joint advantage, but the authors do not provide an analytic proof of this inequality, and the numerical comparison is presented without specifying the grid. Moreover, at α=0 the inequality is an equality: with Q11=8, Q22=8 cosh^2(2λ1) and U12=-8 cosh(2λ1), Eq. (38) gives C_sep1=C_sep2=C_Q(1+T_I) for all λ1. Thus the vacuum-probe case does not support 'always better' in any strict sense. Please either prove the strict inequality for α>0 and handle α=0 separately, or reformulate the conclusion to state that joint estimation is at least as good in the explored region and strictly better in the numerically investigated α>0 cases.","section":"Section IV, Fig. 5 and conclusion"}],"minor_comments":[{"comment":"The text says that Csep1 and Csep2 'are equal', but the optimized formulas Cmin_sepk=1/Qkk+QkkS+2√S are generally different when Q11≠Q22. For α>0 and λ1≠0 these two bounds differ (though they may be close numerically). Please correct the statement and specify which curve is shown in Fig. 5.","section":"Section IV, Eqs. (34)-(38)"},{"comment":"The sentence stating that for φ=0 or φ=π/2 the first moments and variances 'depend only on the sum ... or difference ... thus allowing their joint estimation' is misleading: dependence on only one combination prevents estimation of both parameters jointly. Presumably the intended statement is that the sum and difference can be estimated in separate configurations, or that this illustrates the decoupling mechanism.","section":"Section V, after Eq. (47)"},{"comment":"In the second stepwise scenario, the sentence says the total variance is bounded by Csep1/M, but the quantity defined in Eq. (35) is Csep2. Please correct the label.","section":"Section IV, text near Eq. (35)"},{"comment":"The paper assumes ideal lossless unitary dynamics throughout. Since the conclusions are phrased as unconditional ('always better', 'approaches the optimal precision'), the ideal-model limitation should be restated explicitly in the Conclusions and the claims should be understood as valid for the noiseless model only.","section":"General"},{"comment":"There are several presentation issues: Fig. 2 mentions a 'black region' that is likely a color-region artifact; the notation C=1/det[U] and S=1/det[Q] is easy to confuse with C_Q and should be renamed (e.g., C_inc and S_slop); and Ref. [34] is cited as a 2024 arXiv preprint, which should be updated if a published version is available.","section":"General and references"}],"recommendation":"major_revision","confidential_remarks":"The paper is closely related to the authors' own previous work on overcoming sloppiness (Ref. [34]) and to the companion paper Ref. [39]; its incremental contribution is a continuous-variable, two-squeezing instance of the scrambling idea. The referee's main concern is not the overall approach but the gap between the claims and the evidence: Eq. (30) appears to be mislabeled, the phase optimization is unproven, and the 'always better' joint-versus-stepwise claim rests on an upper-bound comparison that fails to be strict at α=0. These issues are fixable with a corrected formula, an explicit optimization proof or a weakened claim, and a more careful statement of the numerical evidence. I do not see a fundamental error in the model or in the QFIM/Uhlmann derivation, so major revision rather than rejection is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nHere's my take on arXiv:2506.15638. The useful core is a straightforward, explicit application of the scrambling trick from the group's earlier Mach-Zehnder and qubit papers to two successive squeezing unitaries. They derive the QFIM and Uhlmann curvature elements in closed form, give the sloppiness expression, and show numerically that with optimized phases the joint Holevo-bound sandwich C_Q(1+T_I) dips below optimized stepwise bounds for non-vacuum probes. The general-dyne analysis is also a nice addition: they identify a feasible measurement with a classical bound within a factor of about four of the quantum limit at large squeezing.\n\nWhat I'd flag is the central claim in Section IV. The paper says 'joint estimation is always better than separate one' because the stepwise bounds are larger than C_Q(1+T_I) in the entire region. That's an upper bound on C_H, not the bound itself, and the inequality fails at α=0. Using their own Appendix formulas, at α=0 one gets C_sep = C_Q(1+T_I) exactly for all λ1, so there is no strict advantage for a vacuum probe. The claim of universal superiority is therefore not established; at best it holds for α>0 based on numerical inspection over an unspecified grid, and for α=0 the two strategies tie. This is a load-bearing overstatement only in the conclusion's wording; the underlying recipes and formulas remain valid.\n\nThe second soft spot is small: the optimization ϕ=θ=π/4 is asserted from 'inspecting analytically' without proof. I'm confident it can be verified by symmetry or a short calculation, but as written it is an unsupported step. Also, the stepwise comparison uses the joint-model QFIM diagonal elements, which is fine, but it would help to state explicitly the assumption that the first step's estimate of one parameter is used to know the other.\n\nOverall: the mathematics is mostly standard and the derivations in Appendix A check out. The paper is a useful reference for anyone working on sloppiness in CV multiparameter metrology. It deserves a serious referee, though a revision should qualify the joint-vs-stepwise conclusion and either prove or explicitly verify the phase optimization. I'd bring it to a reading group and would cite it for the scrambling technique in CV settings.\n\nBest.","headline":"A clean analytical extension of the group's scrambling recipe to successive squeezings, but the 'joint is always better' claim overreaches: at α=0 the stepwise and joint bounds coincide.","tokens_in":12752,"tokens_out":2907,"would_cite":true,"duration_ms":28281,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Inserting a phase shift between two successive squeezings makes the two squeezing parameters jointly estimable, and with optimized phases joint estimation beats stepwise estimation.","keywords":["quantum metrology","multiparameter estimation","sloppiness","squeezing parameters","quantum Fisher information","Holevo Cramér-Rao bound","general-dyne detection","phase-shift scrambling"],"falsifier":"With the paper's own lossless formulas, one can scan $\\alpha\\ge0$ and $\\lambda_1\\ge0$ and compare $C_{\\mathrm{sep}}^{\\min}$ with $C_Q(1+T_I)$; a single point where $C_{\\mathrm{sep}}^{\\min} \\le C_Q(1+T_I)$ would contradict the statement that joint estimation is always better in this model. Alternatively, in a real experiment, measure the joint and stepwise Fisher informations under the optimized phases $\\varphi=\\theta=\\pi/4$ and see whether the joint advantage survives photon losses.","tokens_in":11819,"feed_emoji":"⚛️","tokens_out":9874,"duration_ms":87569,"temperature":0.7,"pith_summary":"Two successive squeezing operations applied to one optical mode normally hide their individual strengths: the output state depends only on their sum, so the quantum Fisher information matrix is singular and neither parameter can be extracted alone. This paper shows that inserting a tunable phase-shift scrambler between the two squeezings makes the two parameters independently estimable, and that optimizing the scrambler phase and the probe phase at $\\pi/4$ minimizes the remaining sloppiness. Comparing the quantum bounds, the authors find that joint estimation after scrambling is always more precise than stepwise estimation in the explored parameter range. They also show that a realistically implementable general-dyne measurement, with its noise parameter adapted from a first estimate of one squeezing strength, can approach the quantum precision limit in the large-amplitude regime.","feed_headline":"Insert a phase shift, and two squeeze strengths become measurable","feed_subtitle":"A scrambling phase between the two operations restores estimability, and joint estimation beats doing the two measurements one at a time.","key_machinery":"The load-bearing object is the phase-shift scrambler $\\hat V=e^{-i\\varphi\\hat a^\\dagger\\hat a}$, inserted between the two squeezing unitaries $\\hat U_k=e^{-i\\lambda_k\\hat G/2}$ with generator $\\hat G=\\hat a^2+\\hat a^{\\dagger 2}$. Its role is to rotate the field quadratures between the two squeezings so that the output state depends on the individual parameters rather than only on their sum. Alongside it, the calculation relies on the closed-form quantum Fisher information and Uhlmann curvature matrices for the coherent-state probe, and on the optimized choice $\\varphi=\\theta=\\pi/4$, which minimizes the sloppiness quantifier $S=1/\\det Q$. For the detection stage, the paper uses general-dyne POVMs with added noise covariance $\\sigma_m=\\frac{1}{2}\\mathrm{diag}(z,z^{-1})$, optimized at $z=e^{2\\lambda_2}$, to show that a feasible Gaussian measurement nearly saturates the quantum bounds.","core_discovery":"The central claim is that a phase-shift scrambling operation $\\hat V=e^{-i\\varphi\\hat a^\\dagger\\hat a}$ placed between the two squeezing unitaries $\\hat U_k=e^{-i\\lambda_k\\hat G/2}$, with generator $\\hat G=\\hat a^2+\\hat a^{\\dagger 2}$, removes the model's sloppiness and makes the two hyperbolic phases $\\lambda_1$ and $\\lambda_2$ jointly estimable. With the probe in a coherent state $|\\alpha\\rangle$ and optimized phases $\\varphi=\\theta=\\pi/4$, the quantum Fisher information matrix is no longer singular, and the sloppiness $S=1/\\det Q$ is given by a closed expression that decreases with $\\lambda_1$ and with the coherent amplitude. The authors show that the optimized joint bound $C_Q(1+T_I)$ lies below the optimized stepwise bounds $C_{\\mathrm{sep}1}$ and $C_{\\mathrm{sep}2}$ throughout the explored parameter space, so they conclude that joint estimation is always better than separate estimation for this model. For large $\\lambda_1$ the Holevo bound collapses onto the SLD bound, making repeated independent preparations sufficient for optimal precision, while optimized general-dyne detection with $z=e^{2\\lambda_2}$, $\\theta=0$, $\\varphi=\\pi/4$ gives a classical bound $C_g$ satisfying $C_H/C_g\\simeq 1/4$ when both $\\alpha$ and $\\lambda_1$ are large.","pith_inferences":["The same 'open the encoding box and insert a rotation' recipe should apply to other chains of non-commuting unitary encodings; a natural extension is to insert scramblers between every pair of unknown operations in a longer squeezing sequence.","The optimized general-dyne setting assumes knowledge of $\\lambda_2$ to set $z$, so a practical implementation requires a two-stage adaptive protocol; the cost of that first estimate is not quantified here, and one could test whether the factor-4 gap survives when adaptation overhead is included.","In a lossy optical channel the covariance matrix acquires additional terms, and the exact cancellation that makes $\\varphi=\\theta=\\pi/4$ optimal may be disturbed; a concrete next step is to optimize the scrambling phase under photonic loss and check whether the joint advantage over stepwise estimation persists.","The asymptotic ratio $C_H/C_g\\simeq 1/4$ suggests that noise-matched Gaussian measurements can nearly saturate the Holevo bound for multiparameter Gaussian models; whether exact saturation is possible in other regimes is a natural conjecture."],"forward_implications":["Two squeezing parameters that are individually unidentifiable without the scrambler become jointly estimable at finite precision once a phase shift is inserted between the two squeezings.","With optimized phases, the joint-estimation bound $C_Q(1+T_I)$ remains below the best stepwise bounds $C_{\\mathrm{sep}1}^{\\min}$ and $C_{\\mathrm{sep}2}^{\\min}$ over the studied range, so splitting the experimental runs between two separate single-parameter estimates is never preferable in this lossless model.","For large true values of the first squeezing strength $\\lambda_1$, the Holevo bound approaches the SLD quantum Cramér-Rao bound, so repeated independent preparations of the probe are enough to reach the ultimate precision; the small-$\\lambda_1$ regime instead likely requires collective entangled measurements.","A general-dyne measurement with adaptively chosen noise parameter $z=e^{2\\lambda_2}$ yields a classical precision within a factor roughly 4 of the quantum bound in the large-amplitude, large-$\\lambda_1$ regime."],"supporting_citations":[{"why":"Introduces the scrambling approach to mitigate sloppiness in continuous-variable models, which this paper applies to successive squeezings.","marker":"[34]"},{"why":"Extends scrambling optimization to multiparameter qubit estimation, providing the template for optimizing the hidden phase shift.","marker":"[39]"},{"why":"Defines the stepwise estimation strategy and bounds that the paper compares against joint estimation.","marker":"[44]"},{"why":"Provides the weight-dependent Holevo bound chain used to bracket the ultimate joint precision.","marker":"[28]"},{"why":"Supplies the multiparameter estimation framework and the notion of sloppy models with degenerate Fisher information.","marker":"[17]"},{"why":"Establishes the hyperbolic-phase interpretation of squeezing that motivates the estimated parameters.","marker":"[37]"},{"why":"Provides the phase-space description used to derive the general-dyne Fisher information.","marker":"[45]"}],"fun_headline_variants":["Phase shift restores estimability of two squeezing parameters","Phase scrambling makes two squeeze strengths jointly measurable","Joint estimation beats stepwise after phase-shift scrambling","Two squeezes one phase: joint estimation becomes feasible"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the whole process is lossless and ideal: the probe is a pure coherent state and both the squeezing operations and the inserted phase shift are exact unitaries, so no photon loss or excess noise enters the Fisher information; if a real optical channel adds losses, the precision bounds change and the claimed joint advantage may weaken or disappear.","fun_headline_variants_meta":{"raw":{"variants":["Phase shift restores estimability of two squeezing parameters","Phase scrambling makes two squeeze strengths jointly measurable","Joint estimation beats stepwise after phase-shift scrambling","Two squeezes one phase: joint estimation becomes feasible"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000559,"raw_usage":{"total_tokens":2685,"prompt_tokens":999,"completion_tokens":1686,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":615,"completion_tokens_details":{"reasoning_tokens":1625}},"tokens_in":615,"tokens_out":1686,"duration_ms":13170,"temperature":1.0,"reasoning_tokens":1625,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:32:55.205606+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"With the paper's own lossless formulas, one can scan $\\alpha\\ge0$ and $\\lambda_1\\ge0$ and compare $C_{\\mathrm{sep}}^{\\min}$ with $C_Q(1+T_I)$; a single point where $C_{\\mathrm{sep}}^{\\min} \\le C_Q(1+T_I)$ would contradict the statement that joint estimation is always better in this model. Alternatively, in a real experiment, measure the joint and stepwise Fisher informations under the optimized phases $\\varphi=\\theta=\\pi/4$ and see whether the joint advantage survives photon losses.","supporting_citations":[{"cited_title":"Albarelli, J","cited_arxiv_id":null,"evidence_quote":"Provides the weight-dependent Holevo bound chain used to bracket the ultimate joint precision."},{"cited_title":"Albarelli, M","cited_arxiv_id":null,"evidence_quote":"Supplies the multiparameter estimation framework and the notion of sloppy models with degenerate Fisher information."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the hyperbolic-phase interpretation of squeezing that motivates the estimated parameters."},{"cited_title":"Olivares, Quantum optics in the phase space, The European Physical Journal Special Topics203, 3 (2012)","cited_arxiv_id":null,"evidence_quote":"Provides the phase-space description used to derive the general-dyne Fisher information."}],"review_version":2}