{"id":"a07e03b1-0192-4a3a-8054-be4fb1b11b94","arxiv_id":"2505.10148","paper_version":2,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Applying a loss-tolerant GHZ distribution protocol to four-node quantum network sensing yields a quadratic improvement in loss scaling of estimation error (η^-2 vs η^-4), but the promised privacy result is absent from the manuscript.","lead":"The paper analyzes a quantum sensing protocol that distributes entanglement over lossy fiber networks more efficiently and claims lower estimation error for distributed phase parameters. The abstract also promises a privacy-preserving version, but the manuscript body contains no privacy analysis at all.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Figs. 5 and 6 fix N=30 successful events for both protocols instead of accounting for per-attempt success probabilities, so the claimed high-loss variance advantage for multi-parameter sensing may be an artifact of unequal resource accounting.","rationale":"The reader's weakest_assumption, Eq. (6), is not the most load-bearing issue for the stated variance advantage. Even if the diagonal-noise decomposition in Eq. (6) is accepted, the Fisher information per successfully generated copy is bounded by 16p, which is below the direct-transmission value 16. Therefore the only way the proposed scheme can show lower variance is through a much larger effective number of successful copies, i.e., through the eta^(M/2) versus eta^M success-probability scaling. That resource advantage must be counted consistently on both sides. The fixed-N=30 convention in Figs. 5 and 6 discards exactly this factor, so the comparison is not a valid test of the claim. A concrete recomputation with N_eff = N_attempts * P_suc will settle it. I agree with the reader's overall REJECT verdict, but I locate the load-bearing flaw in the resource accounting rather than in Eq. (6). The missing privacy analysis is also a serious independent gap: the abstract promises a private protocol and a hiding parameter, but the body contains no adversary model, privacy definition, or proof. That alone justifies rejection of the paper as submitted, but it is not the technical step on which the variance claim depends.","tokens_in":11718,"tokens_out":9324,"duration_ms":100784,"concrete_test":"Recompute Figs. 5 and 6 under a fixed total attempt count T for both arms. For the proposed protocol, set the number of retained copies of pattern j to N_j = T * P_CS * Pr(pattern j | success), using Eq. (A1) and Table I; for direct transmission set N_j = T * eta^4 for each of the three states. Insert these N_j into the variance expression in Eq. (A4) and plot Delta-theta^2 versus eta. If the proposed curve is not below the direct curve throughout the stated high-loss regime (e.g., eta < 0.1), the central variance-advantage claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The decisive point is the resource normalization in the multi-parameter comparison. For the single-parameter case the paper correctly uses N' = N P_suc (Eq. 31), with P_suc = eta^4 for direct transmission and P_CS from Eq. (A1) for the proposed distribution protocol. For arbitrary linear combinations, however, Figs. 5 and 6 assume that each detection pattern in Table II is obtained exactly 30 times, and that direct transmission sends the corresponding three GHZ states 30 times. No equivalent P_suc factor is applied to either arm. This removes the central resource difference from the comparison. With equal numbers of successful copies, the pure GHZ state used by direct transmission has QFI 16, whereas the heralded state has QFI at most 16p with p < 1 (Eq. 29), so the proposed protocol cannot beat direct transmission on a per-success basis; the plotted advantage must arise from the hidden unequal normalization. The claim of lower variance in the high-loss regime is therefore not established unless both arms are compared under the same total-attempt budget.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript analyzes a four-node quantum network sensing protocol based on the loss-tolerant GHZ-state distribution scheme of Ref. [32]. In the proposed scheme, each sensing station prepares a two-photon state and sends one photon to a central node; upon a heralded detection pattern, the stations share an approximate four-qubit GHZ state. The authors compute the quantum and classical Fisher information for estimating phase combinations such as θ=(θ1−θ2+θ3−θ4)/4, and for arbitrary linear combinations of three phases using three different GHZ patterns. They compare variance lower bounds with a direct-transmission reference in which a pure GHZ state is distributed through lossy fibers, using N′=N×P_suc (Eq. 31) to include the distribution success probability. They report lower estimation variance than direct transmission under high loss, and the abstract promises a privacy feature in which an additional parameter hides the target parameter from all users except the one who controls it. Appendices provide the success-probability formula (imported from Ref. [32]) and the variance propagation for linear combinations.","tokens_in":11727,"tokens_out":10781,"duration_ms":104136,"significance":"If the central claims were established, the result would be practically relevant: a quadratic improvement in distribution success probability (η^{M/2} vs η^M) could partially compensate for loss-induced degradation of multipartite entanglement in quantum network sensing, and the use of an implementable displacement-plus-photon-counting measurement is a useful feature. The paper correctly handles the single-parameter comparison with attempt-based resource accounting (Eq. 31), correctly uses convexity of the quantum Fisher information to bound the noisy-state QFI by 16p, and the direct-transmission QFI/CFI calculation (Eqs. 22–26) checks out. Building on a concrete distribution protocol with an explicit success probability is a strength. However, the multiparameter comparison in Figs. 5–6 does not follow the same resource accounting, and the advertised privacy result is absent from the body of the manuscript. These gaps currently prevent the main claims from being accepted.","major_comments":[{"comment":"The single-parameter comparison correctly uses N′=N×P_suc (Eq. 31), but the multiparameter comparison in Figs. 5 and 6 instead assigns 30 successful copies to each detection pattern for both protocols. This removes the success-probability difference—the core resource advantage of the proposed scheme—from the comparison. Because the direct-transmission state is pure with QFI 16 per copy, while the proposed heralded state has QFI at most 16p with p<1 (Eq. 29), a per-success comparison cannot produce the plotted advantage; the advantage in the figures must come from an implicit unequal total-attempt budget. The claim of lower variance in the high-loss regime for arbitrary linear combinations is therefore not established as presented. Please redo Figs. 5–6 under a common attempt budget, for example N_i′=N_attempts×P_suc for each pattern and N′=N_attempts×η^4 for direct transmission.","section":"Section III, Figs. 5–6 and Eq. (31)"},{"comment":"The abstract and title advertise private quantum network sensing and state that 'an additional parameter can be used to hide the information about the target parameter from everyone, except the person who controls the parameter.' The manuscript body contains no adversary model, no definition of privacy, and no analysis of this hiding mechanism. This is a central advertised contribution, not a side remark. Either add the missing analysis or remove the privacy claims from the abstract and title and adjust the framing accordingly.","section":"Abstract and title; no corresponding analysis"},{"comment":"The decomposition ρ_CS=p|GHZ⟩⟨GHZ|_4+Σ_i r_i|ψ_i⟩⟨ψ_i|, with all noise terms declared diagonal in the photon-number basis and hence θ-independent, is asserted without derivation. This decomposition is the premise for the bound F_Q≤16p (Eq. 29) and for the statement that the noise terms carry no phase information. Please provide the explicit form of the states |ψ_i⟩ and weights r_i (for example from the loss model in Appendix A.1) and show explicitly that they are diagonal under the phase-shift unitaries; if the derivation is already in Ref. [32], state the correspondence precisely and state whether P_CS in Eq. (A1) refers to a single detection pattern or to the union of the two equivalent patterns in Table I.","section":"Section II.A, Eqs. (6)–(7)"},{"comment":"The measurement uses |α⟩⟨α| with the coherent state approximated by e^{−|α|²/2}(|0⟩+α|1⟩). For the chosen α=1/√2 the |2⟩ component is not negligible: about 9% of the population lies in Fock states |n≥2⟩. The reported CCRB values therefore depend on this truncation. Please justify the approximation quantitatively or evaluate the classical Fisher information with the full coherent-state POVM.","section":"Section II.B, Eq. (14) and Figs. 4–6"}],"minor_comments":[{"comment":"The right-hand side should use the quantum Fisher information matrix of the pure GHZ state, not of ρ_CS,θ; as written the inequality would reduce to F_Q(ρ_CS)≤p F_Q(ρ_CS), which is false for p<1.","section":"Eq. (29)"},{"comment":"The notation N is used for the number of stations in Eq. (A1) while the main text uses M=4. In addition, 'η=0.2 [db/km]' is an attenuation coefficient with units dB/km, not a transmittance; the distance dependence of the transmittance should be stated explicitly.","section":"Eq. (A1) and Section III"},{"comment":"Fig. 5(b) is captioned with the parameter (θ1+θ2+2θ3)/4, which is the parameter of Fig. 6, while the text and Fig. 5(a) refer to the average (θ1+θ2+θ3)/3. Please align the caption with the text.","section":"Fig. 5 caption"},{"comment":"The statement that α=1/√2 corresponds to the minimum CCRB is not supported by a derivation or an optimization plot; please provide the calculation or a reference.","section":"Section III, paragraph before Eq. (31)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on the authors' own Ref. [32] for the distribution protocol and success-probability formula, which is imported without derivation; that is acceptable if the correspondence is made precise, but the missing privacy analysis is a scope concern. The resource-accounting error in Figs. 5–6 is the main technical obstacle: it invalidates the multiparameter comparison as presented, although a corrected comparison under a common attempt budget may still show an advantage for the proposed protocol in the high-loss regime. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my take on arXiv:2505.10148. The single-parameter analysis is the real meat and it mostly holds up: conditioning on two-photon detection gives success probability ~η² vs η⁴, and with N' = N P_suc the loss-induced variance advantage is plausible. The construction for estimating arbitrary linear combinations by combining different GHZ detection patterns is a genuinely useful addition. The authors also deserve credit for stating that they idealize direct transmission, which is the right way to search for worst-case advantages.\n\nBut two serious problems keep me from endorsing the paper as is. First, the multi-parameter comparison in Figs. 5 and 6 does not use the same resource accounting as the single-parameter case. The text fixes N=30 successful copies for each detection pattern, and appears to give direct transmission the same number of successful GHZ states, with no P_suc factor on either arm. That removes the core η² vs η⁴ difference from the comparison. On a per-success basis, direct transmission's pure GHZ state has QFI 16, while the heralded state has at most 16p with p<1, so the plotted high-loss advantage cannot come from the protocol's loss tolerance. Unless both arms are compared under the same total-attempt budget, the paper's multi-parameter claim is not established. The stress-test note lands.\n\nSecond, the abstract promises privacy: 'preventing sensitive information from leaking' and 'an additional parameter can be used to hide the information.' The manuscript body contains no privacy analysis at all. That is a mismatch between the advertised contribution and the delivered content. Either add the analysis or remove the claim from the abstract.\n\nMinor points: the classical Fisher information for the displacement measurement is given only graphically, and the coherent-state truncation at α=1/√2 leaves ~7.6% in the |2> component. Both are fixable with a derivation and a numerical check. The state decomposition in Eq. (6) is imported from Ref [32] without derivation here; that is acceptable if the reference truly establishes it, but the paper should point to the exact argument.\n\nWho this is for: people working on distributed quantum sensing with realistic channel loss. The single-parameter result is worth a careful read. The multi-parameter and privacy parts need substantial revision before the claimed scope is justified.\n\nMy recommendation: send it to peer review, but with the expectation of major revision. The core single-parameter idea and the linear-combination construction deserve referee time, but Figs. 5-6 and the abstract need to change. I would not cite the multi-parameter claim as it stands.","headline":"A genuinely interesting single-parameter result is undercut by flawed multi-parameter resource accounting and an unfulfilled privacy promise.","tokens_in":12443,"tokens_out":3423,"would_cite":false,"duration_ms":32859,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.-a","03.65.Ta","42.50.-p"],"model":"deepseek-v4-flash","headline":"The paper claims that a loss-tolerant GHZ distribution scheme lets a four-node quantum sensor network estimate linear combinations of distributed phases with lower variance than direct transmission in the high-loss regime.","keywords":["quantum network sensing","distributed quantum sensing","GHZ states","loss-tolerant entanglement distribution","quantum Fisher information","Cramér–Rao bound","photonic phase estimation","star network"],"falsifier":"Tomographically reconstruct the heralded state $\\rho_{CS}$ after distribution through a lossy channel, then scan the phase $\\theta$ and check whether the non-GHZ part $\\sum_i r_i|\\psi_i\\rangle\\langle\\psi_i|$ changes with $\\theta$; any such phase dependence would falsify the diagonal-noise decomposition on which the bound $F_Q[\\rho_{CS},\\theta]\\le 16p$ depends.","tokens_in":11332,"feed_emoji":"🔗","tokens_out":12581,"duration_ms":122451,"temperature":0.7,"pith_summary":"This paper tries to establish that a loss-tolerant way of distributing GHZ states makes quantum network sensing work over lossy optical fibers where the standard approach fails. In a four-station star network, the proposed distribution has success probability scaling $\\eta^2$ instead of $\\eta^4$, and for the global phase $\\theta=\\frac{1}{4}(\\theta_1-\\theta_2+\\theta_3-\\theta_4)$ the estimation variance stays below the direct-transmission benchmark at high loss. By combining three different heralded GHZ states and using one mode as a reference, the same protocol estimates arbitrary linear combinations of the other three phases. The paper also claims that a user controlling a further reference phase can hide the target parameter from everyone else, giving the protocol a private-sensing capability.","feed_headline":"Loss-tolerant GHZ distribution lowers quantum sensing error","feed_subtitle":"A four-node network shares GHZ states twice as loss-tolerant, shrinking phase-estimation error at long range.","key_machinery":"The load-bearing object is the loss-tolerant GHZ distribution protocol of Ref. [32]: each station prepares a two-photon state and sends one photon to a central node, where a beam-splitter network and single-photon detections herald, on certain two-click patterns, a GHZ state (up to local bit flips) in the retained modes. Its analysis rests on the decomposition $\\rho_{CS}=p|GHZ\\rangle\\langle GHZ|+\\sum_i r_i|\\psi_i\\rangle\\langle\\psi_i|$, with all $|\\psi_i\\rangle\\langle\\psi_i|$ diagonal in the photon-number basis, so they contain no information about $\\theta$ and the quantum Fisher information is bounded by $16p$. The classical Fisher information is computed for the local measurement defined by the binary projectors $|\\alpha\\rangle\\langle\\alpha|$ and $I-|\\alpha\\rangle\\langle\\alpha|$, with $\\alpha=1/\\sqrt{2}$, which is the feasible substitute for the ideal $\\sigma_x$ measurement. The ability to estimate arbitrary linear combinations comes from the pattern table: each two-detector pattern produces a different GHZ state and therefore a different linear phase functional, and the estimators are combined through Eq. (30) with variances adding as in Appendix B.","core_discovery":"The central claim is that the loss penalty for multipartite entanglement need not scale as the worst case of every photon arriving. The paper uses a protocol in which each sensing station keeps one half of a two-mode state $a|00\\rangle+b|11\\rangle$ and sends the other half to a central station, where interference and photon-number detection herald a GHZ state among the four retained modes whenever a two-detector pattern clicks. For the target $\\theta=\\frac{1}{4}(\\theta_1-\\theta_2+\\theta_3-\\theta_4)$, the heralded state is written as $\\rho_{CS}=p|GHZ\\rangle\\langle GHZ|+\\sum_i r_i|\\psi_i\\rangle\\langle\\psi_i|$ whose noise terms are diagonal and phase-independent, and convexity of the quantum Fisher information gives $F_Q[\\rho_{CS},\\theta]\\le 16p$. The paper computes both the quantum and classical Cramér–Rao bounds for this state, with the classical bound evaluated for a feasible measurement made of a displacement followed by photon counting, and finds that the protocol beats the idealized direct-transmission benchmark whenever loss is high. Three different detection patterns yield three GHZ states with different phase functionals, allowing arbitrary linear combinations of three phases to be reconstructed against a reference mode.","pith_inferences":["The abstract's privacy claim is not developed in the main text; a natural extension is to compute the Fisher information about the target parameter available to non-privileged stations and to verify whether the extra reference phase truly suppresses it.","The variance diverges at some phase values, so an adaptive protocol that changes the operating point or merges estimates from several detection patterns could smooth these singularities; the paper leaves that open.","The $\\eta^{M/2}$ scaling invites a direct calculation of the distance at which the crossover to lower variance occurs as a function of loss per kilometer, and of how that crossover moves as $M$ grows; the formulas in Appendix A are enough to generate such a map."],"forward_implications":["For a four-station network with equal-length fibers, the estimation variance of $\\theta=\\frac{1}{4}(\\theta_1-\\theta_2+\\theta_3-\\theta_4)$ becomes smaller than in direct transmission once the fiber loss is sufficiently high, with the crossover visible in the paper's plots around tens of decibels.","Because the distribution success probability scales as $\\eta^{M/2}$ rather than $\\eta^M$ for even $M$, adding sensing stations costs less in lost transmission efficiency than it does under direct distribution.","With one mode used as a reference, the three GHZ states from the three two-click patterns yield three independent phase estimates, and any linear combination of the remaining three phases can be reconstructed with variance equal to the weighted sum of the component variances.","The feasible displacement-and-photon-counting measurement does not reach the quantum Cramér–Rao bound for the noisy heralded state, so the protocol has a known, quantified gap that an improved local measurement could close.","The direct-transmission comparison assumes a deterministic pure GHZ source; including generation noise and failure in the benchmark would only make the proposed protocol look better."],"supporting_citations":[{"why":"It supplies the loss-tolerant GHZ distribution protocol whose $\\eta^{M/2}$ success scaling is the basis of the claimed advantage.","marker":"[32]"},{"why":"It supplies the quantum and classical Cramér–Rao bounds and the Fisher information formalism used to compare estimation variances.","marker":"[4]"},{"why":"It identifies the $\\sigma_x$ measurement as optimal for ideal GHZ phase estimation, which is the benchmark against which the feasible measurement is compared.","marker":"[26]"},{"why":"It provides the displacement-plus-photon-counting implementation of binary projective measurements used for the local measurement.","marker":"[34]"},{"why":"It provides the linear-optics realization of binary projective measurements that supports the classical Fisher information calculation.","marker":"[35]"},{"why":"It supplies the quantum Fisher information matrix formula used to obtain $F_Q=16$ for the ideal GHZ state.","marker":"[40]"}],"fun_headline_variants":["Loss-tolerant GHZ scheme reduces quantum sensing error","Private quantum sensing improved by loss-tolerant GHZ","Efficient GHZ distribution enables private network sensing","Loss-tolerant GHZ shrinks error in private network sensing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole advantage rests on the assumption that the noisy part of the heralded state carries no phase information—that every $|\\psi_i\\rangle\\langle\\psi_i|$ term stays diagonal in the photon-number basis—so only the ideal GHZ component contributes to the Fisher information; if loss creates phase-sensitive coherences in the noise, the bound $F_Q[\\rho_{CS},\\theta]\\le 16p$ and the claimed advantage can fail.","fun_headline_variants_meta":{"raw":{"variants":["Loss-tolerant GHZ scheme reduces quantum sensing error","Private quantum sensing improved by loss-tolerant GHZ","Efficient GHZ distribution enables private network sensing","Loss-tolerant GHZ shrinks error in private network sensing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000344,"raw_usage":{"total_tokens":1910,"prompt_tokens":987,"completion_tokens":923,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":862}},"tokens_in":603,"tokens_out":923,"duration_ms":8648,"temperature":1.0,"reasoning_tokens":862,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:16:16.293300+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Tomographically reconstruct the heralded state $\\rho_{CS}$ after distribution through a lossy channel, then scan the phase $\\theta$ and check whether the non-GHZ part $\\sum_i r_i|\\psi_i\\rangle\\langle\\psi_i|$ changes with $\\theta$; any such phase dependence would falsify the diagonal-noise decomposition on which the bound $F_Q[\\rho_{CS},\\theta]\\le 16p$ depends.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the loss-tolerant GHZ distribution protocol whose $\\eta^{M/2}$ success scaling is the basis of the claimed advantage."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It identifies the $\\sigma_x$ measurement as optimal for ideal GHZ phase estimation, which is the benchmark against which the feasible measurement is compared."},{"cited_title":"Sekatski, S","cited_arxiv_id":null,"evidence_quote":"It provides the displacement-plus-photon-counting implementation of binary projective measurements used for the local measurement."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the quantum Fisher information matrix formula used to obtain $F_Q=16$ for the ideal GHZ state."}],"review_version":1}