{"id":"5b105452-9752-4fbd-bbf3-295d6a9461eb","arxiv_id":"2411.11313","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Photonic circuits with photon subtraction and vacuum measurements can asymptotically invert thermal, displacement, and dephasing noise on bosonic codes, enabling error mitigation and suppression without nonlinear elements.","lead":"This paper proposes linear-optical circuits that reduce noise on bosonic quantum states using interferometers, photon subtraction, and measurement postprocessing. The schemes target thermal, random-displacement, and dephasing noise that corrupt continuous-variable qubits, with numerical tests on common bosonic codes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"VMZ infinite-mode result is not proven for arbitrary states: the M→∞ limit is exchanged with a P-function integral without regularity conditions (Sec. 3.3, Eq. 19→20); a direct trace-class/dominated-convergence argument is needed before the 'any dephasing channel' claim can stand.","rationale":"The reader's weakest assumption correctly identifies the P-function limit exchange in Sec. 3.3 as the main proof gap. I agree that Eq. (20) is not rigorously established for arbitrary states as written, because the paper only computes the coherent-state case and then asserts generalization through P-function linearity without proving that the M→∞ limit commutes with the phase-space integral for distribution-valued P functions. This is a genuine completeness-of-proof problem for the central claim. However, I do not regard it as evidence that the result is false: the conditional Kraus operator for vacuum post-selection is exactly s1^{a†a}, which can be verified from Eq. (17) for coherent states and extends by linearity to all trace-class states, so Eq. (18) is valid without P functions. The remaining issue is then a trace-norm interchange, which is likely closable by Chebyshev plus finite-rank approximation. A secondary technical caveat, not flagged by the reader, is that the inversion step requires |λγ| > 0; for complete dephasing with λγ = 0 the residual channel is vacuum projection and linear amplification is undefined. This qualifies the word 'any' but does not affect the main finite-γ suppression results. The novelty/citation issue raised by the reader is real but does not bear on the correctness of the central physics. Overall, the conditional verdict remains appropriate: the authors should add a rigorous trace-class convergence proof (or state explicit regularity conditions) and also note the |λγ| > 0 requirement for the inversion step.","tokens_in":43005,"tokens_out":16811,"duration_ms":177383,"concrete_test":"Supply a direct proof that for every trace-class ρ, ||E_φ[s1^{a†a}ρs1*^{a†a}] − λγ^{a†a}ρλγ*^{a†a}||_1 → 0 as M→∞, using Chebyshev's inequality and finite-rank truncation. As a numerical cross-check, simulate VMZ for a Fock state |N⟩ with N = 10 and M = 2^10 under central-Gaussian dephasing; if the trace distance to the predicted attenuated state does not decrease with M, the universal limit fails, and if it does decrease, the missing dominated-convergence argument should be added to the paper.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The flagship claim is that, for M→∞, VMZ converts an arbitrary dephasing channel into a rotated linear-attenuation channel for any input state (Sec. 3.3, Eq. (20)). The proof computes the post-selected output for coherent states, Eq. (17), then asserts extension to all states by P-function linearity. That extension requires exchanging lim_{M→∞} with the phase-space integral defining ρ from its Glauber–Sudarshan P function. For Fock states, squeezed states, or other states whose P function is a singular distribution, the interchange is not justified by the text; Chebyshev's inequality (19) controls the random variable s1 pointwise, not the operator-valued average in Eq. (18). The universal claim is therefore not established as written. The gap is plausibly repairable: for each phase tuple the vacuum-projection Kraus operator on the input mode is exactly s1^{a†a}, so Eq. (18) holds for arbitrary trace-class states without invoking P functions; finite-rank approximation plus dominated convergence in trace norm would then close the limit. But that argument is absent. A secondary caveat is that invertibility of the residual attenuation requires |λγ| > 0; for dephasing distributions with λγ = 0 (e.g., complete dephasing), Eq. (20) projects onto vacuum and the subsequent rotated linear amplification is undefined.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops two linear-optical noise-handling protocols for bosonic systems. The first combines amplifying and attenuating photon-subtraction gadgets (PSGs) with probabilistic error cancellation (PEC) to mitigate thermal and random-displacement noise, including an optimized constrained estimator for finite sampling. The second, called VMZ, uses a multimode Mach-Zehnder interferometer with conditional vacuum measurements to suppress dephasing noise; the central claim is that in the infinite-ancilla limit the VMZ setup converts any dephasing channel into a rotated linear-attenuation channel for arbitrary input states, which can then be inverted by linear amplification. The paper also presents numerical demonstrations on binomial, cat, squeezed-cat, and GKP codes, including gate-noise scenarios.","tokens_in":43259,"tokens_out":26595,"duration_ms":251704,"significance":"If the central claims hold, the paper provides genuinely useful linear-optical tools for bosonic error mitigation and suppression: the PSG-PEC scheme addresses thermal and random-displacement noise without Kerr nonlinearities, and the VMZ scheme offers a linear-optical method to turn dephasing into an invertible attenuation channel. The analytical work is mostly self-contained: the PSG map identities, the constrained-estimator MSE formulas in Appendix B, and the finite-M fidelity calculations in Section 3.4 are derived carefully and are supported by numerical simulations on multiple bosonic codes. The authors also explicitly disclose the relation to the prior unitary-averaging work of Ref. [141] and state their additional contributions. The main weakness is a missing rigorous justification of the infinite-M limit for nonclassical input states, which is load-bearing for the universality claim of the VMZ protocol.","major_comments":[{"comment":"The passage from the coherent-state result in Eq. (17) to the arbitrary-state statement in Eq. (20) exchanges the limit M → ∞ with an integral over the Glauber-Sudarshan P function. For states with singular P functions, such as Fock states and squeezed states, this interchange is not justified by the text. Chebyshev's inequality (19) controls the random variable s1 pointwise for a fixed phase tuple, but it does not by itself control the operator-valued average in Eq. (18). The claim that VMZ turns 'any dephasing channel' into a rotated linear-attenuation channel for all input states therefore needs a direct convergence argument, for example by writing the action of s1^{a†a} in the Fock basis, using finite-rank approximations, and applying dominated convergence in trace norm. This is likely repairable, but it is currently absent.","section":"Sec. 3.3, Eq. (19) to Eq. (20)"},{"comment":"The statement that the residual channel can be inverted by the rotated linear amplification λ^{-a†a} requires λγ ≠ 0. For dephasing distributions with λγ = 0, such as uniform dephasing over the full circle, Eq. (20) reduces to a projection onto the vacuum (or is undefined when the input has no vacuum component), and the subsequent linear amplification is not defined. The abstract and Section 3.3 should either restrict the universality claim to dephasing distributions with |λγ| > 0 or explicitly describe the vacuum-projection behavior in the λγ = 0 case.","section":"Sec. 3.3, Eq. (20) and following text"}],"minor_comments":[{"comment":"The overline notation for averaging over the phase tuple is introduced after the equation; it would be clearer to define it before Eq. (18) and to specify that the denominator is the corresponding averaged trace.","section":"Sec. 3.3, Eq. (18)"},{"comment":"The rule-of-thumb expression for g in the thermal-noise case is hard to parse as typeset; the intended parentheses around the square-root expression should be made explicit.","section":"Sec. 2.2, Eq. (5)"},{"comment":"The constrained-estimator formulas rely on the Gaussian approximation (B.3) for the multinomial distribution; the validity condition and the role of the large-N assumption would be easier to follow if stated in the main text rather than only in Appendix B.","section":"Sec. 2.4, Eq. (10)"},{"comment":"The claim that Eq. (25) is accurate even for small m and n is presented as a numerical observation; a brief explanation of why the approximation extends beyond the m,n ≫ 1 regime would strengthen the presentation.","section":"Sec. 3.4, Eq. (25) and Fig. 6"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is well within the scope of the journal and the main ideas are valuable. The key required revision is the missing trace-norm/dominated-convergence argument for the infinite-M VMZ limit, plus a qualification for the λγ = 0 case. Neither issue appears fatal, and both are local to Section 3.3, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the paper is a solid, mostly self-contained linear-optics toolbox: PSG-based PEC for thermal and random-displacement noise, and a multimode interferometer (VMZ) that, in the infinite-mode limit, converts dephasing into a rotated attenuation channel that linear amplification can undo. Second, the headline VMZ claim is not fully proven as written: the step from coherent states to arbitrary states via the P function skips a limit-exchange justification when the P function is singular, and the |λγ|=0 case is not handled. The gap looks repairable.\n\nWhat is actually new: PSG-PEC extends Ref. [83]'s loss-only cancellation to thermal and RDN, and the constrained-estimator analysis is careful. VMZ generalizes the unitary-averaging setup of Swain et al. [141] from Gaussian states to arbitrary states, and adds the linear-amplification inversion plus the finite-M Hadamard optimality. The numerical benchmarks on binomial, cat, squeezed-cat, and GKP codes are credible and match the analytics.\n\nThe soft spots, in proportion. The prior-work disclosure is honest but late—Ref. [141] appears only in the Special Note. That should be moved into the main text, since the VMZ setup is the same as theirs. The central gap: Eqs. (19)-(20) use Chebyshev to show s1→λγ in probability, then write the output as an average of s1^{a†a}ρs1*^{a†a} terms and take M→∞. For coherent states this is fine because the integrand is smooth; for a Fock state the P function is a distribution, and the paper gives no argument for exchanging the integral and the limit. The fix is straightforward—the vacuum-projected Kraus operator is exactly s1^{a†a}, so the expression holds for any trace-class ρ, and dominated convergence in trace norm closes the limit—but that argument is absent, so \"any dephasing channel\" is an overclaim as written. Relatedly, when λγ=0 (complete dephasing), the limiting channel projects onto vacuum and the subsequent amplification is undefined; the paper should state |λγ|>0 or handle the boundary separately.\n\nOne smaller point: \"optimal estimators\" means optimal within the chosen clipped-unconstrained family, not globally; the phrase is a little strong but not misleading in context.\n\nWho this is for: people working on CV error mitigation and bosonic codes will get real use from this. The PSG-PEC part is near-ready; the VMZ part needs the proof gap closed before I would fully trust the infinite-mode claim. It deserves a serious referee. I would send it to review and ask for the citation fix and the rigorous limit argument. If those are addressed, the paper is a strong contribution.","headline":"A genuinely useful linear-optics error-mitigation toolbox, but the flagship VMZ infinite-mode claim overreaches its proof by one limit-exchange step.","tokens_in":43849,"tokens_out":3762,"would_cite":true,"duration_ms":38145,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68","81P73","81V80"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that linear-optical components and photon counting, without nonlinear elements, can mitigate thermal and random-displacement noise and fully suppress dephasing noise in bosonic quantum systems.","keywords":["bosonic error mitigation","probabilistic error cancellation","photon subtraction","dephasing suppression","linear optics","continuous-variable quantum computing","Mach-Zehnder interferometer","bosonic codes"],"falsifier":"Simulate the VMZ protocol with a Hadamard interferometer on a Schrödinger-cat input state, propagating the full multimode state exactly in Fock space for increasing M, and compare the conditional output with the prediction $\\lambda_\\gamma^{a^\\dagger a}\\rho\\,\\lambda_\\gamma^{*a^\\dagger a}/\\mathrm{tr}(\\ldots)$. If the output deviates systematically as M grows, the infinite-M 'any state' claim fails; if it converges, the P-function exchange is supported.","tokens_in":42746,"feed_emoji":"","tokens_out":11285,"duration_ms":105190,"temperature":0.7,"pith_summary":"The paper tries to establish that two classes of bosonic noise—thermal/random-displacement noise and pure dephasing—can be fought with linear optics, photon counting, and classical postprocessing. For the first class, photon-subtraction gadgets placed before and after the channel asymptotically invert the noise, and probabilistic error cancellation turns the required unphysical amplification into a feasible sampling protocol with optimized estimators. For pure dephasing, a multimode Mach–Zehnder interferometer with vacuum measurements on ancillas converts the channel into a linear-attenuation channel, which linear amplification can then undo without Kerr nonlinearity. If these claims hold, bosonic codes such as cat, binomial, squeezed-cat, and GKP qubits gain practical error-mitigation and suppression tools that use currently available integrated photonic components.","feed_headline":"Linear optics alone can undo dephasing noise","feed_subtitle":"A multimode interferometer plus photon counting turns any dephasing channel into a loss channel that linear amplification can reverse.","key_machinery":"The two load-bearing objects are the photon-subtraction gadget (PSG) and the vacuum-based Mach–Zehnder (VMZ) scheme. A PSG is a noiseless linear amplifier or attenuator followed by photon subtraction; on coherent states it acts as $|\\alpha\\rangle\\langle\\alpha|\\mapsto|g\\alpha\\rangle\\langle g\\alpha|$, which is why two PSGs of reciprocal gain sandwiching a thermal or displacement channel leave a scaled displacement mixture that vanishes for large gain. The VMZ scheme is an $M$-mode interferometer $U$, i.i.d. dephasing in the middle, the inverse $U^\\dagger$, and vacuum measurements on $M-1$ ancillas; the random sum $s_1=\\sum_j e^{i\\phi_j}|U_{j1}|^2$ acts as the attenuating operator $s_1^{a^\\dagger a}$, and its concentration via Chebyshev's inequality is what converts dephasing into a known linear-attenuation channel in the large-$M$ limit.","core_discovery":"On the paper's own terms, the central discovery is that dephasing noise is not inherently a nonlinear problem. The VMZ scheme—an $M$-mode interferometer $U$, independent identical dephasing on every mode, the inverse interferometer $U^\\dagger$, and vacuum measurements on $M-1$ ancillas—acts on coherent states as the operator $s_1^{a^\\dagger a}$, where $s_1=\\sum_j e^{i\\phi_j}|U_{j1}|^2$ is a random complex number. Chebyshev's inequality shows that for Hadamard or unitary-two-design interferometers $s_1$ concentrates on $\\lambda_\\gamma=\\overline{e^{i\\phi}}$ as $M\\to\\infty$, so the conditional output becomes $\\lambda_\\gamma^{a^\\dagger a}\\rho\\,\\lambda_\\gamma^{*a^\\dagger a}$ with nonvanishing success probability; this is exactly a phase-space-rotated linear-attenuation channel, invertible by rotated linear amplification. The paper also proves that for thermal and random-displacement noise, amplifying and attenuating photon-subtraction gadgets on either side of the channel produce the same output as a scaled random-displacement channel, which becomes the original state as the gain grows.","pith_inferences":["A natural testable extension is to apply the VMZ scheme to inputs with singular phase-space distributions (e.g., cat states) at finite $M$ and check convergence to Eq. (20), since the proof's limit exchange is not demonstrated for such distributions.","The VMZ 'quantum adder' mechanism suggests that the same interferometric setup could be used to estimate dephasing strength from the success probability $p_{\\mathrm{VMZ}}$ or as a general phase-averaging primitive in continuous-variable protocols.","Because PSG and VMZ both use only linear optics and photon counting, integrating them into a single photonic circuit is a plausible engineering step; the dominant cost would be the nondeterministic success of the amplifying PSG, which the paper bounds via a known linear-optical amplification recipe."],"forward_implications":["For pure dephasing, no Kerr nonlinearity is required: in the large-ancilla limit the residual noise is a known linear-attenuation channel, and linear amplification can invert it.","Even with finite ancilla count, weak central-Gaussian dephasing is always suppressed, and a uniform (Hadamard) interferometer gives the best suppression fidelity.","Thermal and Gaussian-displacement noise can be asymptotically removed from expectation values using amplifying and attenuating photon-subtraction gadgets, with optimal physically-constrained estimators minimizing sampling error.","The two schemes commute, so composite channels such as thermal-plus-dephasing or displacement-plus-dephasing can be treated simultaneously, with improved fidelities for cat, binomial, squeezed-cat, and GKP codes.","With appropriate gate modifications, the same mitigation and suppression works for noise from universal gate operations, as shown numerically."],"supporting_citations":[{"why":"It supplies the method of probabilistic error cancellation for undoing photon losses in photonic systems, which the PSG-PEC protocol extends.","marker":"[83]"},{"why":"It provides the linear-optical scheme that saturates the maximum success probability of noiseless linear amplification, used to set the success rates of the PSG and hybrid protocols.","marker":"[106]"},{"why":"It introduces noiseless-loss suppression for multicomponent cat codes with photon subtraction, the baseline that the PSG gadgets generalize.","marker":"[72]"},{"why":"It shows the equivalence between thermal noise and Gaussian-displacement noise under quantum-limited amplification, used to match noise parameters in the numerical demonstrations.","marker":"[25]"},{"why":"It characterizes the bosonic loss-dephasing channel as a degradable composition, motivating the composite thermal-dephasing noise model treated by the hybrid scheme.","marker":"[40]"},{"why":"It gives an optimal decomposition of arbitrary multiport interferometers into beam splitters and phase shifters, used to derive the gate modifications needed when VMZ suppresses gate noise.","marker":"[116]"},{"why":"It supplies the generalized photon-subtraction method for preparing cat states, used in the robustness tests against state-preparation noise.","marker":"[154]"}],"fun_headline_variants":["Photon counting converts dephasing to a loss channel","Linear optics alone reverses dephasing noise","Dephasing becomes loss; linear amplification undoes it","No Kerr needed: dephasing neutralized by linear optics","Interferometer and photon counting eliminate dephasing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that VMZ converts any dephasing channel into a linear-attenuation channel assumes that the infinite-ancilla limit can be moved inside the phase-space integral representing the input state; this exchange is not proven for input states whose phase-space representation is singular, such as highly nonclassical states.","fun_headline_variants_meta":{"raw":{"variants":["Photon counting converts dephasing to a loss channel","Linear optics alone reverses dephasing noise","Dephasing becomes loss; linear amplification undoes it","No Kerr needed: dephasing neutralized by linear optics","Interferometer and photon counting eliminate dephasing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000875,"raw_usage":{"total_tokens":3853,"prompt_tokens":1078,"completion_tokens":2775,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":694,"completion_tokens_details":{"reasoning_tokens":2697}},"tokens_in":694,"tokens_out":2775,"duration_ms":17929,"temperature":1.0,"reasoning_tokens":2697,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T18:39:46.537571+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the VMZ protocol with a Hadamard interferometer on a Schrödinger-cat input state, propagating the full multimode state exactly in Fock space for increasing M, and compare the conditional output with the prediction $\\lambda_\\gamma^{a^\\dagger a}\\rho\\,\\lambda_\\gamma^{*a^\\dagger a}/\\mathrm{tr}(\\ldots)$. If the output deviates systematically as M grows, the infinite-M 'any state' claim fails; if it converges, the P-function exchange is supported.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the method of probabilistic error cancellation for undoing photon losses in photonic systems, which the PSG-PEC protocol extends."},{"cited_title":"Guanzon, Matthew S","cited_arxiv_id":null,"evidence_quote":"It provides the linear-optical scheme that saturates the maximum success probability of noiseless linear amplification, used to set the success rates of the PSG and hybrid protocols."},{"cited_title":"Clements, Peter C","cited_arxiv_id":null,"evidence_quote":"It gives an optimal decomposition of arbitrary multiport interferometers into beam splitters and phase shifters, used to derive the gate modifications needed when VMZ suppresses gate noise."},{"cited_title":"Generation of optical Schrödinger cat states by generalized photon subtraction","cited_arxiv_id":null,"evidence_quote":"It supplies the generalized photon-subtraction method for preparing cat states, used in the robustness tests against state-preparation noise."}],"review_version":1}