{"id":"ce65ee1f-d069-4a82-859c-6b6cde773c21","arxiv_id":"2505.20382","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The quantum state conditioned on time-averaged measurement records is updated exactly by a Gaussian-averaged tilted-Liouvillian superoperator, which can be expanded to arbitrary order in sqrt(Delta t).","lead":"Continuous quantum measurements are recorded as time-averaged digital bins, not as ideal continuous signals. This paper derives the exact Bayesian quantum-state update for such digitized records, valid at any bin duration, and shows how to compute it numerically or with high-order perturbative expansions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the tilted-Liouvillian derivation is internally consistent and the recursive exact map follows by linearity.","rationale":"The reader's verdict is ACCEPT with high confidence, and I concur. The paper's central identity is a Bayesian filtering result: the robinet state is the conditional expectation of the state given coarse-grained integrated measurement outcomes. The derivation via the tilted Liouvillian is standard and I found no algebraic or stochastic-calculus error. The recurrence is linear in the prior state, which is what makes the filter closed on the conditional mean; this is the crucial property and it holds. The reader's weakest assumption about the ideal digitized signal is a genuine scope limitation but not a correctness flaw, since the paper explicitly defines the digitized record as the time integral of the SME output. The only notable gap is the unproved complete positivity of K_I, which the text asserts without demonstration; however, the map is constructed from a positive post-selection average and the numerical checks are consistent, so this is a minor presentation issue rather than a load-bearing objection. I therefore recommend no change to the reader's verdict.","tokens_in":10280,"tokens_out":24282,"duration_ms":282763,"concrete_test":"Compute the Choi matrix of K_I from Eq. (14) via Gauss-Hermite quadrature for a two-qubit system with noncommuting L and C over a range of Δt values, and check for negative eigenvalues; if positivity fails for any physical parameter set, the completely-positive-instrument claim would need revision, although the conditional-expectation recurrence might still hold.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I independently checked the central derivation. The Itô computation leading to Eq. (9) is consistent: the ordinary exponential of ∫ j dY picks up the 1/2 j² dt correction, and the remaining dW term has zero expectation, so the tilted generator L_j = L + jC + j²/2 I is correct. The factorization in Eq. (10) is valid because the p²/2 term is a scalar superoperator that commutes with L - ipC. The recurrence in Eqs. (12)-(13) is closed because K_I is linear in the prior state, so the conditional mean is a sufficient statistic for the linear Bayesian update. I find no internal inconsistency that would invalidate the central claim. The only real limitation is the one the reader identified: the map is exact for the ideal digitized signal I_k = ∫ dY_t of the SME, and finite detector bandwidth, filtering, dead time, or non-Markovian noise would break exactness. That is a scope condition of the model, not an error in the argument. The paper asserts without proof that K_I is completely positive; this is a presentation gap, but the physical post-selection construction and the numerical tests make the assertion plausible and it does not undermine the conditional-expectation identity.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper addresses the problem of reconstructing the state of a continuously measured quantum system when only a time-averaged (digitized) measurement record is available, which is the situation in most experiments. The authors define the 'robinet state' as the Bayesian conditional expectation of the true state given the digitized record and derive an exact recursive update map K_I (Eqs. (13)-(14)) expressed as a Fourier integral of tilted-Liouvillian exponentials. They also derive a perturbative expansion in powers of sqrt(Δt), recovering and extending existing discretization schemes, and provide numerical verification via Monte Carlo post-selection. Applications to state reconstruction, parameter estimation, and trajectory sampling are presented.","tokens_in":10484,"tokens_out":13767,"duration_ms":154528,"significance":"The central result is a clean, exact formula for the quantum instrument corresponding to time-averaged continuous measurement, valid for arbitrary bin duration. This is an important contribution because it removes the need for the extremely fine binning currently required for unbiased state reconstruction and parameter estimation in experiments with superconducting circuits and other platforms. The derivation via Itô calculus is rigorous and the recursion is simple to implement. The perturbative expansion provides a systematic way to generate higher-order discretization schemes, and the authors demonstrate clear advantages over existing methods. The paper is well written and the numerical checks, while limited, support the claims.","major_comments":[{"comment":"The map K_I is asserted to be completely positive (CP) without proof. The integral representation involves non-Hermitian generators e^{Δt(L - ipC)}, so CP is not manifest. Since K_I is called an 'exact quantum instrument' and is used to generate normalized states, please provide a proof of CP (e.g., via a Kraus representation or by showing the integral is a superposition of CP maps) or cite a reference where this is established.","section":"Exact map, Eq. (14)"},{"comment":"The Monte Carlo post-selection verification for a single point uses only 10 trajectories, giving a statistical uncertainty on the order of 30%. This is not a strong quantitative test of the formula. Please provide error bars or use a more efficient estimator (e.g., importance sampling) to demonstrate convergence as epsilon tends to zero.","section":"Verifying the formula, Figure 1"}],"minor_comments":[{"comment":"The figure caption states that 10 trajectories out of 10^6 are post-selected, and the text says that averaging these 10 states approximates the robinet state; the statistical significance of this comparison should be quantified.","section":"Verifying the formula, Figure 1"},{"comment":"The truncated polynomial-times-Gaussian density used for sampling may take negative values for some values of \\bar{I}; please discuss how the authors ensure a valid proposal density in the rejection sampling procedure.","section":"Evaluating the formula, Eq. (21)"},{"comment":"The symbol I is used both for the measurement signal and for the identity superoperator; please use a distinct symbol (e.g., \\mathbb{I}) for the identity to avoid confusion.","section":"Throughout"},{"comment":"The statement 'for brevity, we assume the Liouvillian is time-independent' is terse; it would be helpful to indicate explicitly how the formula generalizes to time-dependent L (via time-ordered exponentials) in the main text or appendix.","section":"Derivation, Eq. (9)"},{"comment":"The claim that the recursion cost is quadratic in the chosen order is not fully explained; please add a sentence on how the terms are recursively computed.","section":"Appendix, Perturbative expansion"}],"recommendation":"minor_revision","confidential_remarks":"The paper is within the scope of the journal and the central result is correct. The derivation is rigorous, but the missing proof of complete positivity and the weak post-selection verification should be addressed. No issues with novelty or citation pattern were found."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper gives the exact recursive map for reconstructing a quantum state from time-averaged measurement records, which is the right object for experiments with finite digitization. The derivation is straightforward once you see it, but it's new and useful: the tilted-Liouvillian trick is standard, but the recursive construction and the explicit quadrature/perturbative evaluation are solid. The Itô computation is clean, the factorization is valid, and the numerical post-selection verification really checks the conditional average independently of the derivation. The applications show clear advantages over Euler and CPTP-1 schemes, and the perturbative expansion correctly reproduces known low-order terms while extending them to arbitrary order.\n\nThe soft spots are real but minor. The exactness is conditional on the digitized signal being the perfect time integral of the ideal diffusive SME output. Finite detector bandwidth, filtering, dead time, or non-Markovian noise break the map. That is a scope condition, not an error, but it should be stated more prominently than it is. Complete positivity of K_I is asserted rather than proved; it looks plausible from the physical construction, but a proof or at least a more careful statement would tighten the paper. The time-dependent generalization is mentioned in the appendix but not worked out; that is fine for a letter. No code or data are shipped, which is a minor reproducibility gap given the numerical claims.\n\nOverall the central argument holds up. The paper is clearly written, the math is internally consistent, and the authors are honest about what the robinet state is and is not: it is the best estimate given the information actually available, not a reconstruction of the continuous trajectory. I found no load-bearing flaw.\n\nWho is it for? Anyone working with continuous measurement in superconducting circuits or quantum optics where digitization time matters. It deserves a serious referee. I would recommend acceptance with minor revisions: add a short discussion of detector non-idealities, and either prove or explicitly defer the complete-positivity claim. I would cite this in future work on quantum filtering.","headline":"A clean, self-contained derivation of an exact Bayesian filter for digitized continuous measurement records; worth engaging despite its narrow ideal-signal scope.","tokens_in":11001,"tokens_out":1452,"would_cite":true,"duration_ms":18476,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.65.Ta","03.65.Yz"],"model":"deepseek-v4-flash","headline":"This paper derives an exact map $K_I(\\rho)$ that produces the best Bayesian estimate of a quantum state from a binned, time-averaged measurement record for any bin duration $\\Delta t$, and shows it can be evaluated exactly or expanded to…","keywords":["continuous quantum measurement","stochastic master equation","Bayesian state estimation","state reconstruction","parameter estimation","tilted Liouvillian","quantum trajectory"],"falsifier":"Run a homodyne or heterodyne experiment on a system with known Hamiltonian and monitoring strength, digitize at a coarse $\\Delta t$ where Euler schemes fail, histogram many runs' bin integrals, and compare with $\\operatorname{Tr}[K_{I}(\\rho_0)]$ computed by Gauss-Hermite quadrature; any discrepancy beyond sampling error refutes exactness. A sharper version inserts a low-pass filter before digitization: the paper's premise says $K_I$ must change, so the unmodified map should visibly mispredict the histogram.","tokens_in":10071,"feed_emoji":"⚛️","tokens_out":14545,"duration_ms":127049,"temperature":0.7,"pith_summary":"Experiments record continuous quantum measurements only as digitized bins: averages of the instantaneous signal over intervals of duration $\\Delta t$. The paper's central claim is that the best Bayesian estimate of the quantum state given only these bins—not the unobservable continuous trajectory—can be obtained by an exact, completely positive map that iterates in time and is valid for any bin duration, however large. This matters because standard practice treats the binned signal as though it were the continuous one, which is only approximately correct for very small $\\Delta t$ and can produce unphysical states or biased parameter estimates. The paper shows the exact map can be evaluated numerically (Gauss-Hermite quadrature) or expanded in powers of $\\sqrt{\\Delta t}$ to arbitrary order, recovering and extending existing low-order discretization schemes. If correct, it removes the numerical-stability constraint on digitization, allowing coarser data storage and unbiased inference at the physically relevant timescale.","feed_headline":"Exact formula fixes quantum state reconstruction from coarse data","feed_subtitle":"No more tiny bins: an exact Bayesian filter reconstructs the state and removes bias at any bin duration.","key_machinery":"The load-bearing object is the tilted Liouvillian $\\mathcal{L}_j = \\mathcal{L} + j\\mathcal{C} + \\frac{j^2}{2}\\mathcal{I}$, a deformation of the measurement generator that, when exponentiated over a bin, propagates the state weighted by the Fourier phase of that bin's integrated signal. Its time-ordered exponential over $n$ bins factors into a product of single-bin superoperators $\\Phi_p$, and because each $\\Phi_p$ is linear, the Fourier integrals from the Dirac deltas collapse into a recursion on the previous linear robinet state. The completely positive map $K_I$ is the quantum instrument for a bin: it produces both the unnormalized conditional state and, by its trace, the exact probability density of the binned signal. The same tilted-generator trick supplies the jump-measurement version and the multi-channel version, and the Gaussian integration against the phase $e^{ipI}$ gives the Hermite-polynomial perturbative expansion in $\\sqrt{\\Delta t}$.","core_discovery":"The construction is built on the linear robinet state: an unnormalized state carrying the conditional expectation of $\\rho_{n\\Delta t}$ weighted by a product of Dirac deltas that enforce the measured bin integrals. By writing each delta as a Fourier integral, the expectation becomes a Gaussian-weighted Fourier integral of a state propagated by the tilted Liouvillian $\\mathcal{L}_j = \\mathcal{L} + j\\mathcal{C} + (j^2/2)\\mathcal{I}$ with piecewise-constant imaginary $j$ on each bin. Linearity of the resulting superoperator $\\Phi_p = e^{\\Delta t \\mathcal{L}_{-ip}}$ makes the $n$-bin expression factorize, yielding the one-bin recurrence $\\rho_k = K_{I_k}(\\rho_{k-1})/\\operatorname{Tr}[K_{I_k}(\\rho_{k-1})]$ with $$K_I(\\rho) = \\frac{1}{2\\pi}\\int_{\\mathbb{R}} dp\\, e^{ipI - \\$\\Delta$ t $p^{2}$/2} e^{\\$\\Delta$ t(\\mathcal{L} - ip\\mathcal{C})}(\\rho).$$ The normalization is the exact probability density of the binned signal $I_k$. The paper claims this is the exact Bayesian conditional average $E[\\rho_{n\\Delta t}\\mid I_1,\\dots,I_n]$ for any $\\Delta t$, that it extends to jump measurements (with a bounded $p$-integral) and multiple monitored operators, and that its perturbative expansion reproduces the Milstein correction and known higher-order completely positive discretizations while extending them to arbitrary order; for unit efficiency, up to fifth order in $\\sqrt{\\Delta t}$ the map is a single Kraus operator.","pith_inferences":["Editorial inference: the exact map gives a natural generative model of coarse-grained trajectories; sampling bins from the exact signal density and updating produces statistics identical to the full stochastic master equation, which could be used for fast, memory-light simulation of long experiments.","Editorial inference: existing experimental datasets recorded at coarse $\\Delta t$ could be re-analyzed with the exact map without re-running the experiment, potentially correcting state estimates and parameter biases in previously published results.","Editorial inference: the derivation's reliance on the ideal time-average suggests a concrete stress test—insert a calibrated filter before digitization and check whether a modified, filter-dependent map is needed; where the ideal premise fails, the method would need to be re-derived for realistic detector responses.","Editorial inference: the same Fourier/tilted-generator construction should apply to time-dependent Liouvillians and to nonlinear or colored-noise measurement records, though the paper explicitly treats only constant Liouvillians and ideal averaging."],"forward_implications":["State reconstruction from a digitized record at coarse $\\Delta t$ yields the true best Bayesian estimate, where Euler and CPTP-1 schemes can produce unphysical states and fidelities lower than even the unobserved Lindblad average.","Maximum-likelihood parameter estimation built from the exact map is unbiased at fixed $\\Delta t$, so binning can be chosen at the information-relevant timescale instead of the numerical-stability timescale.","Digitized quantum trajectories can be sampled directly at finite $\\Delta t$ by drawing $I_k$ from $\\operatorname{Tr}[K_{I_k}(\\rho_{k-1})]$ and applying the update, with the perturbative expansion giving schemes of arbitrarily high order.","The perturbative expansion places existing discretization schemes, including the Milstein correction, on a common Bayesian footing and extends them to all orders in $\\sqrt{\\Delta t}$.","For perfect detection efficiency, purity loss from time averaging appears only at third order in $\\Delta t$; up to fifth order in $\\sqrt{\\Delta t}$ the map remains a single Kraus operator."],"supporting_citations":[{"why":"Defines the diffusive stochastic master equation and measurement record that the exact map is conditioned on.","marker":"[3]"},{"why":"Supplies the Itô-lemma step and the jump tilted Liouvillian used in the derivation and generalization.","marker":"[4]"},{"why":"Provides the tutorial jump-measurement stochastic master equation whose tilted generator appears in the jump generalization.","marker":"[5]"},{"why":"Introduced the Milstein correction that the perturbative expansion reproduces at second order.","marker":"[6]"},{"why":"Derived completely positive quantum trajectory terms recovered by the third- and fourth-order expansion.","marker":"[7]"},{"why":"Derived higher-order completely positive trace-preserving unravelings that the expansion reproduces and extends.","marker":"[8]"},{"why":"Supplies the engineered-dissipation superconducting-cavity experiment used to benchmark state reconstruction.","marker":"[14]"},{"why":"Defines the CPTP-1 reconstruction scheme compared against in the state-reconstruction benchmark.","marker":"[15]"}],"fun_headline_variants":["Exact formula for coarse-bin quantum measurement","Quantum state from binned records: exact solution","No bias in quantum trajectory reconstruction","Exact Bayesian quantum state from digitized data"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction assumes the recorded bin value is exactly the time integral of the ideal continuous measurement output obeying the stochastic master equation, with no detector bandwidth limit, filtering, dead time, or non-Markovian noise; if the physical record deviates from this perfect average, the map $K_I$ is not the true Bayesian update.","fun_headline_variants_meta":{"raw":{"variants":["Exact formula for coarse-bin quantum measurement","Quantum state from binned records: exact solution","No bias in quantum trajectory reconstruction","Exact Bayesian quantum state from digitized data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000313,"raw_usage":{"total_tokens":1849,"prompt_tokens":1089,"completion_tokens":760,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":705,"completion_tokens_details":{"reasoning_tokens":705}},"tokens_in":705,"tokens_out":760,"duration_ms":7575,"temperature":1.0,"reasoning_tokens":705,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T13:56:58.472866+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a homodyne or heterodyne experiment on a system with known Hamiltonian and monitoring strength, digitize at a coarse $\\Delta t$ where Euler schemes fail, histogram many runs' bin integrals, and compare with $\\operatorname{Tr}[K_{I}(\\rho_0)]$ computed by Gauss-Hermite quadrature; any discrepancy beyond sampling error refutes exactness. A sharper version inserts a low-pass filter before digitization: the paper's premise says $K_I$ must change, so the unmodified map should visibly mispredict the histogram.","supporting_citations":[{"cited_title":"Rouchon, A tutorial introduction to quantum stochas- tic master equations based on the qubit/photon system, Annual Reviews in Control54, 252 (2022)","cited_arxiv_id":null,"evidence_quote":"Provides the tutorial jump-measurement stochastic master equation whose tilted generator appears in the jump generalization."},{"cited_title":"Rouchon and J","cited_arxiv_id":null,"evidence_quote":"Introduced the Milstein correction that the perturbative expansion reproduces at second order."},{"cited_title":"Wonglakhon, H","cited_arxiv_id":null,"evidence_quote":"Derived higher-order completely positive trace-preserving unravelings that the expansion reproduces and extends."}],"review_version":1}