{"id":"c0f6be2b-79f6-4d8a-9cf0-590e1b43f0bb","arxiv_id":"2506.05509","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":2.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"A numerical study of correlated dephasing noise on the 2-qubit Deutsch-Jozsa algorithm whose central non-monotonicity claim is contradicted by a direct analytic calculation.","lead":"Simulating the 2-qubit Deutsch-Jozsa algorithm under Ornstein-Uhlenbeck dephasing noise, this paper claims the success probability varies non-monotonically with noise correlation time and that Markovian noise models overestimate fidelity. The non-monotonic claim contradicts the paper's own circuit model, where fidelity is a monotonic function of correlation time.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed non-monotonic fidelity vs τc is impossible in the stated model: the third phase insertion after the final Hadamard is a measurement no-op, and the exact average fidelity 1/2 + (1/2)exp[-σ²Δt²(1+e^{-Δt/τc})] is strictly decreasing in τc.","rationale":"My independent derivation of the circuit confirms the reader's weakest_assumption: the third Rz after the final Hadamard multiplies only the |1> amplitude and leaves P(|1>) unchanged, so the effective noise duration is 2Δt, not 3Δt. The exact average is F = 1/2 + (1/2)exp[−σ²Δt²(1+e^{−Δt/τc})], which strictly decreases as τc grows because the OU autocorrelation ρ = e^{−Δt/τc} increases the variance of φ1+φ2. This makes the paper's central claim—non-monotonic fidelity versus τc with intermediate times 'more detrimental'—mathematically false within the stated model. I checked robustness to Rz phase-sign conventions: changing the sign of any subset of the three phases still yields a variance quadratic in ρ with no interior minimum on [0,1], so non-monotonicity cannot arise from the three-point model; it would require at least four alternating-sign insertions, which is not the paper's setup. The reported numbers compound the problem: for σ=5, the analytic curve runs from 0.889 to 0.803, while the paper reports 0.8426 at τc=0.1 and 0.7871 at τc=1.0—the latter is below the entire exact range and ~6σ from the analytic value given the stated statistics, so it cannot be attributed to finite sampling. The secondary claim that the matched Markovian model overestimates fidelity for large τc is directionally consistent with the monotone curve (F_OU→0.803 < F_Markov≈0.875), but it does not rescue the central non-monotonicity claim. This is a soundness failure, not a dispute with an alternative interpretation: the abstract, results, discussion, and conclusion all rest on a non-monotonicity that does not exist in the described circuit. No formal verification, code, or data accompanies the manuscript. The reader's REJECT verdict is appropriate; nothing in my stress-test changes it.","tokens_in":8076,"tokens_out":32807,"duration_ms":310660,"concrete_test":"Symbolically evaluate the paper's circuit: show P(|1>) = cos²((φ1+φ2)/2), with φ3 absent. Then compute the exact average F(τc) = 1/2 + (1/2)exp[−σ²Δt²(1+e^{−Δt/τc})] at σ=5, Δt=0.1 for τc = 0.1, 1.0, 10.0 and overlay the paper's Fig. 1 data points with per-point standard errors √[Var(P)/Ntraj] using Ntraj=100, Nshots=1024. If the reported minimum at τc=1.0 (0.7871) lies more than 3σ below the analytic curve, the simulation disagrees with its own model specification; if it lies within 3σ, the claimed non-monotonicity is an unsupported sampling artifact.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Following the circuit in Secs. IIA-C, the query qubit just before the final Hadamard is (|0> − e^{i(φ1+φ2)}|1>)/√2 (up to global phase). After the final H and the third Rz(φ3), the |1> amplitude carries an extra e^{iφ3}, so P(|1>) = |(1+e^{i(φ1+φ2)})/2|² = cos²((φ1+φ2)/2), independent of φ3. Hence the third noise insertion (Sec. IIC, point 3) cannot affect the measured fidelity, contradicting Sec. IV's 'approximately 3Δtstep' effective-duration argument. Averaging over the stationary OU phases, with Var(φk)=σ²Δt² and Cov(φ1,φ2)=σ²Δt²e^{−Δt/τc}, gives the exact fidelity F = 1/2 + (1/2)exp[−σ²Δt²(1+e^{−Δt/τc})], monotonically decreasing in τc. The central claim of non-monotonicity (Abstract, Sec. IIIA, Sec. V) is therefore false for the described model. The reported dips are also numerically inconsistent with this curve: for σ=5, Δt=0.1, the analytic F is 0.889 (τc=0.1), 0.855 (τc=1.0), and 0.803 (τc→∞), while the paper reports 0.8426 and 0.7871 at the first two points; with Ntraj=100 and Nshots=1024 the per-point standard error is ≈0.008–0.010, so these are ~6σ deviations, not sampling noise. The paper gives no code, no data, and no error bars, so the non-monotonic signature cannot be traced to a well-specified mechanism.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper numerically studies the fidelity of the 2-qubit Deutsch-Jozsa algorithm (balanced oracle) under dephasing noise modeled as an Ornstein-Uhlenbeck (OU) process, applied as discrete Rz rotations at three specified points in the circuit. The authors vary the noise strength sigma_OU and correlation time tau_c, and report that the fidelity depends non-monotonically on tau_c at higher noise strengths, with certain intermediate correlation times being more detrimental. They also compare with a Markovian phase-damping model matched via Eq. (4) and conclude that the Markovian approximation overestimates fidelity for tau_c > 0.1 Delta t. The results are obtained from Qiskit simulations with Ntraj = 100 and Nshots = 1024, but no code, data, or error bars are provided.","tokens_in":8557,"tokens_out":9261,"duration_ms":94184,"significance":"If the claimed non-monotonic dependence on tau_c were correct, it would be an interesting demonstration that noise memory timescales resonant with algorithmic operations can be particularly harmful, with practical implications for noise characterization and error mitigation. The comparison between correlated and Markovian dephasing is a relevant question. However, the central claim is analytically false for the model described in the paper: the fidelity is an exactly computable, strictly decreasing function of tau_c. The paper therefore does not establish its main finding, and the absence of code, data, and error bars prevents independent verification of the numerical results. The topic is relevant, but the core result is not defensible as presented.","major_comments":[{"comment":"The numerical evidence for non-monotonicity is not statistically quantified. No error bars, code, or raw data are provided, so the reader cannot assess whether the reported local minima are anything beyond Monte Carlo sampling noise. A rough estimate using the analytic distribution gives a standard error of the mean fidelity of order 0.02 for Ntraj = 100 at the relevant parameters, so the deviations of the reported values from the exact monotonic curve are not statistically significant. More fundamentally, the exact solution in the first comment shows that a genuine non-monotonic dependence is impossible in this model, so the burden is on the authors to identify and fix the implementation error.","section":"Section III and Fig. 1"}],"minor_comments":[{"comment":"Figure 1 shows curves without error bars and without specifying whether the lines are interpolations or guides to the eye. Given that the central claim rests on fine features of these curves, error bars are essential.","section":"Section II.D"}],"recommendation":"reject","confidential_remarks":"The paper's central claim is analytically impossible for the stated model, which is a load-bearing error that cannot be fixed within the manuscript's scope. The reported Markovian fidelity values suggest the implementation may have inadvertently used only two effective noise points, but the authors' interpretation assumes three, indicating a likely implementation bug. The absence of code, data, and error bars further prevents any salvage of the empirical claims. I recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead this one so you don't have to. The central claim—that the DJ fidelity depends non-monotonically on the OU correlation time τc—is wrong, and the paper's own circuit makes that clear. The third phase insertion after the final Hadamard is a measurement no-op: it multiplies the |1> amplitude by e^{iφ3}, which cancels in the computational-basis probability. The fidelity is exactly 1/2 + 1/2 exp[−σ²Δt²(1+e^{−Δt/τc})], strictly decreasing in τc. So there are no intermediate correlation times that are 'more detrimental.' The reported dips are sampling noise from 100 trajectories, and the paper gives no error bars.\n\nCredit where due: the paper is clearly written, the OU discretization is standard, and it honestly discusses its limitations (single qubit noise, 2-qubit toy). The idea of comparing with an equivalent Markovian channel is sensible. But these virtues don't outweigh the load-bearing flaw.\n\nThe specific errors: (1) Section IIC says three noise points, but the third doesn't affect the measurement; Section IV's 'approximately 3Δtstep' is false. (2) The non-monotonicity claim in the abstract, Sec. IIIA, and Sec. V is not supported by the analytic expression or by the data once sampling uncertainty is accounted for. (3) No code or data is provided, so the reported simulations can't be checked. The Markovian matching (Eq. 4) is a crude approximation and isn't derived properly, though that's a minor issue compared to the main one.\n\nThe stress-test note's '6σ' deviation is overstated—the correct standard error for 100 independent noise trajectories is about 0.02, not 0.008, so the reported points are consistent with the monotonic curve plus noise. But that doesn't help; it just means the data doesn't support the claim.\n\nThis paper doesn't clear the bar for refereeing. The central result is false, and the meaningful content reduces to a textbook calculation. Send it back without referee time. It might be worth using in a journal club as an example of why you check the circuit before trusting simulation curves.","headline":"The paper's central non-monotonicity claim is false: the third noise insertion is a no-op, and the exact fidelity is monotonic in τc.","tokens_in":8994,"tokens_out":8325,"would_cite":false,"duration_ms":89143,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68"],"pacs":["03.67.Lx","03.65.Yz"],"model":"deepseek-v4-flash","headline":"The paper claims that 2-qubit Deutsch-Jozsa fidelity depends non-monotonically on noise correlation time, with intermediate memory timescales most damaging, and that Markovian models overestimate fidelity when correlations persist.","keywords":["dephasing noise","non-Markovian noise","Ornstein-Uhlenbeck process","Deutsch-Jozsa algorithm","quantum algorithm fidelity","temporal noise correlations","quantum decoherence","noise simulation"],"falsifier":"Computing the conditioned success probability of the paper's balanced-oracle circuit as a function of the three phase errors $\\phi_1,\\phi_2,\\phi_3$ gives $\\cos^2((\\phi_1+\\phi_2)/2)$, independent of $\\phi_3$; averaging the two correlated Gaussian phases from the OU update yields $P=\\frac12+\\frac12\\exp[-\\sigma_{\\mathrm{OU}}^2\\Delta t_{\\mathrm{step}}^2(1+e^{-\\Delta t_{\\mathrm{step}}/\\tau_c})]$, which decreases monotonically with $\\tau_c$, and comparing this closed form with Fig. 1 (or deleting the third phase error in the simulation) would settle whether the reported dips are real.","tokens_in":7894,"feed_emoji":"⚛️","tokens_out":13685,"duration_ms":147088,"temperature":0.7,"pith_summary":"This paper asks whether temporally correlated dephasing noise degrades a basic quantum algorithm in ways that memoryless noise models miss. The author models the environment as an Ornstein-Uhlenbeck process and applies phase errors at three points in a 2-qubit Deutsch-Jozsa circuit with a balanced oracle. The central claim is that the algorithm's fidelity---the probability of measuring the correct answer, $|1\\rangle$---is non-monotonic in the noise correlation time $\\tau_c$, so an intermediate correlation time can be more damaging than very fast or very slow noise. The paper also claims that a Markovian dephasing channel matched to the same per-step error variance overestimates fidelity whenever $\\tau_c$ is not extremely short. If true, this implies that noise characterization for near-term devices must track correlation times, not just average error rates.","feed_headline":"Deutsch-Jozsa fidelity dips at intermediate noise memory times","feed_subtitle":"A 2-qubit Deutsch-Jozsa test suggests noise memory times, not just strength, set algorithm success.","key_machinery":"The central object is the Ornstein-Uhlenbeck (OU) process, a stationary Gaussian stochastic process with a finite correlation time $\\tau_c$, defined by the update $\\nu_{k+1}=\\nu_k e^{-\\Delta t_{\\mathrm{step}}/\\tau_c}+\\sigma_{\\mathrm{OU}}\\sqrt{1-e^{-2\\Delta t_{\\mathrm{step}}/\\tau_c}}N_k(0,1)$. Each phase error inserted into the circuit is $\\phi_k=\\nu_k\\Delta t_{\\mathrm{step}}$, applied as an $R_z(\\phi_k)$ rotation on the query qubit at three fixed circuit locations. This process is the only carrier of temporal correlation in the model: tuning $\\tau_c$ at fixed $\\sigma_{\\mathrm{OU}}$ changes how strongly successive phase errors are aligned, which is what the paper argues produces the non-monotonic fidelity profile. The comparison object is a Markovian phase-damping channel with damping parameter $\\lambda_{\\mathrm{pd}}=(\\Delta t_{\\mathrm{step}}\\sigma_{\\mathrm{OU}})^2$, chosen to match the OU single-step phase variance.","core_discovery":"On its own terms, the paper's discovery is that the success probability of the 2-qubit Deutsch-Jozsa algorithm under Ornstein-Uhlenbeck dephasing has a non-trivial dependence on the noise memory time $\\tau_c$. In the simulations, low noise ($\\sigma_{\\mathrm{OU}}=1.0$) leaves fidelity near 0.99 with only small $\\tau_c$ variations, while stronger noise ($\\sigma_{\\mathrm{OU}}=4.0,5.0$) produces dips at intermediate correlation times, for example a minimum of 0.7871 at $\\tau_c=1.0\\,\\Delta t_{\\mathrm{step}}$ for $\\sigma_{\\mathrm{OU}}=5.0$. The author interprets these dips as a memory effect: when $\\tau_c$ is comparable to the roughly $3\\,\\Delta t_{\\mathrm{step}}$ interval over which the three phase insertions act, consecutive phase errors do not average out and can accumulate. The paper further claims that a Markovian phase-damping channel with matched single-step variance predicts higher fidelity for $\\tau_c>0.1\\,\\Delta t_{\\mathrm{step}}$, with the gap growing with noise strength, and that even at the shortest simulated $\\tau_c$ the OU result is slightly below the Markovian prediction.","pith_inferences":["The paper's circuit admits an exact averaged success-probability formula; deriving it and overlaying it on the numerical curves would show which features of Fig. 1 are robust rather than artifacts of finite sampling.","The same phase-error-at-three-points structure should appear in other single-query interference algorithms, so the correlation-time sensitivity is a general design consideration, not a peculiarity of Deutsch-Jozsa.","An engineered-noise experiment that sweeps $\\tau_c$ at fixed $\\sigma_{\\mathrm{OU}}$ could directly test the predicted worst-case memory time.","Removing the third noise insertion or moving it to the ancilla would isolate which of the three phase points actually controls the fidelity, a direct way to test the paper's 'effective duration of about $3\\Delta t_{\\mathrm{step}}$' interpretation."],"forward_implications":["For noise strengths $\\sigma_{\\mathrm{OU}}\\ge 2.0$, the simulated fidelity versus $\\tau_c$ is non-monotonic, with the deepest minimum at $\\sigma_{\\mathrm{OU}}=5.0$, $\\tau_c=1.0\\,\\Delta t_{\\mathrm{step}}$ (fidelity 0.7871).","A Markovian dephasing model with matched per-step variance is accurate only near $\\tau_c=0.1\\,\\Delta t_{\\mathrm{step}}$; for larger $\\tau_c$ it overestimates fidelity, and the overestimate grows with noise strength.","Even the shortest simulated correlation time gives slightly lower fidelity than the Markovian channel, indicating residual memory or a distributional mismatch beyond the matched variance.","Benchmarks based only on Markovian error rates, such as $T_2$-type times, may overstate the success of interference-based algorithms in environments with noise correlations.","Noise characterization should include the correlation time or spectral shape of the dephasing environment, not just its strength."],"supporting_citations":[{"why":"Defines the Deutsch-Jozsa algorithm whose 2-qubit fidelity is the quantity being studied.","marker":"[6]"},{"why":"Provides the phase-oracle formulation of quantum algorithms used to write the oracle as a phase on the query qubit.","marker":"[4]"},{"why":"Introduces the Ornstein-Uhlenbeck process used as the temporally correlated dephasing noise model.","marker":"[17]"},{"why":"Supplies the exact discretized update formula for simulating the OU process.","marker":"[7]"},{"why":"Defines the Markovian open-systems framework that the paper sets out to test against correlated noise.","marker":"[2]"},{"why":"Documents non-Markovian dynamics in open quantum systems, motivating why temporal correlations matter.","marker":"[5]"},{"why":"Shows real qubit noise has measurable temporal correlations, supporting the premise that $\\tau_c$ is a relevant device parameter.","marker":"[3]"},{"why":"Provides the quantum circuit simulation tool used to generate the numerical fidelity data.","marker":"[14]"}],"fun_headline_variants":["Noise memory time matters: Deutsch-Jozsa fidelity dips at medium correlation times","Intermediate noise memory hurts 2-qubit Deutsch-Jozsa success most","Non-Markovian dephasing: fidelity drops at specific correlation times"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the phase error inserted after the final Hadamard gate contributes to the measured fidelity, so that all three noise insertions count equally; in the paper's own balanced-oracle circuit, that final phase error cannot change the probability of measuring $|1\\rangle$, leaving only the first two phase errors to set the outcome.","fun_headline_variants_meta":{"raw":{"variants":["Noise memory time matters: Deutsch-Jozsa fidelity dips at medium correlation times","Intermediate noise memory hurts 2-qubit Deutsch-Jozsa success most","Non-Markovian dephasing: fidelity drops at specific correlation times"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000316,"raw_usage":{"total_tokens":1848,"prompt_tokens":1061,"completion_tokens":787,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":677,"completion_tokens_details":{"reasoning_tokens":730}},"tokens_in":677,"tokens_out":787,"duration_ms":8393,"temperature":1.0,"reasoning_tokens":730,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T10:21:00.146860+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Computing the conditioned success probability of the paper's balanced-oracle circuit as a function of the three phase errors $\\phi_1,\\phi_2,\\phi_3$ gives $\\cos^2((\\phi_1+\\phi_2)/2)$, independent of $\\phi_3$; averaging the two correlated Gaussian phases from the OU update yields $P=\\frac12+\\frac12\\exp[-\\sigma_{\\mathrm{OU}}^2\\Delta t_{\\mathrm{step}}^2(1+e^{-\\Delta t_{\\mathrm{step}}/\\tau_c})]$, which decreases monotonically with $\\tau_c$, and comparing this closed form with Fig. 1 (or deleting the third phase error in the simulation) would settle whether the reported dips are real.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the Ornstein-Uhlenbeck process used as the temporally correlated dephasing noise model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the Markovian open-systems framework that the paper sets out to test against correlated noise."},{"cited_title":"Breuer and F","cited_arxiv_id":null,"evidence_quote":"Documents non-Markovian dynamics in open quantum systems, motivating why temporal correlations matter."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows real qubit noise has measurable temporal correlations, supporting the premise that $\\tau_c$ is a relevant device parameter."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the quantum circuit simulation tool used to generate the numerical fidelity data."}],"review_version":1}