{"id":"5e5fdba8-7d57-4b31-ab8b-c59da0ffad6e","arxiv_id":"2411.13437","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Flux-pulse-assisted readout on a fluxonium qubit achieves 94.3% assignment fidelity in 280 ns and an SNR-limited fidelity of 99.9% in 360 ns, the fastest fluxonium readout reported.","lead":"Researchers show that pulsing a fluxonium qubit's magnetic flux during measurement can make the readout faster and more accurate. Their approach achieves the fastest fluxonium readout reported so far, using a simple setup and no quantum-limited amplifier.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 99.9% SNR-limited fidelity is the least secure part of the central claim: it rests on Gaussian fits without error bars and is only 'verified' by a model with fitted n̄ and a 40 ns offset.","rationale":"The reader's weakest assumption centered on the efficiency calibration, but the 99.9% SNR-limited fidelity is actually computed from the raw histograms, not from η. The more direct vulnerability is the Gaussian-fit assumption and the absence of error bars. The assignment fidelity of 94.3% is robust. The model's two adjustable parameters (n̄, offset) weaken the claimed verification but do not affect the measured data. A non-parametric reanalysis would settle the 99.9% claim. This does not change the conditional verdict; it sharpens the condition.","tokens_in":14401,"tokens_out":11705,"duration_ms":119114,"concrete_test":"Using the provided data repository, take the FPA single-shot records at 360 ns integration time and compute the assignment error with a threshold optimized directly on the empirical |0> and |1> distributions, without assuming Gaussian shapes. Also bootstrap the SNR (Eq. 2) to obtain a 95% confidence interval. If the non-parametric assignment error exceeds 0.5% or the bootstrap interval for the SNR-limited error includes values above 0.5%, the 99.9% fidelity claim is not substantiated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The directly measured 94.3% assignment fidelity at 280 ns is well supported and not in question. The headline 99.9% SNR-limited fidelity at 360 ns, however, is computed by fitting the two single-shot histograms with single Gaussians and using Eq. 3. The paper gives no error bars or statistical uncertainty on this quantity. At the same integration time the assignment fidelity is only about 94%, so the 99.9% is an idealized overlap-based estimate. The authors themselves note that high photon numbers and proximity to the |6>-|0> transition can cause measurement-induced state transitions, which would produce non-Gaussian tails and make the Gaussian-fit SNR an overestimate. The semi-classical model said to verify the result requires n̄=75 while the calibration yields n̄=52, plus a 40 ns timing offset (Sec. III and App. C), so it does not independently corroborate the 99.9% value. If the true distribution tails matter, the 99.9% figure could be substantially overstated, weakening a central quantitative claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental demonstration of flux-pulse-assisted readout of a fluxonium qubit in a setup without a parametric amplifier. By flux-pulsing the qubit to a bias point with a larger dispersive shift, the authors achieve a directly measured assignment fidelity of 94.3% at 280 ns integration time, and they quote an SNR-limited assignment fidelity of 99.9% at 360 ns obtained from Gaussian fits to single-shot histograms. The results are compared with conventional sweet-spot readout and with prior fluxonium readout experiments, and a semi-classical model based on the authors' earlier proposal is used to reproduce the SNR-limited error curves. Measurement data and analysis code are made publicly available.","tokens_in":14605,"tokens_out":2672,"duration_ms":30992,"significance":"If the central claims hold, this is a useful advance: it shows that fast fluxonium readout does not require a quantum-limited parametric amplifier, and the 280 ns integration time is shorter than previously reported fluxonium readouts. The paper's strengths are the direct experimental measurement of assignment fidelity, the reproducible data and code repositories, the careful device characterization, and the explicit comparison table of prior work. The directly measured 94.3% assignment fidelity at 280 ns is well supported. However, the headline 99.9% SNR-limited fidelity is a derived quantity that currently lacks statistical uncertainty and depends on a model whose photon number and timing are fitted; this weakens the quantitative weight that can be placed on that specific number, though it does not undermine the core experimental demonstration.","major_comments":[{"comment":"The 99.9% SNR-limited fidelity at 360 ns is computed by fitting the |0> and |1> single-shot histograms with single Gaussians and applying Eq. (3), but the paper reports no confidence intervals, fit residuals, or goodness-of-fit statistics for these fits. Since the assignment fidelity at the same integration time is only about 94%, the 99.9% figure is an idealized overlap estimate, and the authors themselves note in Section III that high photon numbers and the proximity of the |6>->|0> transition to 3*omega_r can cause measurement-induced state transitions that would produce non-Gaussian tails. As this number appears in the abstract and in the 'fastest reported readout' claim, the authors should provide error bars on the Gaussian-fit parameters, show the histogram fits, and discuss or bound the effect of non-Gaussian tails on the SNR estimate.","section":"Section III, Eq. (3) and Fig. 3"},{"comment":"The semi-classical simulation is said to verify the experimental SNR-limited error, but the agreement is obtained with an average photon number of n_bar = 75 in the simulation, while the calibration in Appendix C yields n_bar = 52.1, and an additional 40 ns timing offset is introduced. Because these are fitted parameters, the model does not independently corroborate the 99.9% SNR-limited fidelity. The authors should quantify the sensitivity of the simulated SNR-limited error to n_bar, the timing offset, and the measured efficiency eta = 6.04%, and state explicitly what range of these parameters is consistent with the data; without such an analysis, the 'verification' claim is overstated.","section":"Section III and Appendix C"},{"comment":"The conversion from drive amplitude to in-cavity photon number relies on the dephasing model Gamma_d = 8*chi^2*n_bar/kappa and on the assumptions of a square measurement pulse and negligible ring-up time. The resulting n_bar = 52.1 is used both to justify the operating point and to set the simulation photon number, so an uncertainty in eta or in the dephasing model propagates directly into the quoted SNR-limited fidelity. The paper should provide an uncertainty budget for the calibration parameters a and sigma and show how the inferred n_bar changes under plausible variations of the model assumptions.","section":"Appendix C, Eq. (C1)"}],"minor_comments":[{"comment":"Several axis labels in the figures appear garbled in the manuscript text (e.g., the sequence '/uni00000013/uni00000011/...'), which makes the figures difficult to parse; the rendered labels should be fixed.","section":"Fig. 2 and Fig. 3"},{"comment":"The notation in Eq. (5), where '+(-)' corresponds to the |1>(|0>) state, is confusing because the symbol '±' already carries a sign ambiguity; please define the correspondence explicitly and unambiguously.","section":"Section III, Eq. (5)"},{"comment":"In Eq. (C1), the quantity tau_total is defined as the measurement pulse duration plus the idling time, but the text does not explain why the idling time should enter the photon-number conversion; a short justification or a reference to the protocol of Ref. [56] would help.","section":"Appendix C"},{"comment":"The code repository link should include a version tag or commit hash so that the analysis can be reproduced exactly as used in the paper.","section":"Data availability"}],"recommendation":"major_revision","confidential_remarks":"The directly measured 94.3% assignment fidelity at 280 ns is the solid core of the paper and appears reproducible from the shipped data. The main risk is the prominence given to the 99.9% SNR-limited fidelity, which is a model-dependent extrapolation rather than a directly measured quantity; the revision should either provide a rigorous uncertainty analysis or soften the claim. The paper fits the scope of the journal, and the comparison table is a useful contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis is a solid experimental paper. The group demonstrates flux-pulse-assisted readout on a fluxonium qubit, achieving 94.3% assignment fidelity in 280 ns without a parametric amplifier. That is a real record for fluxonium readout speed, and the central number is directly measured from single-shot experiments. Data and analysis code are public, which earns trust.\n\nWhat's new: this is the first experimental implementation of the flux-pulse-assisted readout scheme they proposed in Ref [24], and it parallels the dynamic-dispersive-shift readout shown with transmons in Ref [25]. The paper also quantifies a ~20% increase in dispersive shift and gives a semi-classical model that captures the trend. Calibration is careful, and they flag their setup's limitations (thermal population, measurement-induced transitions, no parametric amplifier) honestly.\n\nThe soft spots are around the 99.9% SNR-limited fidelity quoted for 360 ns. That number comes from Gaussian fits to the single-shot histograms, and the paper gives no error bars on it. The semi-classical model used as verification requires a mean photon number of 75 against the measured ~52, plus a 40 ns timing offset, both acknowledged but still fitting. So the 99.9% should be read as a model-dependent estimate of the SNR limit, not an independently confirmed value. The directly measured 94.3% assignment fidelity is not affected by this, and the paper's main claim holds up.\n\nCitation pattern is fine: they cite their own proposal, the transmon analog, and the comparison table against prior fluxonium readouts is useful. The paper is clearly written and mostly avoids overselling; the abstract gives the 99.9% figure more prominence than it can bear, but the body is transparent.\n\nWho should read this: anyone working on fluxonium readout, QEC with fluxonium, or fast dispersive readout. It deserves a serious referee. I'd suggest asking for error bars on the SNR-limited numbers and a clearer separation between measured and derived quantities, but the core experimental result is solid and worth publishing.","headline":"Solid experimental demonstration of fast fluxonium readout; the direct 94.3% fidelity at 280 ns holds up, but the 99.9% SNR-limited figure is a model-dependent estimate without error bars.","tokens_in":15187,"tokens_out":3820,"would_cite":true,"duration_ms":34934,"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":"Dynamic flux pulsing shrinks fluxonium readout to 280 ns while reaching 94.3% assignment fidelity, the fastest reported for this qubit type.","keywords":["fluxonium qubit","dispersive readout","flux pulse","dynamic flux control","assignment fidelity","quantum measurement","parametric amplifier","superconducting qubits"],"falsifier":"Perform an independent measurement of the in-cavity photon number during the flux-pulse-assisted readout, for example a calibrated AC-Stark shift at the same drive amplitude. If the resulting photon number disagrees with the η≈6% calibration, then the 99.9% SNR-limited fidelity is not supported; if it agrees, the central claim is confirmed.","tokens_in":14215,"feed_emoji":"⚡","tokens_out":9602,"duration_ms":84771,"temperature":0.7,"pith_summary":"The paper reports an experimental demonstration that reading out a fluxonium qubit can be made dramatically faster by pulsing its magnetic flux during the measurement. By temporarily shifting the qubit from its sweet spot to a flux bias where the dispersive shift between qubit and readout resonator is about 20% larger, the authors achieve an assignment fidelity of 94.3% with a 280 ns integration time, without using a parametric amplifier. The same data imply a signal-to-noise-limited fidelity of 99.9% at 360 ns, which the paper identifies as the fastest fluxonium readout reported so far. This matters because fast, high-fidelity readout is a prerequisite for mid-circuit measurement and feedback in quantum error correction, and fluxonium qubits have lagged behind transmons in this respect.","feed_headline":"Flux pulse hits 94.3% readout fidelity in 280 ns","feed_subtitle":"A 20% larger dispersive shift from a flux pulse delivers the fastest fluxonium readout reported.","key_machinery":"The central mechanism is a fast flux pulse applied during the readout window that moves the fluxonium from its sweet spot at $\\Phi_{\\mathrm{ext}}/\\Phi_0=0.5$ to a bias point $\\Phi_{\\mathrm{ext}}/\\Phi_0 \\approx 0.6567$, where the higher-energy transition $|2\\rangle \\to |0\\rangle$ is closer to the readout resonator, increasing the dispersive shift magnitude by roughly 20%. The readout signal is modeled by a differential equation for the intracavity coherent amplitude, $\\dot{\\alpha} = -i\\Delta_{\\pm}\\langle\\sigma_z\\rangle\\alpha - \\frac{1}{2}\\kappa\\alpha - \\sqrt{\\kappa}\\alpha_{\\mathrm{in}}$, where $\\Delta_{\\pm}$ captures the time-dependent detuning of the readout tone from the state-dependent resonator frequencies; the model converts the cavity output into an SNR and hence an SNR-limited error. The calibration that converts room-temperature pulse amplitude to in-cavity photon number, using a measurement-efficiency protocol giving $\\eta=6.04\\%$, is also part of the machinery, as the quoted 99.9% depends on it.","core_discovery":"The central claim is that flux-pulse-assisted readout makes fluxonium qubits readable as fast as or faster than transmon-based readout, even with an inefficient (≈6%) measurement chain. The experiment shifts the qubit from $\\Phi_{\\mathrm{ext}}/\\Phi_0 = 0.5$ to $0.6567$ during the readout window, raising the magnitude of the dispersive shift from $0.92$ MHz to $1.09$ MHz. This yields an assignment fidelity of 94.3% at 280 ns; if only the separation of the measured signal distributions is considered, the fidelity is 99.9% at 360 ns. The paper also shows that a simple semi-classical model of the cavity field reproduces the observed integration-time dependence, provided the model uses an average photon number $\\bar{n}=75$ and a 40 ns timing offset, and it attributes the gap to uncertainties in the measurement-efficiency calibration.","pith_inferences":["The technique should transfer directly to other flux-tunable qubit platforms whose dispersive shift changes sharply near an avoided crossing, not just fluxonium; the transmon demonstration cited as Ref. [25] already hints at this.","Because the assignment error plateaus at roughly 6% for integration times above 300 ns, the next bottleneck is initialization and relaxation, not readout; faster reset protocols like sideband driving could push assignment fidelity toward the 99.9% SNR limit.","The mismatch between the simulated photon number (75) and the calibrated one (≈52) suggests the measurement-efficiency calibration carries a systematic uncertainty of order 30%, so the 99.9% figure is an estimate until independently verified.","If the flux-pulse-assisted readout is combined with a parametric amplifier, the authors' own reasoning implies assignment fidelities above 99% at sub-300 ns integration times are plausible, which would put fluxonium readout on par with transmon readout for error correction."],"forward_implications":["Fluxonium readout no longer requires a quantum-limited parametric amplifier to reach sub-300 ns measurements; a flux pulse suffices to get 94.3% assignment fidelity at 280 ns.","Sub-microsecond, mid-circuit-capable fluxonium readout is within reach, since 280 ns is short compared to typical fluxonium coherence times near 1 ms.","Adding a parametric amplifier to the same readout chain should allow fewer photons per measurement, reducing measurement-induced transitions and enabling even higher fidelity or larger dispersive shifts.","The semi-classical model validated in the paper can be used to select the optimal flux-pulse bias and readout frequency for other fluxonium devices without full quantum simulation."],"supporting_citations":[{"why":"Proposes flux-pulse-assisted readout of fluxonium; the present experiment is the demonstration of that proposal.","marker":"[24]"},{"why":"Supplies the measurement-efficiency extraction protocol that gives η=6.04%, the calibration behind the SNR-limited fidelity.","marker":"[56]"},{"why":"Demonstrates dynamic control of the dispersive shift in transmons, providing the complementary experimental evidence for the technique.","marker":"[25]"},{"why":"Reports fluxonium readout fidelity at 1.5 µs integration without a parametric amplifier; key comparison for the 'fastest reported' claim.","marker":"[45]"},{"why":"Reports high-fidelity fluxonium readout using a parametric amplifier; benchmark showing competitive performance without one.","marker":"[15]"},{"why":"Reports fluxonium readout at 0.6 µs with a parametric amplifier; another comparison point in the readout-fidelity figure.","marker":"[47]"}],"fun_headline_variants":["Flux pulse hits 94.3% fidelity in 280 ns","Fluxonium readout record: 280 ns via flux pulse","No amplifier needed: flux pulse speeds readout","20% shift boost gives fast fluxonium readout","Flux pulsing delivers 94.3% fidelity at 280 ns"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quoted 99.9% SNR-limited fidelity rests on the calibration that maps readout pulse amplitude to in-cavity photon number (a measurement-efficiency value of η≈6%); the numerical model needs 75 photons while the calibration yields about 52, so if that calibration is off, the fidelity numbers move.","fun_headline_variants_meta":{"raw":{"variants":["Flux pulse hits 94.3% fidelity in 280 ns","Fluxonium readout record: 280 ns via flux pulse","No amplifier needed: flux pulse speeds readout","20% shift boost gives fast fluxonium readout","Flux pulsing delivers 94.3% fidelity at 280 ns"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000606,"raw_usage":{"total_tokens":2852,"prompt_tokens":998,"completion_tokens":1854,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":614,"completion_tokens_details":{"reasoning_tokens":1767}},"tokens_in":614,"tokens_out":1854,"duration_ms":13353,"temperature":1.0,"reasoning_tokens":1767,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T16:24:15.369483+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Perform an independent measurement of the in-cavity photon number during the flux-pulse-assisted readout, for example a calibrated AC-Stark shift at the same drive amplitude. If the resulting photon number disagrees with the η≈6% calibration, then the 99.9% SNR-limited fidelity is not supported; if it agrees, the central claim is confirmed.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the measurement-efficiency extraction protocol that gives η=6.04%, the calibration behind the SNR-limited fidelity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports fluxonium readout fidelity at 1.5 µs integration without a parametric amplifier; key comparison for the 'fastest reported' claim."}],"review_version":1}