{"id":"82910f25-6cb7-42dc-8c1f-f358d7e07238","arxiv_id":"2412.02120","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Quantum ptychography can estimate pure multiqubit states on a superconducting quantum processor with high fidelity for up to four qubits when the final measurement uses an approximate quantum Fourier transform.","lead":"This paper tests quantum ptychography, a technique for estimating unknown pure quantum states, on a real IBM quantum processor. It shows the method works for small numbers of qubits, and that using an approximate Fourier transform makes it more resilient to device noise.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The load-bearing assumption is PIE convergence; the unexplained W-state failure for AQFT m=1 shows state-dependence, so the high fidelities reported for a few hand-picked states may not generalize.","rationale":"The reader's conditional verdict is well calibrated. The paper is honest about its scope: the simulations for AQFT use only five states, and the W m=1 failure is explicitly unexplained. The strongest claim, however, is specifically about the tested states, and for those states the data support it: with QFT and AQFT m=2, fidelities in Figs. 6, 7, and 9 are high, and the AQFT improvement over QFT is consistent with reduced circuit depth. My concern does not challenge those numbers; it challenges the inference that the method is a reliable estimator for arbitrary pure states. The PIE algorithm has no convergence guarantee, and the m=1 W-state anomaly demonstrates that the empirical behavior is not fully understood. This does not invalidate the paper's demonstration, but it does justify keeping the verdict conditional and asking for a random-state stress test and, ideally, code and data.","tokens_in":21639,"tokens_out":18193,"duration_ms":191080,"concrete_test":"Run Algorithm 1 exactly as in Sec. IV A (beta=2, Delta-beta=0.04, 50 iterations) in noiseless simulation with 2^13 shots for U = QFT and U = AQFT m=2, on 1,000 Haar-random pure states plus 1,000 random separable states for n=3, 4, and 5. Record the distribution of final fidelities and the fraction of states with mean fidelity below 0.99. If a non-negligible fraction fails, the reported 'all tests' claim does not generalize and the method's reliability is not established; if all pass, the convergence concern is empirically resolved for these protocols, and the W m=1 anomaly can be treated as an isolated artifact of that specific unitary.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is an empirical demonstration, so the key condition is that the PIE pipeline (Algorithm 1) actually converges to a state close to the true prepared state for the QFT and AQFT m=2 protocols. This is assumed, not proven. More importantly, it is not merely a missing proof: Sec. V A 2 reports that with the Hadamard transform (AQFT m=1) the W state is correctly estimated only for odd numbers of qubits, and the authors state 'we cannot explain this behavior' (Fig. 8(a)). That is direct evidence that the same algorithm, with the same projectors, is state- and unitary-dependent in a way that is not understood. The QFT and AQFT m=2 experiments used only the five states of Table III per n; no random-state stress test is reported for those protocols. If those five states happen to lie in a favorable region, the abstract's 'high fidelities in all tests' does not establish a reliable estimation method. The variable-feedback schedule (Delta-beta = 0.04, 50 iterations) was also chosen after 'several tests,' increasing the risk of overfitting to the tested states.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental and numerical implementation of quantum ptychography for estimating pure multiqubit states on an IBM superconducting processor. The protocol uses 3n circuits, each consisting of a single-qubit Pauli measurement followed by a final projective measurement in a basis generated by the quantum Fourier transform (QFT) or its approximation (AQFT). The authors present numerical simulations for up to 10 qubits and experiments on 2–5 qubits, reporting high fidelities for QFT up to n=3 and for AQFT of degree m=2 up to n=4. They also test separable unitary operations, which yield good estimates only for separable states. The paper introduces a variable-feedback modification to the ptychographic iterative engine (PIE) and a simple measurement-error-mitigation technique, and it compares the results with the PZD pure-state estimation method.","tokens_in":21883,"tokens_out":18632,"duration_ms":169003,"significance":"If the findings hold, this is a useful demonstration: quantum ptychography is implemented on a gate-based device and the AQFT is shown to be a more noise-robust measurement basis than the full QFT. The numerical simulations with random states for the QFT protocol are a strength, as are the high experimental fidelities for the tested states and the linear scaling of the number of circuits. However, the general reliability claim is weakened by the absence of a convergence proof for the PIE algorithm, an unexplained state-dependence observed for the Hadamard transform, and a non-matched cross-device comparison with the PZD method. The paper is likely to be of interest to the quantum characterization and NISQ benchmarking community, but the load-bearing points need clarification.","major_comments":[{"comment":"The text states that the AQFT of degree m is obtained by removing controlled-phase gates with index k > m and that for m=1 the transform reduces to the Hadamard product H^{⊗n}. However, Eq. (7) restricts the phase sum to n−m ≤ a+b, so for m=1 the sum includes the terms with a+b = n−1, which correspond to controlled-phase (controlled-Z) gates, not to H^{⊗n}. For example, for n=3 and m=1, Eq. (7) retains phases involving pairs (0,2) and (1,1), which is not the Hadamard transform. Since the PIE algorithm in Algorithm 1 uses the matrix U explicitly, this discrepancy changes the unitary assumed in the reconstruction and may explain the odd W-state behavior reported in Sec. V A 2. Please clarify which definition was implemented in the simulations and experiments and make Eq. (7) consistent with the textual description.","section":"Sec. V A 1 and Eq. (7)"},{"comment":"The performance comparison of ptychography with PZD is made between the average fidelity of ptychography and the median fidelity of PZD. Mean and median are not interchangeable unless the fidelity distribution is symmetric, which is not demonstrated; if the PZD distribution is left-skewed, this comparison can overstate the advantage of ptychography. For a fair comparison, either both statistics should be reported or the full distributions should be shown. This is load-bearing because the claim that ptychography outperforms PZD for n ≥ 6 is based on this comparison.","section":"Sec. IV B, Figs. 4(c) and 4(d)"},{"comment":"The authors report that for the Hadamard transform (AQFT m=1) the W state is correctly estimated only for odd numbers of qubits, and add 'we cannot explain this behavior.' This is direct evidence that the PIE algorithm's convergence is state- and basis-dependent in an uncontrolled way. The general claim in Sec. I that 'the algorithm will make an initial random guess converge to the true pure state that generated the data' is therefore not supported. A convergence analysis or at least a random-state stress test for the QFT and AQFT m=2 protocols would be needed to establish that the high fidelities are not an artifact of the specific five target states.","section":"Sec. V A 2, Fig. 8(a)"},{"comment":"The experimental comparison with the PZD method is performed on a different IBM device (ibm_perth vs ibmq_montreal) with different CNOT error rates and numbers of shots. Although the authors acknowledge these differences, the claim that the ptychographic results 'compare favorably' with PZD is confounded by the hardware generation gap. A more convincing comparison would use the same device or a noise-matched simulation. As written, the comparison in Table V should be interpreted only as an indicative benchmark, not a head-to-head evaluation.","section":"Sec. VI A and Table V"}],"minor_comments":[{"comment":"The update step '|˜φ_corr⟩ ← √Ω_ℓ' is a shorthand: the algorithm must replace the moduli of the amplitudes of U|φ_ℓ⟩ with the square roots of the measured probabilities while preserving their phases. The notation should say this explicitly, as written it is not a well-defined vector assignment.","section":"Algorithm 1"},{"comment":"The variable feedback schedule (β0=2, Δβ=0.04 or 0.1) was chosen 'after several tests' on the same data. The paper would be stronger with a sensitivity analysis showing that the reported fidelities are stable over a range of schedules, rather than optimized for the displayed states.","section":"Sec. IV A 2"},{"comment":"The statement that data are available upon reasonable request is not as useful as a public repository. Given the many free parameters (shuffling, transpilation, PIE runs, β schedule), releasing the Qiskit code and the experimental data would greatly improve reproducibility.","section":"Data Availability"},{"comment":"The W-state anomaly for odd vs even n is reported but not analyzed; at minimum, the authors could compute the exact unitary for m=1 and check whether the odd-even pattern is consistent with the phase structure of the reconstructed state. Adding this would help readers assess whether the failure is fundamental or an artifact of Eq. (7).","section":"Fig. 8(a)"}],"recommendation":"major_revision","confidential_remarks":"The most serious technical issue is the inconsistency between the textual definition of the AQFT and Eq. (7), which directly affects the m=1 and m=2 simulation results. The mean-vs-median comparison in Sec. IV B is also a common but serious flaw that should be corrected. The empirical demonstration itself is valuable, but the authors should temper the general convergence claim and either provide a convergence analysis or explicitly restrict the claim to the tested protocols and states."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing to know: this is the first circuit-based implementation of quantum ptychography on a real processor, and the AQFT substitution plus variable-feedback PIE are useful, real tweaks. The experiments show what they claim: with QFT, high fidelities for 2- and 3-qubit states; with AQFT m=2, high fidelities for up to 4 qubits including GHZ and W states. That is a genuine step beyond the earlier optical demonstrations, and the comparison with PZD is fair, with the caveat that they ran on different devices (they acknowledge this).\n\nWhat is good: the noise-free simulations up to 10 qubits with fixed shot counts are clean, the PZD comparison is informative (ptychography wins at larger n), and the error mitigation is simple and clearly described. The variable beta is a small but honest improvement; Fig. 3 shows the stagnation with fixed beta. The authors also report an unexplained failure transparently rather than hiding it.\n\nSoft spots, in proportion: the load-bearing assumption is PIE convergence, and there is no proof. More concretely, the m=1 AQFT result (Fig. 8a) shows the W state is estimated well only for odd n, and the authors openly say they cannot explain it. That is direct evidence of state-dependence in the pipeline, so the \"high fidelities in all tests\" for five hand-picked states per n does not by itself establish a reliable estimator. Second, the feedback schedule was chosen after \"several tests\"; without a random-state stress test for the experimental QFT/AQFT m=2 protocol, empirical overfitting to the chosen states cannot be ruled out. The simulations do include 100 random states per n, which partially addresses this, but those ran with different settings and did not test AQFT m=2 at scale. Third, no code or data beyond \"available upon request\" is a real gap for a methods paper. The missing convergence proof is a known feature of PIE in general, so I would not treat that alone as a flaw.\n\nWho this is for: people working on pure-state estimation, quantum device characterization, and NISQ-era methods. It deserves a serious referee: the demonstration is new, the comparison is useful, and the limitations are mostly addressable. My recommendation is to send to peer review, conditional on releasing code/data and on addressing the state-dependence question, at minimum by adding random-state experiments for the AQFT m=2 protocol or explaining the W-state anomaly.","headline":"A genuine first circuit-based demonstration of quantum ptychography on real hardware, with an honest but unexplained state-dependent failure; the core empirical claim holds, but the tuning and missing data make it a conditional accept.","tokens_in":22395,"tokens_out":1798,"would_cite":true,"duration_ms":18850,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P50","81P68"],"pacs":["03.67.-a","03.67.Lx"],"model":"deepseek-v4-flash","headline":"Quantum ptychography estimates pure multiqubit states on real quantum hardware, and an approximate Fourier transform pushes reliable reconstruction to four entangled qubits.","keywords":["quantum ptychography","pure state estimation","approximate quantum Fourier transform","PIE algorithm","Pauli measurements","superconducting quantum processor","NISQ devices","state fidelity"],"falsifier":"Run a noise-free simulation for n=6 with a random entangled state, use the degree-2 AQFT final basis and a very large number of shots (for example $10^{6}$), and check whether the PIE algorithm from many random starting estimates converges to fidelity above 0.99; if it does not, the method fails to estimate arbitrary pure states with those circuits.","tokens_in":1875,"feed_emoji":"⚛","tokens_out":7559,"duration_ms":140801,"temperature":0.7,"pith_summary":"The paper establishes that quantum ptychography, a phase-retrieval method that reconstructs an unknown pure state from overlapping projections followed by one final measurement, can be implemented on a circuit-based quantum computer. Using single-qubit Pauli measurements as the overlapping projections keeps the number of circuits linear in the number of qubits, 3n instead of the 3^n needed for full Pauli tomography. With the exact quantum Fourier transform as the final measurement basis, experiments on a superconducting processor estimate separable and entangled states of two and three qubits with high fidelity. Substituting a degree-2 approximate quantum Fourier transform reduces circuit depth and noise enough to also estimate four-qubit entangled states reliably. The paper compares favorably with a recently proposed scalable pure-state estimator and points to further improvements, such as nonseparable intermediate projections, for scalability on noisy devices.","feed_headline":"Approximate QFT extends quantum ptychography to 4 qubits","feed_subtitle":"Swapping the QFT for a degree-2 approximation cuts noise enough to estimate entangled 4-qubit states.","key_machinery":"The load-bearing object is the ptychographic iterative engine (PIE), an iterative phase-retrieval loop that starts from a random estimate, applies each overlapping projector, changes basis with the final unitary, corrects amplitudes with measured data, transforms back, and updates the estimate with a feedback parameter. The paper's key algorithmic modification is a decreasing feedback parameter, which acts like a learning-rate schedule and significantly improves convergence over a constant parameter. The second piece of machinery is the family of final unitaries: exact QFT, approximate QFT of degree m (which removes controlled-phase gates beyond distance m), and random separable unitaries. The approximate QFT of degree 2 reduces the transpiled circuit's two-qubit gate count enough to make the method viable on noisy hardware.","core_discovery":"The central claim is that the ptychographic protocol, originally proposed as a concept, works in practice on a noisy quantum computer when the overlapping projections are implemented through single-qubit Pauli measurements and the final basis is generated by a unitary operation. The authors show experimentally that the exact QFT yields median fidelities near 0.986 for two-qubit states and near 0.95 for three-qubit states, while a degree-2 AQFT improves three-qubit results (median 0.974) and makes four-qubit entangled state estimation reliable (GHZ fidelity 0.894, W fidelity 0.879). They also find that random separable final unitaries provide good estimations only for separable states, not for entangled states. Noise-free simulations up to ten qubits show average fidelities above 0.98 for both separable and arbitrary states, and in the same simulations the ptychographic method outperforms a recent scalable pure-state estimator for more than five qubits.","pith_inferences":["The success of degree-2 AQFT suggests that any low-depth unitary sufficiently biased away from product bases might serve as an effective final measurement basis; this could be probed systematically with random low-depth circuits.","A rigorous proof or counterexample for PIE convergence with the chosen projectors would determine whether the method is reliable for all pure states or only for the empirically tested families.","Combining nonseparable intermediate projections (such as Bell-state projectors between connected qubits) with a separable final basis is a concrete next step toward a fully low-depth, scalable ptychographic scheme, and is testable on existing hardware.","The observed advantage of AQFT echoes known decoherence-resilience results for approximate Fourier transforms, suggesting that optimizing the approximation degree m per device noise level could further improve fidelities."],"forward_implications":["Quantum ptychography can serve as a practical pure-state verification tool on current NISQ processors, needing only 3n circuits rather than 3^n.","Substituting low-degree AQFT for exact QFT improves estimation fidelity on noisy hardware, suggesting a general noise-versus-accuracy trade-off in phase-retrieval state estimation.","In noise-free simulations, ptychography matches or outperforms a recent scalable pure-state estimator beyond five qubits, indicating good scaling as system size grows.","Separable final measurement bases are insufficient for entangled states within this protocol, so scalability for entangled states requires nonseparable bases or nonseparable intermediate projections.","The variable-feedback PIE modification improves convergence and can be adopted in other implementations of quantum ptychography."],"supporting_citations":[{"why":"Original quantum ptychography protocol with single-qubit Pauli projectors and QFT; this paper implements and extends it.","marker":"36"},{"why":"Introduced the ptychographic iterative engine (PIE) used for reconstruction.","marker":"43"},{"why":"Provided the phase-retrieval algorithm foundation for the iterative estimation.","marker":"44"},{"why":"Defined the approximate quantum Fourier transform that the paper tests as a noise-reducing final unitary.","marker":"45"},{"why":"Showed that approximate Fourier transforms can outperform exact ones under decoherence, motivating the AQFT choice.","marker":"58"},{"why":"A recent scalable pure-state estimation method used as the comparison baseline in simulations and experiments.","marker":"38"},{"why":"Standard Pauli tomography, framing the exponential circuit count that ptychography avoids.","marker":"28"}],"fun_headline_variants":["AQFT ptychography estimates entangled 4-qubit states","Ptychography with approximate QFT reaches 4 qubits","Ptychography cuts circuits from 3^n to 3n","Approximate QFT enables 4-qubit quantum ptychography"],"cache_read_input_tokens":24576,"weakest_assumption_plain":"The PIE algorithm converges to the true pure state from any random initial estimate for the chosen projectors and final unitary, which is demonstrated only empirically for the tested states and noise levels, not proven.","fun_headline_variants_meta":{"raw":{"variants":["AQFT ptychography estimates entangled 4-qubit states","Ptychography with approximate QFT reaches 4 qubits","Ptychography cuts circuits from 3^n to 3n","Approximate QFT enables 4-qubit quantum ptychography"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000768,"raw_usage":{"total_tokens":3426,"prompt_tokens":990,"completion_tokens":2436,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":606,"completion_tokens_details":{"reasoning_tokens":2363}},"tokens_in":606,"tokens_out":2436,"duration_ms":19553,"temperature":1.0,"reasoning_tokens":2363,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T23:47:33.593014+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a noise-free simulation for n=6 with a random entangled state, use the degree-2 AQFT final basis and a very large number of shots (for example $10^{6}$), and check whether the PIE algorithm from many random starting estimates converges to fidelity above 0.99; if it does not, the method fails to estimate arbitrary pure states with those circuits.","supporting_citations":[{"cited_title":"Pereira , author L","cited_arxiv_id":null,"evidence_quote":"Introduced the ptychographic iterative engine (PIE) used for reconstruction."},{"cited_title":"Zambrano , author L","cited_arxiv_id":null,"evidence_quote":"Provided the phase-retrieval algorithm foundation for the iterative estimation."},{"cited_title":"Tariq , author A","cited_arxiv_id":null,"evidence_quote":"Defined the approximate quantum Fourier transform that the paper tests as a noise-reducing final unitary."},{"cited_title":"Goyeneche , author G","cited_arxiv_id":null,"evidence_quote":"A recent scalable pure-state estimation method used as the comparison baseline in simulations and experiments."}],"review_version":1}