{"id":"f7094942-87f3-4d03-be2a-0aae0fc012c0","arxiv_id":"2507.15955","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A 108-mode simulation of a GKP-based measurement-based quantum computer shows a three-qubit Grover search beats the classical one-query bound only above about 10 dB of GKP squeezing.","lead":"Researchers simulated a photonic quantum computer that stores quantum bits in special light states, tracking more than 100 physical light modes at once. Under ideal no-loss conditions, a small quantum search beats a classical search only when the light is squeezed to about 10 dB.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The ~10 dB crossover is computed under the e^{-epsilon N} photon-dampening model and is only as 'fundamental' as that model; a two-parameter physical GKP envelope could shift the threshold.","rationale":"The reader's weakest assumption is exactly the load-bearing point: the finite-energy GKP state is modeled by e^{-epsilon N}, and the cited equivalence to other approximate GKP models is not enough to establish that the computed RB error rates and the 10 dB Grover crossover are independent of the envelope prescription. I agree with that identification. I considered two other candidates but did not make them the primary attack: (1) the FMPS simulations use randomized SVD and fixed domains without convergence certificates (Appendix B.1, B.5) - this is a real reproducibility risk, but the internal agreement between RB and analytics in the mid-squeezing range and the agreement between Grover data and the depolarizing estimate give some support that truncation errors are not dominant near 10 dB; (2) Appendix A.2 contains a phase inconsistency: Eq. 4 defines the magic Bell state with relative phase e^{i pi/4}, while the text after Eq. 23 says to add e^{i pi/8}. If the code followed the latter, the simulated T gate would not be the standard pi/8 gate and the Grover oracle would be wrong. This should be checked against the repository, but I am not basing the verdict on it because it is likely a typographical error. The primary concern remains model fidelity, and it can be settled by the concrete two-parameter envelope test. The verdict should stay CONDITIONAL: the paper is a solid simulation study, but the 'fundamental threshold' wording outruns the evidence, and the numerical claims are not yet independently verified with a different physical noise model or a pinned code/data release.","tokens_in":13347,"tokens_out":16167,"duration_ms":195103,"concrete_test":"Rerun the Sec. 5.2 Grover simulation with the same FMPS pipeline, truncation parameters, and 108-mode QRL layout, but initialize qunaught and Bell states with a two-parameter finite-squeezed GKP model, e.g. a state with independently controlled q and p envelope widths set by a single squeezing parameter, or the explicit squeezed-comb model psi(q) proportional to sum over exp(-(q - 2n sqrt(pi))^2 / (2 sigma^2)) with envelope exp(-r q^2), instead of e^{-epsilon N}. Repeat at 9, 10, and 11 dB with matched effective squeezing. If the success probability at 10 dB moves across the 13/28 line by more than the 95% confidence interval, the claimed fundamental threshold is model-dependent; if it does not move, the 10 dB crossover is robust to the envelope prescription.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central numerical claim in Sec. 5.2 and the Abstract is that ~10 dB of GKP squeezing is a fundamental lower bound for the three-qubit Grover search to beat the classical one-query bound of 13/28. This claim rests on the one-parameter non-unitary model e^{-epsilon N} used for all physical GKP states (Eq. 3, Appendix B.2). The only justification offered for calling this model 'fundamental' is the sentence in Sec. 2 that other prominent models 'have previously been shown to be equivalent' [17]. That cited equivalence is not derived here, and it is not shown that the transformations used in ref. [17] preserve the specific quantities that set the threshold: the gadget-level logical X/Z error rates, the RB error rate in Fig. 3, and the Grover depolarizing estimate in Eq. 8. In fact, Appendix B.2 explicitly notes that the spherical phase-space symmetry of the envelope is a special feature of e^{-epsilon N}. Real finite-squeezed GKP states produced with a single squeezing parameter generally have quadrature-dependent envelopes, and any additional loss or homodyne inefficiency breaks the symmetry further. If the envelope is anisotropic or the teeth have different q/p widths, the per-gadget error rates and the depolarizing fit change, which can move the 13/28 crossover away from 10 dB. Since the internal consistency checks in Figs. 3 and 7 compare the simulation only with analytical estimates built on the same noise premise, they cannot validate the premise itself. The threshold claim is therefore conditional on an untested modeling assumption, not a model-independent fundamental bound.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents an FMPS-based simulation of measurement-based quantum computation with GKP qubits in a quad-rail lattice (QRL), using the photon-dampening approximation e^{-epsilon N} for finite-energy GKP states. It performs randomized benchmarking for squeezing levels between 5 and 15 dB and simulates a three-qubit Grover search that involves 108 physical GKP modes. The central claim is that the Grover success probability exceeds the one-query classical bound 13/28 only above approximately 10 dB of GKP squeezing, which the authors call a fundamental squeezing threshold. The paper also contributes a deterministic magic-state injection scheme for T gates and a Pauli-frame tracking procedure for non-Clifford gates in the QRL.","tokens_in":13492,"tokens_out":7871,"duration_ms":87106,"significance":"If the numerical results are reliable, this is a significant demonstration that an FMPS method can simulate measurement-based CV circuits with more than 100 GKP modes and explicit finite-squeezing noise, going well beyond earlier few-qubit simulations. The agreement between the simulated RB error rates and the analytical estimates of Walshe et al. is a useful cross-check in an intermediate squeezing window, and the release of the simulation code is a practical asset. The significance is however qualified by two issues: the 'fundamental threshold' is computed under a single one-parameter noise model, and the manuscript does not report convergence tests for the FMPS truncation and fixed-domain approximations. The quantitative 10 dB number should therefore be regarded as model- and method-dependent until these points are addressed.","major_comments":[{"comment":"The central claim that ~10 dB is a fundamental lower bound rests entirely on the photon-dampening model e^{-epsilon N}. The paper cites ref. [17] for equivalence of approximate GKP models, but it does not show that the equivalence preserves the specific quantities that set the threshold: the gadget-level logical X/Z error rates, the RB error rate in Fig. 3, or the depolarizing estimate in Eq. (8). Appendix B.2 itself notes that spherical phase-space symmetry is a special feature of this model, and the fixed-domain simplification of Appendix B.5 relies on that symmetry. If real finite-squeezed GKP states have asymmetric envelopes, or if loss and detection inefficiency are included, the crossover can move. The authors should either establish the equivalence for the relevant observables or rephrase the claim as a prediction under this model; ideally they should also test at least one alternative envelope, such as individual squeezed teeth with unequal q- and p-widths.","section":"Section 2, Eq. (3), Appendix B.2"},{"comment":"The analytical curve in Fig. 7 is not an independent check of the numerical simulation. It uses r from the analytical dashed line in Fig. 3 rather than from the numerically estimated RB decay parameters, and the depolarizing survival probability p = 1 - 4r/3 in Eq. (8) is inserted by construction. Thus the agreement between the points and the curve in Fig. 7 validates the depolarizing error-propagation model only under the same noise premise; it does not validate the premise itself. The numerical points are independent, but the caption should state explicitly that the line is based on the analytical error-rate model and is not a fit to the Grover data.","section":"Eq. (8), Fig. 7"},{"comment":"No convergence data are reported for the two main FMPS approximations: the truncated randomized SVD and the fixed quadrature domain. The manuscript states that internal dimensions are 'capped somewhere on the order 64' and that the fixed domain loses no accuracy 'as long as the domain is big enough', but it does not state the domain size or truncation tolerance used, nor does it show that the 10 dB crossover is stable under tightening these parameters. Since the crossover is the main quantitative result, a convergence test, such as repeating the RB and Grover simulations at higher bond dimension and larger domain, is needed.","section":"Appendices B.1 and B.5"},{"comment":"The agreement between the numerical RB error rates and the analytical estimates is explicitly reported to fail below 7 dB and above 11 dB. The 10 dB crossover in Fig. 7 lies close to the upper edge of this deviation region, and the text offers only speculative 'interference effects' as an explanation. The authors should quantify the resulting systematic uncertainty in the analytical curve used in Fig. 7, or use the numerical RB rates for the Grover estimate, before claiming a precise threshold.","section":"Section 4.2, Fig. 3"}],"minor_comments":[{"comment":"The name of the algorithm is misspelled as 'Grovers' in the Abstract and in Section 5.1; it should be 'Grover's algorithm'.","section":"Abstract and Section 5.1"},{"comment":"The sentence 'the most prominent models including this one has previously been shown to be equivalent' has a grammar error and does not specify under which metric or for which quantities the equivalence is claimed; please clarify.","section":"Section 2"},{"comment":"The notation T^{dagger_{Xi in P}}_i is confusing and easy to misread; define the conditional exponent explicitly, for example by writing T_i when X_i is not in P and T_i^dagger when X_i is in P.","section":"Appendix A.3, Eqs. (16)-(17)"},{"comment":"The caption says 'The two different shades are only used to distinguish data sets' but the shades are not otherwise labeled; a legend or explicit per-curve labels would improve readability.","section":"Fig. 2 caption"},{"comment":"The statement that logical purity shows no significant decay with circuit depth is not supported by any figure or table; either show the data or soften the claim.","section":"Section 4.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is a useful numerical study, but the headline claim of a 'fundamental squeezing threshold' is stronger than what the model and the convergence evidence support. A more cautious framing, an explicit statement that the threshold is model-dependent, and robustness checks against at least one alternative GKP envelope would make the contribution suitable for publication. No concerns about citation practice or originality of the simulations themselves; the FMPS method and the QRL construction are prior work, but the large-scale simulations and the T-gate injection scheme are new."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing to know: this paper delivers the first circuit-level simulations of GKP-based measurement-based computation at 108 modes, and it makes a crisp, testable claim—three-qubit Grover beats the classical one-query bound only above about 10 dB of GKP squeezing. That is a real step beyond prior work, which was stuck at fewer than ten qubits or relied on abstract noise models. The FMPS method is not new, but applying it to the QRL with randomized benchmarking and a full algorithm is a genuine contribution.\n\nThe strongest part is the internal consistency. The RB estimates agree with the analytical Walshe et al. curves across the mid-squeezing range, and the Grover success probabilities track a depolarizing model built from the RB decay. The absence of logical purity decay is a nice diagnostic supporting the depolarizing approximation. The authors also state the limitations of their noise model and their fixed-domain approximation rather than hiding them.\n\nWhere the paper is softer: the word \"fundamental\" is doing too much work. The 10 dB threshold is computed under the e^{-epsilon N} photon-dampening model, which gives a spherically symmetric envelope. Real GKP states produced by a single squeezing parameter have anisotropic envelopes, and the equivalence cited to ref. [17] is not shown to preserve the specific gadget error rates or the Grover threshold. So the threshold is a property of that model, not a model-independent bound. The authors do say they exclude external noise, but they still claim a \"fundamental lower bound\" from intrinsic noise; that claim should be hedged to \"under the photon-dampening model.\"\n\nOther issues are minor but worth fixing: RB excludes depths below 7 and the fit residuals grow at the low and high squeezing ends, so the benchmarking validation is not as clean as the abstract implies. The FMPS simulations rely on random SVD and fixed domains without convergence certificates, which matters for trusting the quantitative threshold. Reproducibility is incomplete—the repo lacks a commit hash and data artifacts.\n\nNet: this is a serious, competently executed study. The central claim is plausible but conditional on a modeling premise; the right framing is \"under the e^{-epsilon N} model, ~10 dB is the crossover.\" The paper deserves peer review and will be useful to anyone working on GKP hardware or CV simulation. I would cite it for the simulation scale and the concrete squeezing target. Send it to a serious referee, but ask them to push on the noise-model dependence and reproducibility.","headline":"Solid simulation advance with a concrete GKP squeezing target, though the claimed 'fundamental threshold' is only as strong as the one-parameter noise model it assumes.","tokens_in":14248,"tokens_out":1334,"would_cite":true,"duration_ms":16894,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P68","81P45"],"pacs":["03.67.Lx","03.67.Pp","42.50.Ex"],"model":"deepseek-v4-flash","headline":"A classical simulation of 108 GKP modes finds that a three-qubit Grover search on the quad-rail lattice beats classical search only above about 10 dB of GKP squeezing.","keywords":["GKP qubits","measurement-based quantum computation","quad-rail lattice","functional matrix product states","finite squeezing noise","randomized benchmarking","Grover's algorithm","photonic quantum computing"],"falsifier":"Run the same three-qubit Grover circuit under an alternative noise model, such as a GKP code with asymmetric quadrature envelopes or with added photon loss and finite detection efficiency, and check whether the success probability crosses the classical $13/28$ bound at a squeezing level significantly different from 10 dB. A direct experiment preparing GKP states with measured squeezing near 10 dB and executing one query of a two-solution three-qubit Grover oracle would also settle the claim.","tokens_in":12987,"feed_emoji":"⚛️","tokens_out":8650,"duration_ms":78263,"temperature":0.7,"pith_summary":"The paper reports the first detailed simulation of a measurement-based quantum computation using Gottesman-Kitaev-Preskill (GKP) qubits in a quad-rail lattice (QRL) cluster state, involving over 100 GKP modes. Using the functional matrix product states (FMPS) method, it models the coherent noise that finite squeezing intrinsically puts on GKP states. Randomized benchmarking between 5 and 15 dB matches analytical error estimates for high-quality GKP qubits, supporting the use of simple depolarizing-noise models. As a practical demonstration, a three-qubit Grover search is simulated and found to beat the classical one-query bound only above roughly 10 dB of GKP squeezing, which the paper interprets as a fundamental lower bound on squeezing for quantum behaviour in this setting.","feed_headline":"10 dB of squeezing unlocks quantum search advantage in GKP simulation","feed_subtitle":"Simulating a 108-mode photonic circuit pins the squeezing needed for a three-qubit Grover search to beat classical odds.","key_machinery":"The machinery is the functional matrix product state (FMPS) representation of continuous-variable states, which keeps simulations tractable by truncating singular values and using randomized range finding for approximate SVDs. The circuit model is the quad-rail lattice (QRL), built from qunaught states and Bell pairs entangled by beam splitters, with homodyne measurements implementing Knill-style error-corrected Clifford gates and injected magic Bell pairs implementing $T$ gates. Noise enters through the non-unitary photon dampening operator $e^{-\\epsilon \\hat N}$, which models finite-energy GKP states, and simulations use fixed domains because the resulting envelopes are spherically symmetric and inherited from the ancillas.","core_discovery":"On the paper's own terms, the central discovery is that the FMPS formalism can simulate QRL computations with 108 physical GKP modes, and that under the photon-dampening noise model a three-qubit Grover's algorithm has a squeezing threshold of approximately 10 dB below which intrinsic finite-energy noise destroys any quantum advantage. The authors show through randomized benchmarking that the average gate error rate follows the analytical estimates of ref. [12] in the intermediate squeezing range, and that logical errors are well described as independent depolarizing noise because logical purity does not decay with circuit depth. They emphasize that this threshold is for systems with no external noise, so it is a fundamental rather than practical bound.","pith_inferences":["If the 10 dB threshold persists under other noise models, photonic experiments aiming for algorithmic quantumness would need GKP squeezing near 10 dB before any external loss is added; loss and detection inefficiency would only push the threshold higher.","The FMPS approach should extend to other bosonic codes and alternative cluster-state geometries, since its cost depends on the tensor-network bond dimension rather than on the specific code or entanglement pattern.","The proposed deterministic magic Bell state injection for $T$ gates could be tested in existing photonic platforms as a lower-overhead alternative to heralded magic state preparation.","The agreement between simulated randomized benchmarking and analytical estimates suggests that small-circuit error rates could be used to extrapolate hardware requirements for larger GKP computations."],"forward_implications":["At squeezing levels around and above roughly 10.5 dB, the analytical incoherent-noise estimates used for the QRL match the simulated gate error rates.","A three-qubit Grover search on the QRL succeeds with probability above the classical $13/28$ one-query bound only for GKP squeezing above about 10 dB.","Logical-level depolarizing noise is a good approximation of the physical error mechanisms across deep circuits.","The FMPS approach can handle over 100 GKP modes, so larger near-term GKP circuits are classically simulable.","Since logical purity does not decay with circuit depth, the dominant errors are logical errors from incorrect decoding of gadget syndromes, not degradation of the GKP states themselves."],"supporting_citations":[{"why":"introduces the photon dampening model $e^{-\\epsilon \\hat N}$ and the fault-tolerant CV framework used here","marker":"[1]"},{"why":"introduces the FMPS method that makes the large-scale simulations computationally feasible","marker":"[9]"},{"why":"introduces the quad-rail lattice (QRL) and provides the analytical gate error estimates that the randomized benchmarking results are compared against","marker":"[12]"},{"why":"establishes equivalence among approximate GKP codes, supporting the photon dampening model as representative of finite-energy GKP states","marker":"[17]"},{"why":"supplies the performance analysis of GKP error correction that justifies the Knill-type error correction and decoder assumptions in the QRL","marker":"[7]"},{"why":"provides the continuous-variable gate teleportation and error correction formalism used to derive gadget displacements and syndrome decoding","marker":"[19]"},{"why":"gives the logical density matrix extraction method used to define logical purity and success probabilities","marker":"[18]"}],"fun_headline_variants":["10 dB squeezing threshold for GKP quantum search advantage","Simulated 108-GKP-mode circuit sets 10 dB Grover threshold","GKP Grover needs 10 dB squeezing to beat classical odds","FMPS simulates 108-GKP-mode Grover with 10 dB threshold","Quantum search advantage requires 10 dB squeezing in GKP"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a real finite-energy GKP state is faithfully represented by applying the photon-dampening operator $e^{-\\epsilon \\hat N}$ to the ideal state, so that the only noise in the simulated circuits is this intrinsic coherent envelope noise; if physical GKP states have asymmetric or non-Gaussian envelopes, or if photon loss and detection inefficiency are added, the simulated error rates and the 10 dB crossover could shift.","fun_headline_variants_meta":{"raw":{"variants":["10 dB squeezing threshold for GKP quantum search advantage","Simulated 108-GKP-mode circuit sets 10 dB Grover threshold","GKP Grover needs 10 dB squeezing to beat classical odds","FMPS simulates 108-GKP-mode Grover with 10 dB threshold","Quantum search advantage requires 10 dB squeezing in GKP"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000818,"raw_usage":{"total_tokens":3519,"prompt_tokens":818,"completion_tokens":2701,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":434,"completion_tokens_details":{"reasoning_tokens":2610}},"tokens_in":434,"tokens_out":2701,"duration_ms":19311,"temperature":1.0,"reasoning_tokens":2610,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:22:22.985604+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same three-qubit Grover circuit under an alternative noise model, such as a GKP code with asymmetric quadrature envelopes or with added photon loss and finite detection efficiency, and check whether the success probability crosses the classical $13/28$ bound at a squeezing level significantly different from 10 dB. A direct experiment preparing GKP states with measured squeezing near 10 dB and executing one query of a two-solution three-qubit Grover oracle would also settle the claim.","supporting_citations":[{"cited_title":"Functional matrix product state simulation of continuous variable quantum circuits","cited_arxiv_id":null,"evidence_quote":"introduces the FMPS method that makes the large-scale simulations computationally feasible"},{"cited_title":"Streamlined quantum computing with macronode cluster states","cited_arxiv_id":"2109.04668","evidence_quote":"introduces the quad-rail lattice (QRL) and provides the analytical gate error estimates that the randomized benchmarking results are compared against"},{"cited_title":"Continuous-variable gate teleportation and bosonic-code error correction","cited_arxiv_id":null,"evidence_quote":"provides the continuous-variable gate teleportation and error correction formalism used to derive gadget displacements and syndrome decoding"}],"review_version":1}