{"id":"f947d863-2dd0-487f-b828-8b8a59e6c88b","arxiv_id":"2509.09395","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"A quantum-circuit-proposed MCMC sampler recovers the same cosmological posteriors as classical emcee on SNe Ia and CMB data, but with no demonstrated speedup.","lead":"This paper tests a hybrid sampler that uses a quantum circuit to propose each MCMC step, with accept/reject decided classically, for cosmological parameter estimation. It finds the same posterior contours as the classical sampler emcee on Planck and Pantheon+ data, but does not yet show a practical advantage.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"QMCMC's proposal distribution is not characterized: f in Eq. 7 is undefined and the circuit's randomness is unspecified, so the MH symmetry cancellation is unproven and convergence to the target posterior is not established.","rationale":"The reader's weakest_assumption and my read coincide. The only mechanism linking the chain to the target posterior is the Metropolis-Hastings ratio in Eq. 6. That ratio is computable only if the proposal density q is known or symmetric. Eq. 7 defines q through an unspecified step function f(Im(v)) and an incompletely specified quantum circuit; the paper does not say whether v comes from repeated statevector simulation (deterministic) or from measurement shots, and no gate angles are given. Consequently q is not a well-defined density, so the claimed symmetry cancellation is unverifiable. The empirical contour matches with emcee are suggestive but cannot carry the argument: many deterministic or asymmetric proposal rules could produce similar-looking chains on these posteriors without sampling π. This is an internal gap, not a disagreement with consensus. I therefore keep the reader's REJECT verdict; the fix is straightforward (define f and the circuit, or estimate the q-ratio), so a revised version could be assessed on that basis.","tokens_in":10850,"tokens_out":7034,"duration_ms":90556,"concrete_test":"Obtain from the authors (or the released code) the exact definition of f, the rotation angles of the circuit in Fig. 1, and the sampling mechanism that makes v stochastic. For a fixed current point θ, simulate the circuit N=10^5 times to obtain shifts s and estimate the proposal kernel q(θ'|θ); repeat from θ'=θ+s to estimate the reverse kernel q(θ|θ'). If the two kernels differ by more than Monte Carlo sampling error, Eq. 6 must include the ratio q(θ|θ')/q(θ'|θ); rerun the Ackley (or Pantheon+) analysis with this corrected acceptance and check whether the posterior contours shift. If all N draws of v are identical, the proposal is deterministic and the method is not a Metropolis-Hastings sampler at all.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that QMCMC yields posterior distributions consistent with emcee—requires that the Markov chain target the posterior π. In Section III, the proposal is θ' = θ + i·Re(v)·f(Im(v)) (Eq. 7). For the Metropolis-Hastings acceptance (Eq. 6) to be valid with the q-ratio cancelled, q(θ'|θ) must be symmetric. The paper neither proves this nor defines f, and it does not specify the rotation angles or the randomness that generates v from the circuit in Fig. 1—e.g., repeated statevector simulation would give a deterministic v, not a stochastic proposal. Without a characterization of the distribution of s, q is not a well-defined density; the proof that the chain has stationary distribution π is missing. The empirical agreement with emcee therefore is not evidence of correct sampling: it could arise from a deterministic walk, an asymmetric proposal with a different stationary distribution, or a proposal accidentally tuned to the target. This is an internal gap in the derivation, not a disagreement with external consensus.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a hybrid quantum-classical MCMC sampler (QMCMC) for cosmological parameter estimation. The proposal step is generated from a parameterized quantum circuit: a statevector v is computed and a shift s = i Re(v) f(Im(v)) is added to the current point; the candidate is accepted or rejected with classical Metropolis-Hastings. The authors apply the method to the Ackley test function in 8 dimensions, to a wCDM fit of SNe Ia (Pantheon+) with parameters (w, Omega_M, H0), and to a LambdaCDM fit of Planck TT with five parameters. They report contours, z-scores, and Pearson coefficients showing close agreement with emcee, and quote convergence diagnostics, acceptance rates, and effective sample sizes. The central claim is that QMCMC produces posterior distributions consistent with classical methods.","tokens_in":11075,"tokens_out":7215,"duration_ms":86502,"significance":"If correct, this would be a useful proof of concept for quantum-assisted sampling in cosmology, with an appealing scaling of qubit number as log2(d). The external comparison with emcee is a reasonable and non-circular benchmark, and the use of two real cosmological likelihoods goes beyond toy examples. The paper also has the merit of being candid about its limitations, including the use of a simulator rather than real quantum hardware. However, the manuscript currently lacks a well-defined proposal distribution and a proof of detailed balance; without these, the sampler has no guaranteed stationary distribution equal to the target posterior. The empirical agreement is suggestive but cannot validate the algorithm. If the missing definitions and correctness checks are supplied and the numerical experiments re-run with a fully specified transition kernel, the contribution could be of interest to the astrostatistics and quantum-computing communities.","major_comments":[{"comment":"The proposal distribution is not defined. Eq. (7) defines the shift using f(Im(v)), but f is never specified; the circuit's rotation angles and number of layers n_l are not given; and the text does not state whether v is obtained by a projective measurement (and in what basis) or by exact statevector simulation. On a statevector simulator, repeated calls return the same v, so the 'proposal' would be deterministic and no transition density q(theta'|theta) exists. Without q, Eq. (6) cannot be evaluated and Algorithm 1 does not define a Markov chain with a known stationary distribution. This is a load-bearing gap: all posterior results depend on this undefined kernel.","section":"Section III, Eq. (7) and Algorithm 1"},{"comment":"The acceptance rule uses Metropolis-Hastings, but the proposal ratio is omitted. The text states q cancels 'if it is symmetric' and then proceeds as if it did. No symmetry proof is given for the shift s = i Re(v) f(Im(v)); for generic rotation angles and a step function f, the distribution of s will generally be asymmetric, especially with the initial Hadamard layer producing real coefficients before rotations. If q is asymmetric, the chain's stationary distribution is not pi. The authors must either prove symmetry for the exact implementation, restrict the circuit/function family to make it symmetric, or compute and include q(theta|theta')/q(theta'|theta).","section":"Section II (Eq. (6)) and Section III (Eq. (7))"},{"comment":"The empirical agreement with emcee is a necessary consistency check, not sufficient validation of correct sampling. Agreement in contours and low z-scores could also arise from a deterministic or wrong-kernel walk on the smooth likelihood surfaces considered. The authors should add a formal validation: e.g., compare against known analytic posteriors on a multimodal target, estimate the proposal ratio from simulations, and verify detailed balance or reversibility of the implemented kernel. Without this, the central claim that the results are consistent with classical methods is not established.","section":"Section IV, Figs. 2 and 3"}],"minor_comments":[{"comment":"The symbol i is used for the initial step size, which is easy to confuse with the imaginary unit appearing in the same expression; use e.g. epsilon or sigma.","section":"Section III, Eq. (7)"},{"comment":"'The overall algorithm is summarized in III' should refer to Algorithm 1. Also, in the pseudocode, 'If stepmod n==0' should be typeset with a space.","section":"Section III, Algorithm 1"},{"comment":"Typos: 'the shiftss' should be 'the shifts'; 'aersimulator' should be 'aer simulator'.","section":"Section III and IV"},{"comment":"The z-score formula is not typeset correctly; define it in display math.","section":"Section IV"},{"comment":"The mapping from the 2^n-dimensional statevector to the d-dimensional parameter shift is not described for d that is not a power of two (e.g., d=5). Please specify how the statevector components are assigned to dimensions.","section":"Section III, step 2"}],"recommendation":"major_revision","confidential_remarks":"The paper reads as an early-stage proof of concept. The missing definitions and lack of a detailed-balance proof are substantial but appear fixable within the manuscript's scope: the authors can specify a stochastic measurement protocol, define f, and either prove symmetry or include the proposal ratio. If they do so and re-run the experiments, the contribution may become publishable. There is no evidence of circularity or dishonest benchmarking; the issue is incompleteness of the algorithmic definition."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the QMCMC paper. The empirical match to emcee is real, but the proposal distribution is not characterized well enough to know whether the chain is sampling the target. That's the thing to tell anyone before they cite it.\n\nWhat's new: using both real and imaginary parts of the statevector to build MH proposals, then applying that to SNe Ia and CMB likelihoods. I don't know of that specific combination in the literature. The paper is clearly written, and the pseudocode is helpful. The validation is standard: contour overlap, z-scores, Pearson rho, and they report acceptance rates and ESS. They also acknowledge the obvious limitations (emulator only, no hardware, no automation of step size). That honesty counts.\n\nThe soft spot is not a small one. Eq. 7 defines the shift as s = i * Re(v) * f(Im(v)), but f is never actually defined—\"step function\" is not a specification. The circuit's rotation angles are not given either, and the algorithm says \"generate quantum statevector v\" without saying how randomness enters. On a simulator, a fixed circuit gives a deterministic v. If v is deterministic, the proposal is deterministic, and the MH acceptance rule (Eq. 6) doesn't have a well-defined q(theta'|theta) at all. The symmetry cancellation is then unproven, and the chain need not converge to the posterior. The empirical agreement with emcee could come from the starting points or from the deterministic walk accidentally exploring the relevant region; it doesn't rescue the derivation.\n\nI want to be fair: the authors do say 'if it is symmetric,' so they flag the assumption. But they don't prove it, and given the all-positive Hadamard initialization and the unspecified f, it's unlikely to hold. So the central correctness claim is open.\n\nWho gets value: readers interested in whether quantum circuits can drive MCMC proposals will find a clear, honest attempt with some encouraging but not yet convincing results. It deserves a serious referee—send it to peer review, but the bar for acceptance should be major revision: fully specify f and the circuit/randomization, prove symmetry or use the correct MH ratio, and show the sampler works on a case where a deterministic walk would fail.","headline":"The empirical match with emcee is genuine, but the proposal distribution is underspecified—and possibly deterministic—so the sampler's correctness is not established.","tokens_in":11658,"tokens_out":3081,"would_cite":false,"duration_ms":35419,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A hybrid quantum-classical MCMC sampler reproduces classical posterior contours for supernova and CMB cosmological data.","keywords":["quantum Markov chain Monte Carlo","cosmological parameter estimation","Metropolis-Hastings","statevector proposal","wCDM","ΛCDM","CMB temperature spectrum","hybrid quantum-classical algorithm"],"falsifier":"Record the shift vectors proposed by the quantum circuit over many thousands of iterations, build histograms of s and of -s (or of forward and reverse proposal probabilities for pairs θ, θ'), and test whether they coincide. If they differ measurably, the proposal is asymmetric and the chain is not guaranteed to converge to the target posterior, so the reported agreement with the classical sampler would need reinterpretation. This check can be done with the same emulator used in the paper.","tokens_in":10714,"feed_emoji":"⚛️","tokens_out":8017,"duration_ms":88124,"temperature":0.7,"pith_summary":"This paper aims to show that a Markov Chain Monte Carlo sampler can use a quantum circuit to generate its proposed steps and still converge to the same posterior distributions as a classical sampler. The authors implement a hybrid algorithm in which the shift vector for each proposal is read from the statevector of a shallow quantum circuit, while the Metropolis-Hastings accept/reject decision is computed classically. They apply it to a multimodal test function and to two real cosmological likelihoods — wCDM fitted to Type Ia supernova distances and ΛCDM fitted to CMB temperature data — and report that the resulting contours, means, correlations, and z-scores match the classical reference. If this holds, it would mean quantum-generated proposals can be dropped into standard Bayesian inference for cosmology without biasing the inferred parameters, and the history-independent step generation could help escape local minima.","feed_headline":"Quantum sampler reproduces CMB and supernova posteriors","feed_subtitle":"A circuit-proposed MCMC returns the same cosmological constraints as classical sampling, with history-free steps.","key_machinery":"The proposal engine is a shallow quantum circuit: Hadamard gates at both ends, three layers of rotations about the y and z axes, and conditional y-rotations that entangle all qubits. Its output statevector v — the list of complex amplitudes describing the circuit's output state — is converted into a parameter-space shift by s = i·Re(v)·f(Im(v)), where i is a step-size hyperparameter and f is a step function applied to the imaginary part of v. This shift becomes the proposal in a classical Metropolis-Hastings update, whose acceptance rule is written with the proposal ratio cancelled, an assumption that the proposal distribution is symmetric. The circuit uses log2(d) qubits and has depth 2 + n","core_discovery":"The central claim is that the QMCMC algorithm, which proposes steps via the statevector of a quantum circuit and accepts or rejects them with classical Metropolis-Hastings, produces posterior distributions for cosmological parameters that are statistically consistent with those obtained from a standard classical MCMC sampler. The paper's evidence is that the 68% and 95% credible contours overlap almost completely in both the CMB and supernova runs, the z-scores of the means are near zero, and the pairwise Pearson correlation coefficients are nearly identical. On this basis the authors state that the algorithm correctly locates the region around the global minimum for the test function and fo","pith_inferences":["The paper's consistency claim rests on an unproven symmetry of the proposal distribution; a direct measurement of the shift distribution from the circuit would settle whether the acceptance rule is correctly specified or whether the agreement is specific to these test cases.","Because the quantum circuit is only a source of random shift vectors, the same algorithmic shell could work with any cheap history-independent random generator; the quantum origin is not essential to the Metropolis-Hastings framework, only to the particular implementation.","The 'avoids local minima' motivation is not demonstrated by the tests shown — the test function has local minima but the chains were not adversarially initialized there — so a targeted test with multimodal cosmological posteriors and poor initial conditions would be needed to see the claimed benefit.","On real quantum hardware, gate errors and connectivity constraints will perturb the statevector and therefore the proposal distribution; the authors acknowledge this, and the practical value of the method will depend on whether those perturbations corrupt the symmetry assumption used in the acceptance rule."],"forward_implications":["If the reported consistency is correct, quantum-generated proposals can be spliced into classical Metropolis-Hastings pipelines for cosmological parameter estimation without changing the inferred posterior.","The logarithmic qubit scaling means the circuit proposal remains small even for higher-dimensional cosmological problems, so the approach is in principle scalable beyond the 5-parameter case tested.","A history-independent proposal mechanism of this kind could reduce the risk of chains getting trapped in local minima on multimodal likelihood surfaces, a property the authors highlight as the motivation for the design.","The reported effective sample sizes and acceptance rates indicate the sampler can produce usable chains, but the paper does not claim a runtime speedup over classical sampling; the bottleneck is the classical likelihood evaluation and circuit transpilation."],"fun_headline_variants":["Quantum MCMC matches classic cosmic posteriors","Quantum sampler reproduces CMB and supernova results","Cosmology: quantum MCMC agrees with classical","Quantum Markov chain nails cosmological constraints","Quantum steps, classical acceptance: same cosmology"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing assumption is that the proposal distribution generated by Eq. 7 is symmetric, so q(θ|θ')/q(θ'|θ) cancels in the Metropolis-Hastings acceptance; the paper states this possibility without proof, and the step function f(Im(v)) is never defined, leaving the actual distribution of the proposed shift uncharacterized.","fun_headline_variants_meta":{"raw":{"variants":["Quantum MCMC matches classic cosmic posteriors","Quantum sampler reproduces CMB and supernova results","Cosmology: quantum MCMC agrees with classical","Quantum Markov chain nails cosmological constraints","Quantum steps, classical acceptance: same cosmology"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000118,"raw_usage":{"total_tokens":860,"prompt_tokens":626,"completion_tokens":234,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":370,"completion_tokens_details":{"reasoning_tokens":166}},"tokens_in":370,"tokens_out":234,"duration_ms":2796,"temperature":1.0,"reasoning_tokens":166,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T19:07:50.872842+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Record the shift vectors proposed by the quantum circuit over many thousands of iterations, build histograms of s and of -s (or of forward and reverse proposal probabilities for pairs θ, θ'), and test whether they coincide. If they differ measurably, the proposal is asymmetric and the chain is not guaranteed to converge to the target posterior, so the reported agreement with the classical sampler would need reinterpretation. This check can be done with the same emulator used in the paper.","supporting_citations":[],"review_version":1}