{"id":"485255e2-364f-4d07-a344-1f482b4441e1","arxiv_id":"2411.15973","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"A pairwise Bell-state entanglement trick reduces the trainable parameter count in a quantum diffusion model, but the reported image quality scores are too low to support the claimed advances.","lead":"This paper proposes a quantum diffusion model that uses pairwise Bell-state entanglement to cut trainable parameters and speed up image generation on small qubit counts. The reported gains are undercut by very poor generation quality and a missing explanation of how the quantum output becomes an image.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The measurement-to-image readout is never specified; every reported metric depends on this undefined mapping, so the central performance claim is unverifiable and likely impossible with only computational-basis measurements.","rationale":"The reader's weakest_assumption identifies exactly the same load-bearing concern: the readout step is undefined. I agree that this is the most critical gap. The paper's central claim is about performance metrics, and every one of those metrics depends on converting quantum measurement outcomes into images. Without a specified observable, shot count, and reconstruction map, the experimental section is not reproducible, and the NISQ claim is unsupported. The parameter-reduction mechanism may be internally coherent, but the headline claim of improved FID/SSIM/PSNR cannot be evaluated. A secondary issue is that the training objective in Eq. (6) is not a standard diffusion objective, but even if the model were relabeled as a variational quantum autoencoder, the readout problem would remain. The proposed concrete test is decisive: it forces the authors to either reveal the measurement procedure or show that finite-shot computational-basis measurements cannot produce the reported results. If the authors supply code with full statevector access, the NISQ claim fails; if they supply a valid finite-shot reconstruction map, the metrics can be checked. Without either, the verdict of REJECT stands.","tokens_in":8617,"tokens_out":5175,"duration_ms":55129,"concrete_test":"Obtain (or re-implement from the paper) the exact EEQDM code and locate the measurement stage. Replace the final circuit evaluation with an explicit computational-basis measurement using a finite number of shots (e.g., 1024) on all qubits, and require the published reconstruction map to convert these outcomes into a 16x16 image. If the original code calls qml.state() or otherwise accesses the full statevector, the NISQ claim fails; if it uses samples, check that the number of real values extracted per image is at least 256 and that the Table 1 / Fig. 5 metrics reproduce. If no code or reconstruction map is supplied, run the same experiment with only computational-basis measurements; if SSIM/PSNR/FID do not beat the random baseline, the reported metrics necessarily came from an unspecified non-NISQ procedure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The single load-bearing gap is the undefined readout in the EEQDM pipeline. Section 3.1 describes amplitude encoding of N-pixel images into ceil(log2(N)) qubits, pairwise Bell-state preparation, and a PQC applied to a subset of qubits; Section 3.3 then defines loss via \"reconstructed data\" x-hat, and Section 5 reports FID/SSIM/PSNR against images. Nowhere does the paper state (i) which observable is measured, (ii) how many shots are used, or (iii) how the classical outcomes c in Fig. 1 are mapped back to an N-pixel image. For 16x16 MNIST, N=256 values are encoded in 8 qubits; a computational-basis measurement of 8 qubits yields at most 8 bits, which is a single index, not 256 pixel intensities. Reconstructing the full image therefore requires either full statevector access (contradicting the NISQ claim) or an undisclosed tomography/post-processing step. Since every reported improvement in Section 5 is computed from these reconstructed images, the central claim collapses if the mapping is absent or impossible.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes an Entanglement-Enhanced Quantum Diffusion Model (EEQDM) that amplitude-encodes images into ceil(log2 N) qubits, applies Hadamard-CNOT Bell-pair entanglement across the register, runs a parameterized quantum circuit on a two-qubit subset, and then measures the output. Training is described as minimizing an MSE reconstruction loss, and the model is benchmarked against a quantum denoising diffusion model (QDDM) and a classical DDPM on 8x8 and 16x16 MNIST and 16x16 CIFAR-10 images. The paper claims 40-47% parameter reduction, faster execution, and improved FID, SSIM, and PSNR. I assess these claims against the text, equations, and reported tables.","tokens_in":32,"tokens_out":7971,"duration_ms":125390,"significance":"If the proposed architecture worked as claimed, a parameter-efficient quantum generative model trainable on NISQ hardware would be a useful contribution. The paper deserves credit for providing explicit parameter counts and runtime comparisons and for benchmarking against both quantum and classical baselines. However, the central performance claims are not supported by the reported numbers (FID 382.36, SSIM 0.0433, PSNR 10.65 dB on the main evaluated tasks), and the experimental pipeline is incompletely specified, so the significance is not established as written. The paper also contains no code, no data-release statement, and no formal proofs of the expressibility claims.","major_comments":[{"comment":"The measurement-to-image readout is never specified. Amplitude encoding uses 8 qubits for a 16x16 image (256 pixels), and Fig. 1 shows only a generic \"Measurement stage\" with classical outcomes c. Section 3.3 defines the loss in Eq. (6) on \"reconstructed data\" x-hat, and Table 1 and Fig. 5 report image metrics, but the paper never states which observable is measured, how many shots are taken, or how the measurement outcomes are mapped back to 256 pixel intensities. A computational-basis measurement on 8 qubits yields one index, not 256 values; reconstructing the full image therefore requires either full statevector access (contradicting the NISQ claim) or an undisclosed tomography/post-processing step. Every reported metric depends on this unspecified step, so the central experimental claim is unverifiable.","section":"Section 3.1, Figure 1, Section 3.3"},{"comment":"The training objective is not a diffusion objective. Eq. (2) defines a forward noising process q(xt|xt-1), but no noisy samples or noise schedule appear in the loss. Eq. (6) is a plain MSE between the original and reconstructed data, with no timestep conditioning and no denoising term; Eq. (5) simply iterates the same circuit on its own output. Consequently the model is, as formulated, a quantum autoencoder rather than a diffusion model, and the \"quantum diffusion\" claim in the title and abstract is not supported by the training procedure.","section":"Section 3.2, Eq. (2), Section 3.3, Eq. (6)"},{"comment":"The reported quality metrics contradict the abstract's claim of high performance. For MNIST, Table 2 gives a mean FID of 382.36 (std 74.66), a value that indicates essentially no resemblance to the data distribution; Section 5.3 reports SSIM of 0.0433 and PSNR of 10.65 dB for EEQDM on CIFAR-10, which are near the level of random or severely corrupted images. The improvements over QDDM (e.g., SSIM 0.0433 vs 0.0169) are numerically small and qualitatively meaningless. These numbers do not support \"significant improvements in performance metrics\" as stated in the abstract.","section":"Section 5.4, Table 2, Section 5.3"},{"comment":"The parameter-reduction claim is true by construction but is presented as an advantage. Because the PQC is applied only to qubits q0 and q1, the parameter count scales with the depth of that small circuit and is therefore 40-47% lower than a PQC acting on the full register. This reduction is a design choice, not evidence that pairwise Bell entanglement allows a two-qubit circuit to \"access information from the entire input state\" as asserted in Section 3.1; no expressibility analysis, functional test, or controlled experiment supporting that assertion is provided. The efficiency and performance claims therefore rest on an unvalidated assumption.","section":"Section 3.1, Fig. 2"}],"minor_comments":[{"comment":"There is a typo in the first paragraph: \"and and complexity\" should read \"and complexity\".","section":"Section 1"},{"comment":"The heading reads \"CIFAFR10\" and should be \"CIFAR-10\".","section":"Section 5.3 heading"},{"comment":"The notation for the number of features and qubits is conflated: Eq. (1) uses n as the dimension of the data vector, while Section 3.1 states amplitude encoding uses log(n) qubits; please define n consistently and specify the logarithm base.","section":"Section 3.1, Eq. (1)"},{"comment":"The symbol x0 is used for both random noise and clean data, and the reverse process is written as x1,...,xT = ReverseProcess(x0), which is reversed relative to standard diffusion notation; please clarify the indexing.","section":"Section 3.2, Eqs. (3)-(5)"},{"comment":"The text says execution time increases exponentially with parameter count, but Table 1 shows an approximately linear increase (994 s at 150 parameters versus 5097 s at 750 parameters); please correct the description.","section":"Section 5.2, Table 1"},{"comment":"The classical model's PSNR of -46.77 dB is incompatible with Eq. (9) for any nonnegative MSE; please check the normalization or calculation of this metric.","section":"Section 5.3"},{"comment":"The axes of Fig. 2 are unlabeled, so it is unclear whether the parameter count is total or per layer.","section":"Figure 2"},{"comment":"The text says images are flattened as 2D, but CIFAR-10 images have three color channels; please specify how the color channels are flattened and amplitude-encoded.","section":"Section 4.1"}],"recommendation":"reject","confidential_remarks":"The undefined readout alone would justify rejection, and the reported metrics show that the model is not performing at a level that supports the paper's claims. I do not see a straightforward revision path within the manuscript's current scope; a resubmission would need a fully specified measurement protocol, a genuine diffusion training objective, and substantially better or recalibrated claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe thing to know about arXiv:2411.15973: the circuit idea is real but the paper leaves out the step that connects the quantum measurement to the reported image metrics. Without that readout specification, the FID/SSIM/PSNR numbers could be anything.\n\nWhat's new: the specific assembly—amplitude encoding, pairwise Bell-state preparation on first/second half of the register, then a PQC acting only on a subset of qubits—does not appear in the cited QDDM papers. The parameter reduction is genuine, though it follows by construction from applying rotations to two qubits. The reported comparison against QDDM shows consistent improvements in loss, SSIM, PSNR, and runtime, and the FID is lower (382 vs 420), even if both values are near the random-noise level for MNIST.\n\nThe soft spots are substantial. The readout is never defined. A 16x16 image holds 256 pixel values; amplitude encoding puts them into 8 qubits. A computational-basis measurement of 8 qubits gives one index, not 256 intensities. To reconstruct the image you need either full statevector access (which contradicts the NISQ claim) or an unstated tomography/post-processing step. Every metric in Section 5 depends on this mapping, so the central claim is unverifiable until the mapping is specified. Second, Eq. (6) is a plain reconstruction MSE with no forward noising schedule; calling this a diffusion model is a stretch. Third, the abstract's 'significant improvements' is not supported by a mean FID of 382. Fourth, there are citation errors and no code or data artifacts.\n\nWho gets value from this? Someone working on parameter-efficient quantum generative models could take the entanglement trick as a starting idea, but the paper as written does not establish its central claim. A serious referee would ask the authors to specify the readout, the shot count, and the mapping back to pixels; if the readout turns out to require full statevector simulation, the NISQ claim collapses.\n\nMy recommendation: do not accept as is. But I would send it to peer review rather than desk reject—the novelty is enough to warrant expert eyes, and the readout issue, while severe, is fixable in principle if the authors can describe a legitimate measurement scheme.","headline":"Novel entanglement trick, but the undefined measurement-to-image readout makes the reported performance unverifiable.","tokens_in":9351,"tokens_out":3559,"would_cite":false,"duration_ms":32783,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that pairwise Bell-state entanglement lets a quantum diffusion circuit act on a subset of qubits while still accessing the whole input, cutting parameters by 40-47% and improving image metrics.","keywords":["quantum diffusion models","Bell state entanglement","parameterized quantum circuits","generative modeling","NISQ devices","image generation","amplitude encoding","FID"],"falsifier":"Reproduce the reported numbers with an explicit measurement stage: amplitude-encode 16x16 MNIST into 8 qubits, apply the Bell-pair and parameterized circuit, and reconstruct all 256 pixel values using only computational-basis measurement outcomes with a finite shot budget. If the loss, SSIM, PSNR, and FID values cannot be matched unless the full statevector is accessed, the claim that EEQDM is a NISQ-ready generative model is wrong.","tokens_in":8402,"feed_emoji":"⚛️","tokens_out":9841,"duration_ms":80506,"temperature":0.7,"pith_summary":"The paper tries to establish that entangling qubits in pairs before the trainable circuit runs makes a quantum denoising diffusion model substantially more parameter-efficient without sacrificing expressive power. The proposed Entanglement-Enhanced Quantum Diffusion Model (EEQDM) amplitude-encodes an image by mapping its pixel values onto the amplitudes of $\\lceil \\log_2(N)\\rceil$ qubits, creates Bell pairs between the first and second half of the register, and then applies a parameterized quantum circuit to only part of the register. The authors report a 40-47% reduction in trainable parameters relative to an existing quantum diffusion model, together with improved loss, SSIM, PSNR, and FID on downsampled MNIST and CIFAR-10 images and a roughly twofold speedup on the larger color images. If true, this makes the model a candidate for running generative tasks on noisy intermediate-scale quantum (NISQ) devices, because the entanglement pairs are supposed to give the reduced set of qubits access to information from the entire input state.","feed_headline":"Bell-pair entanglement cuts quantum diffusion parameters by 40-47%","feed_subtitle":"Entangling the first and second half of the register lets fewer qubits carry the whole image and improves FID, SSIM, and PSNR.","key_machinery":"The load-bearing object is the pairwise Bell-state preparation step: Hadamard gates on the first half of the qubit register followed by CNOT gates that entangle each of those qubits with a corresponding qubit in the second half. The Bell pairs act as information pathways, so the subsequent parameterized quantum circuit (rotation gates and CNOTs) can be applied to a subset of the qubits and still process the whole encoded input. Amplitude encoding supplies the exponential compression, and the diffusion forward and reverse Markov chain supplies the training objective; together they turn the entanglement pattern into a parameter-saving generative model.","core_discovery":"The central claim is that pairwise Bell-state entanglement, inserted between amplitude encoding and the parameterized circuit, establishes information pathways that allow the circuit to operate on a reduced set of qubits while still accessing information from the entire input state. On the paper's account, this is what produces the 40-47% parameter reduction and the improved reconstruction metrics: the Bell pairs correlate each qubit in the first half of the register with a partner in the second half, so the trainable gates do not need to touch every qubit. The paper further claims that directly predicting the denoised data, rather than the noise component, works better empirically, and that EEQDM outperforms a quantum diffusion model without the entanglement strategy as well as a classical U-Net-based diffusion baseline, with the largest gains on 16x16 CIFAR-10 color images.","pith_inferences":["The paper leaves the readout step unspecified; every reported metric therefore presupposes a reconstruction rule that is never stated, and the claimed NISQ compatibility would fail if the experiments required full statevector access rather than finite-shot measurements.","A likely implicit trade-off is that the Bell-pair correlations store information in a way that does not respect spatial locality; the paper's own future-work suggestion about correlation-preserving encoding indicates that exploiting image structure would require a different pairing scheme.","A testable extension would be to compare the Bell-pair circuit against a full strongly-entangling circuit with the same reduced parameter count, isolating whether the entanglement pattern itself, rather than merely fewer parameters, drives the reported gains.","On real hardware, the same circuit should be run for 8x8 images and the measured outcome statistics compared against the simulator loss curves; the unstated shot count and observable would make this comparison decisive."],"forward_implications":["The reported parameter reduction means the model can represent a 256-pixel image with roughly 750 trainable parameters, an order of magnitude fewer than the classical U-Net baseline used in the comparison.","Because the trainable circuit touches only a subset of qubits, execution time scales more slowly with depth, which is why the paper observes EEQDM running about twice as fast as QDDM on 16x16 CIFAR-10 at higher depths.","The authors' direct-prediction choice (output the denoised data rather than the predicted noise) is presented as an empirical improvement, so future quantum diffusion implementations should consider it as a design option.","If the entanglement strategy delivers the claimed access to the full input, it can be combined with spatial-correlation-preserving encodings to push the same parameter savings to higher-resolution images."],"supporting_citations":[{"why":"Supplies the baseline quantum diffusion model whose parameter counts, execution times, and metrics EEQDM is directly compared against.","marker":"[6]"},{"why":"Supplies the prior intermediate-measurement approach that EEQDM extends and avoids by reducing parameters inside the quantum circuit rather than with a classical autoencoder.","marker":"[8]"},{"why":"Defines the denoising diffusion probabilistic model whose forward Markov chain and reverse-process structure the quantum implementation mirrors.","marker":"[3]"},{"why":"Defines the Fréchet Inception Distance used to compare the distribution of generated images with real images.","marker":"[9]"},{"why":"Defines the Structural Similarity Index used to measure reconstruction quality in the experiments.","marker":"[21]"},{"why":"Supplies the MNIST dataset used for the 8x8 and 16x16 image generation experiments.","marker":"[11]"},{"why":"Supplies the CIFAR-10 dataset used to test the model on more complex color images.","marker":"[12]"},{"why":"Supplies the amplitude-encoding construction that allows a full image to be represented by logarithmically many qubits, the basis of the claimed qubit savings.","marker":"[14]"},{"why":"Supplies the NISQ hardware context that motivates the demand for low parameter counts and shallow circuits.","marker":"[22]"}],"fun_headline_variants":["Bell-pair entanglement cuts quantum diffusion parameters by 40-47%","Pairwise Bell states reduce qubit needs in quantum diffusion","Entangle qubits in pairs to slash quantum diffusion costs","Quantum diffusion on NISQ improved with Bell-state pairing","Bell pairs enable high-dim image generation on few qubits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported results all depend on an unspecified readout: the paper never states which observable is measured, how many shots are taken, or how measurement outcomes are mapped back to a full image after amplitude encoding. If that mapping is not physically realizable, the central efficiency claim collapses.","fun_headline_variants_meta":{"raw":{"variants":["Bell-pair entanglement cuts quantum diffusion parameters by 40-47%","Pairwise Bell states reduce qubit needs in quantum diffusion","Entangle qubits in pairs to slash quantum diffusion costs","Quantum diffusion on NISQ improved with Bell-state pairing","Bell pairs enable high-dim image generation on few qubits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000589,"raw_usage":{"total_tokens":2727,"prompt_tokens":870,"completion_tokens":1857,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":486,"completion_tokens_details":{"reasoning_tokens":1773}},"tokens_in":486,"tokens_out":1857,"duration_ms":11886,"temperature":1.0,"reasoning_tokens":1773,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:40:38.684598+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Reproduce the reported numbers with an explicit measurement stage: amplitude-encode 16x16 MNIST into 8 qubits, apply the Bell-pair and parameterized circuit, and reconstruct all 256 pixel values using only computational-basis measurement outcomes with a finite shot budget. If the loss, SSIM, PSNR, and FID values cannot be matched unless the full statevector is accessed, the claim that EEQDM is a NISQ-ready generative model is wrong.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the denoising diffusion probabilistic model whose forward Markov chain and reverse-process structure the quantum implementation mirrors."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the Fréchet Inception Distance used to compare the distribution of generated images with real images."},{"cited_title":"C., Sheikh, H","cited_arxiv_id":null,"evidence_quote":"Defines the Structural Similarity Index used to measure reconstruction quality in the experiments."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the MNIST dataset used for the 8x8 and 16x16 image generation experiments."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the CIFAR-10 dataset used to test the model on more complex color images."},{"cited_title":"J., Bergholm, V., & Salomaa, M","cited_arxiv_id":null,"evidence_quote":"Supplies the amplitude-encoding construction that allows a full image to be represented by logarithmically many qubits, the basis of the claimed qubit savings."}],"review_version":1}