{"id":"dc0bceca-68d0-43fe-b697-35fcf5eba03d","arxiv_id":"2506.02441","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":2.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"A vacuum-noise QRNG based on homodyne detection is built and characterized, but the claimed 5 bits of quantum entropy per sample and the LFSR-based extraction are not adequately supported.","lead":"The authors built a quantum random number generator that measures vacuum noise with a balanced homodyne detector and tested the output with NIST and Dieharder suites. The result is a laboratory-scale implementation of an established method, but the specific claims about quantum entropy and the post-processing are not well supported.","discovery_kind":"replication","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The LFSR post-processing is a deterministic linear operation, so it cannot flatten the Gaussian vacuum-noise distribution or add entropy; the claimed 5-bit quantum contribution and certified randomness after extraction are therefore unsupported.","rationale":"The paper has genuine value as a survey, and the homodyne setup follows a standard architecture; the reported shot-noise scaling and statistical test results are consistent with random-looking output. However, the central claim is not merely that the device outputs statistically random bits, but that it produces certified quantum randomness from vacuum fluctuations at about 50 Mbps, with 5 bits of quantum entropy per sample. That claim depends on two conditions: the post-processing must be a valid randomness extractor, and the 5-bit min-entropy must be attributable to the quantum source. The paper's description of the LFSR as flattening a Gaussian distribution contradicts the linear-algebraic nature of an LFSR; no randomized seed or leftover-hash bound is supplied. Additionally, the NIST SP800-90B estimate is made on the total ADC output, and the paper gives no source model that separates quantum and classical contributions to the min-entropy. The conclusion itself concedes that higher-speed operation would require 'rigorous hashing and extracting algorithms like the Toeplitz function,' which suggests the present LFSR step is not being treated as a rigorous extractor. Without an LO-off control, passing NIST and Dieharder cannot certify quantum character, because classical electronic noise could be whitened by the same deterministic post-processing. This concern matches the reader's weakest assumption and reinforces the reader's REJECT verdict: the authors would need to supply a proper extractor analysis, a source-model quantum entropy bound, and LO-off control data to restore the central claim.","tokens_in":16840,"tokens_out":7778,"duration_ms":81990,"concrete_test":"Run a local-oscillator-off control: with the laser blocked, record the same 10 MSa/s 8-bit ADC data under otherwise identical conditions (electronic noise only), then apply exactly the same post-processing—keep the 5 MSBs, apply the same 63-bit LFSR serial transformation, and test the resulting bitstream with NIST SP800-90B, NIST STS, and Dieharder. If the LO-off classical-noise stream also flattens to near-uniform and passes the same suites with comparable min-entropy per sample, then the LFSR, not the vacuum source, produces the observed randomness and the 5-bit quantum claim collapses. If raw data cannot be released, the same test can be run on the described pipeline using synthetic Gaussian noise with the published electronic-noise variance; indistinguishable passing results would have the same implication.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section V states that 'the biased profile of quantum noise (Gaussian) is flattened by using Linear Feedback Shift Register (LFSR) as a potential extractor algorithm' and that after 5-MSB selection and serial LFSR the extracted data are 'equally probable in all bins and maximally unpredictable.' This is the load-bearing step, but no seed, tap polynomial, output mapping, or leftover-hash argument is given. A fixed LFSR is a deterministic linear map over GF(2): it cannot increase entropy, it cannot turn a Gaussian-shaped source distribution into a uniform one, and it cannot create fresh randomness. If the intended claim is that the LFSR is a cryptographic extractor, then the input must already have sufficient conditional min-entropy and the extractor must be seeded; neither condition is established. The NIST SP800-90B min-entropy estimate of about 5 bits per 8-bit sample is applied to the full raw ADC output, not to a quantum-conditioned source model, so discarding 3 bits and attributing the remaining 5 bits to vacuum fluctuations is unsupported. Because the apparent uniformity of the final stream is produced by a deterministic post-processing stage, passing NIST and Dieharder cannot certify the quantum origin of the bits. The absence of a control run with the local oscillator blocked leaves no way to distinguish quantum shot noise from classical electronic noise as the source of the extracted entropy.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reviews quantum random number generator approaches and presents an experimental vacuum-fluctuation QRNG based on balanced homodyne detection. It reports an SP800-90B min-entropy estimate of about 5 bits per 8-bit sample, attributes this entropy to quantum vacuum noise, applies a 63-bit LFSR as a 'potential extractor' after discarding 3 bits, and claims that the extracted stream passes the NIST, Dieharder, compression, and autocorrelation tests, yielding a certified quantum random bit rate of about 50 Mbps.","tokens_in":17141,"tokens_out":5449,"duration_ms":45068,"significance":"If the central claims were supported, the paper would offer a simple and accessible vacuum-fluctuation QRNG with a clear entropy assessment. The authors provide a broad literature overview, an indigenous balanced homodyne detector, and a systematic battery of statistical tests. However, the load-bearing steps—the LFSR extractor and the attribution of all estimated min-entropy to quantum noise—are not justified; the statistical tests cannot certify quantum origin. The paper's value as a standalone research contribution is therefore limited, though it may be useful as a tutorial-style account of building and testing a homodyne-based QRNG.","major_comments":[{"comment":"The manuscript states that 'the biased profile of quantum noise (Gaussian) is flattened by using LFSR as a potential extractor algorithm' and that the extracted counts are 'equally probable in all bins.' A 63-bit LFSR is a deterministic linear map over GF(2); it cannot increase entropy, cannot transform a Gaussian-shaped source distribution into a uniform one, and cannot generate fresh randomness. No seed, tap polynomial, output mapping, or leftover-hash argument is provided. Consequently, the uniformity of the processed stream and the passing of NIST and Dieharder tests do not certify the quantum origin or the unpredictability of the bits; at best they show that the post-processed stream resembles a pseudo-random sequence.","section":"Section V, Fig. 7"},{"comment":"The SP800-90B min-entropy estimate of about 5 bits per 8-bit sample is applied to the full raw ADC output, which comprises vacuum noise, electronic noise, and environmental noise. The paper offers no model that separates the quantum component from the classical components (e.g., a shot-noise calibration or a conditional-entropy calculation), so the statement that 'the contribution of quantum noise is 5 bits' (Section V) is unsupported. The same estimate is used both to set the extraction rate (discarding 3 bits) and to assert the quantum nature of the source, which renders the attribution circular.","section":"Section V.A"},{"comment":"The claim that 10 mW is the 'shot-noise limit' is not established. The variance-versus-power data are shown on a log plot, but the paper does not provide a quantitative fit to σ²_total = σ²_c + σ²_q, nor a comparison with the expected shot-noise scaling, nor a verification that the excess variance is free of classical intensity noise. The description of the electronic-noise measurement with the LO switched off is qualitative; a quantitative control experiment is required to support the claim that the added variance is quantum-limited.","section":"Section V, Fig. 9"},{"comment":"The claimed output rate of about 50 Mbps is not derivable from the stated parameters: 10 MSa/s × 8 bits/sample = 80 Mbps raw, and after discarding 3 bits and applying the LFSR the output rate depends on the exact post-processing mapping, which is not specified. The paper should state the number of output bits per input sample and the throughput of the LFSR stage, and it should clarify whether the 50 Mbps figure is measured or inferred.","section":"Section V.B and Conclusion"}],"minor_comments":[{"comment":"The phrase 'chona's circuit' should read 'Chua's circuit'.","section":"Section I.A"},{"comment":"The phrase 'Continues Wave' should read 'Continuous Wave'.","section":"Section IV"},{"comment":"The phrase 'Teoplitz function' should read 'Toeplitz function'.","section":"Conclusion"},{"comment":"The data availability statement says all data are presented graphically and therefore no data are associated, which contradicts the claimed 100 Mb NIST file and 50 GB Dieharder file; raw data or a link should be provided for reproducibility.","section":"Section VII"},{"comment":"The sentence 'The min-entropy estimator is based on the assumption that the probability distribution of the i.i.d. random variables is uniform' is incorrect; SP800-90B min-entropy estimators do not assume uniformity, and the following two sentences conflate the estimator with a uniformity assumption.","section":"Section V.A"},{"comment":"Reference [39] (multiphoton blockade) is not relevant to the claim about super-Poissonian scattering; please replace or remove it.","section":"References"},{"comment":"The captions of Figures 9, 12, and 13 do not fully specify the meaning of all curves, dots, and bars; each caption should identify every plotted quantity and the significance threshold used.","section":"Figures 9, 12, and 13"}],"recommendation":"reject","confidential_remarks":"This manuscript is a lab-report-style account with a broad review and a working experimental setup, but the scientific claims go beyond what the data and analysis support. The LFSR extractor misunderstanding and the unvalidated entropy attribution are fundamental flaws that cannot be fixed by local edits. I recommend rejection, though the authors could resubmit a substantially revised version that includes a proper source model, a seeded extractor with leftover-hash analysis, and a quantitative shot-noise calibration."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hi,\n\nHere's my take on 2506.02441. It's a vacuum-noise QRNG built from a balanced homodyne detector, an 852 nm laser, an ADC, and an LFSR post-processor. The physics of the source is standard; the paper is essentially a lab report. What it does well: it gives a thorough walkthrough of QRNG methods, explains min-entropy, describes the home-built 10 MHz BHD, and reports a variance-versus-LO-power curve that looks like real shot-noise behavior. The authors also ran NIST, Dieharder, and compression tests on the output. As a recipe for someone building a lab-scale QRNG, the paper has some value.\n\nThe problems are load-bearing. The authors claim the Gaussian raw noise is 'flattened' by a 63-bit LFSR used as an extractor, and that the result is 'maximally unpredictable.' An LFSR is a deterministic linear map over GF(2). It cannot increase entropy, cannot turn a Gaussian distribution into a uniform one, and cannot generate fresh randomness. If the final bits pass statistical tests, that only shows the LFSR output is a good-looking pseudo-random sequence derived from the raw samples; it does not certify the quantum origin. No seed, tap polynomial, or output mapping is provided, so the description is at best seriously incomplete.\n\nThe entropy claim has the same problem. SP800-90B gives about 5 bits of min-entropy per 8-bit sample for the full ADC output. The paper then states that 'the contribution of quantum noise is 5 bits.' But that estimator is agnostic about the noise source; it measures the total noise distribution, not the quantum component. Without a model that separates shot noise from electronic and environmental noise, the attribution is unsupported. They did measure electronic noise with the LO off, but they don't show that the final extracted bits are unpredictable even given that classical noise.\n\nThere is no raw data or code, only graphical summaries, so the NIST/Dieharder results can't be independently checked. The data availability statement says all data is in the figures.\n\nAll that said, the experiment is probably real, and the raw variance behavior is consistent with a working homodyne setup. The paper is not a waste of time; it's a useful demonstration that needs a proper entropy model and a correct post-processing description. For a serious venue, I'd want the authors to provide the LFSR details, a control experiment with LO blocked, or a proper argument for why the extracted bits contain quantum entropy. As is, the central claim is unsupported.\n\nMy recommendation: reject in current form, but send to a referee if the venue is willing to consider a revised lab-notes paper. I'd cite it only if the authors fix the extractor issue.","headline":"A workmanlike vacuum-noise QRNG lab report whose central claim—certified 5-bit quantum entropy per sample—is unsupported because the LFSR post-processing cannot do what the paper says and the entropy estimate does not separate quantum from classical noise.","tokens_in":17691,"tokens_out":3357,"would_cite":false,"duration_ms":31539,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A home-built balanced homodyne detector extracts 5 bits of quantum vacuum noise from each 8-bit sample, and the post-processed stream passes all 15 tests of one statistical suite and all 31 tests of another, at about 50 Mbps.","keywords":["quantum random number generation","vacuum fluctuations","balanced homodyne detection","min-entropy","LFSR extractor","statistical randomness tests","shot-noise limit","autocorrelation"],"falsifier":"With the local oscillator switched off, record the electronic noise alone and push it through the identical 8-bit sampling, 5-most-significant-bit selection, and 63-bit LFSR post-processing. If the LO-off stream has a per-sample min-entropy comparable to 5 bits and passes the same 15-plus-31 statistical tests, then the reported tests and entropy estimate do not distinguish quantum vacuum noise from classical electronic noise, and the central claim fails.","tokens_in":16639,"feed_emoji":"🎲","tokens_out":9953,"duration_ms":84657,"temperature":0.7,"pith_summary":"This paper reports a tabletop quantum random number generator that draws randomness from the quantum vacuum: an 852 nm local oscillator, a polarizing beam splitter whose unused port admits the vacuum, and a home-built 10 MHz balanced homodyne detector convert the vacuum's amplitude-quadrature fluctuations into a measurable current. Sampling that current at 8 bits per point, the authors estimate a worst-case min-entropy of 5 bits per sample, attribute that entropy to quantum vacuum noise, and then pass the extracted bit stream through a 63-bit linear feedback shift register used as an extractor. They report that the final stream passes all 15 tests of one standard statistical suite and all 31 tests of another battery, that autocorrelation drops to a few thousandths after one bit, that the file resists compression, and that the generation speed is about 50 Mbps. If true, the result matters because it offers a relatively simple QRNG that avoids single-photon detectors, interferometric delay lines, and complex calibration.","feed_headline":"Vacuum fluctuations produce 50 Mbps of random bits","feed_subtitle":"A balanced-homodyne setup extracts 5 quantum bits per 8-bit sample and passes all 46 statistical randomness tests.","key_machinery":"The balanced homodyne detector is the physical engine: two matched photodiodes receive the two outputs of a 50:50 beam splitter, their photocurrents are subtracted, and the difference is directly proportional to the vacuum amplitude quadrature, so the local oscillator amplifies vacuum-scale fluctuations above electronic noise. The post-processing chain is the statistical engine: 8-bit sampling, a min-entropy estimate of 5 bits per sample, retention of the 5 most significant bits, and application of a 63-bit linear feedback shift register (a shift register whose new bit is a linear function of previous bits) in serial as the claimed extractor.","core_discovery":"The central discovery is that the homodyne difference current, $\\Delta i = S\\hbar\\omega\\tau (2\\bar{\\alpha}_{\\mathrm{lo}})\\delta X_{1,\\mathrm{vac}}$, isolates the vacuum quadrature $\\delta X_{1,\\mathrm{vac}}$ on top of a fixed classical electronic background; the variance of the total noise grows with local-oscillator power until it saturates at about 10 mW, which the paper takes as the shot-noise limit. At that operating point the raw 8-bit samples have a Gaussian profile whose worst-case min-entropy is about 5 bits per 8-bit sample, so the authors treat 5 bits of each sample as the quantum contribution and discard the remaining 3 bits. Keeping the 5 most significant bits and applying a 63-bit LFSR in serial is claimed to flatten the Gaussian profile into an equiprobable distribution; the extracted stream then passes all 15 tests of the standard statistical suite and all 31 tests of the second battery on a 50 GB file, with autocorrelation reduced to a few thousandths after lag one. The paper's conclusion is that a certified, environment-robust QRNG with about 50 Mbps output can be built from these components, with a plausible upgrade path to 500 Mbps.","pith_inferences":["Editorial inference: because an LFSR is a linear reversible map, it cannot add entropy; the certified-randomness claim therefore rests on the 5-bit min-entropy measured before extraction, and a stronger certification would use a keyed universal hash rather than an LFSR.","Editorial inference: the same post-processing would plausibly make any stationary Gaussian electronic noise look random to the same test batteries; a decisive check is to run the full 8-bit-to-LFSR pipeline on the local-oscillator-off data and show that its per-sample min-entropy falls well below 5 bits.","Editorial inference: if the entropy estimate is correct, the device's 50 Mbps rate is an estimate of certified quantum randomness; the unextracted raw output, with its Gaussian bias and temporal correlations, should not be used directly."],"forward_implications":["A QRNG for cryptographic and simulation use can be assembled from a laser, a half-wave plate, a polarizing beam splitter, two photodiodes, and a 10 MHz difference circuit, with no single-photon detectors.","The 5-bit-per-8-bit-sample min-entropy estimate provides a conservative entropy budget: three bits per sample can be discarded in post-processing while preserving quantum randomness.","Operating at the 10 mW shot-noise-limit point maximizes the quantum contribution to the total noise variance, giving a simple calibration rule for similar setups.","The reported pass on all 15-plus-31 statistical tests means the output is statistically indistinguishable from ideal random bits on the tested file size, which is what applications require.","Raising the ADC sampling rate and replacing USB with PCI data transfer is claimed to raise output to about 500 Mbps, with stronger hashing needed once faster sampling adds more bias."],"supporting_citations":[{"why":"Provides the foundational vacuum-state QRNG result that this work builds on with a home-built 10 MHz detector.","marker":"[16]"},{"why":"Supplies the balanced-homodyne quadrature measurement method used for the vacuum-noise extraction.","marker":"[79]"},{"why":"Provides the quantum-optics relation that the measured variance saturates at the shot-noise limit, fixing the 10 mW operating point.","marker":"[80]"},{"why":"Prior vacuum-fluctuation QRNG whose variance-saturation curve is compared when setting the local-oscillator power.","marker":"[81]"},{"why":"Supports the estimate of extractable quantum entropy in vacuum-based QRNGs used to justify the 5-bit per-sample figure.","marker":"[82]"},{"why":"The entropy-assessment suite whose software gives the worst-case min-entropy of 5 bits per 8-bit sample.","marker":"[83]"},{"why":"The 15-test statistical battery used to certify the extracted bit stream.","marker":"[90]"},{"why":"The 31-test randomness battery reported as fully passed on a 50 GB file.","marker":"[91]"}],"fun_headline_variants":["Vacuum noise yields 50 Mbps of certified random bits","Homodyne vacuum probe passes 46 randomness tests","Shot-noise limit reached, 5 quantum bits per sample","Entropy from vacuum: 50 Mbps QRNG with LFSR","Quadrature vacuum fluctuations power a 50 Mbps QRNG"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that a serial 63-bit LFSR, a linear bit operation, actually flattens the Gaussian raw-noise distribution into a uniform one and so acts as a randomness extractor; since a linear map cannot increase entropy, if that flattening is not a genuine extraction, the test passes certify only the LFSR's own pseudorandom behavior, not quantum vacuum randomness.","fun_headline_variants_meta":{"raw":{"variants":["Vacuum noise yields 50 Mbps of certified random bits","Homodyne vacuum probe passes 46 randomness tests","Shot-noise limit reached, 5 quantum bits per sample","Entropy from vacuum: 50 Mbps QRNG with LFSR","Quadrature vacuum fluctuations power a 50 Mbps QRNG"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000469,"raw_usage":{"total_tokens":2346,"prompt_tokens":967,"completion_tokens":1379,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":583,"completion_tokens_details":{"reasoning_tokens":1295}},"tokens_in":583,"tokens_out":1379,"duration_ms":8406,"temperature":1.0,"reasoning_tokens":1295,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:24:16.429566+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"With the local oscillator switched off, record the electronic noise alone and push it through the identical 8-bit sampling, 5-most-significant-bit selection, and 63-bit LFSR post-processing. If the LO-off stream has a per-sample min-entropy comparable to 5 bits and passes the same 15-plus-31 statistical tests, then the reported tests and entropy estimate do not distinguish quantum vacuum noise from classical electronic noise, and the central claim fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The 31-test randomness battery reported as fully passed on a 50 GB file."},{"cited_title":"Gabriel, C","cited_arxiv_id":null,"evidence_quote":"Provides the foundational vacuum-state QRNG result that this work builds on with a home-built 10 MHz detector."},{"cited_title":"Zhang, F","cited_arxiv_id":null,"evidence_quote":"Supplies the balanced-homodyne quadrature measurement method used for the vacuum-noise extraction."},{"cited_title":"Gerry, P","cited_arxiv_id":null,"evidence_quote":"Provides the quantum-optics relation that the measured variance saturates at the shot-noise limit, fixing the 10 mW operating point."},{"cited_title":"Zheng, Y","cited_arxiv_id":null,"evidence_quote":"Prior vacuum-fluctuation QRNG whose variance-saturation curve is compared when setting the local-oscillator power."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supports the estimate of extractable quantum entropy in vacuum-based QRNGs used to justify the 5-bit per-sample figure."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The entropy-assessment suite whose software gives the worst-case min-entropy of 5 bits per 8-bit sample."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The 15-test statistical battery used to certify the extracted bit stream."}],"review_version":1}