{"id":"d4eb7037-cdaa-4999-8a5d-cacfa54f2cea","arxiv_id":"2608.01765","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Under uniform background noise, two-mode SPADE outperforms direct imaging for one-vs-two-source discrimination and approaches the quantum Chernoff limit in the sub-Rayleigh regime, as shown experimentally.","lead":"A reduced SPADE measurement using only the two lowest Hermite-Gaussian modes keeps near-quantum-limited error rates for deciding whether a dim source is single or double, even with background light at 11% of the signal per detector. The result suggests practical noise-robust superresolution for astronomy and sensing.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Experimental validation uses an Airy PSF while theory assumes a Gaussian PSF; the resulting modal overlaps and Chernoff exponents are uncomputed, so the claimed agreement is unsupported.","rationale":"The reader's weakest assumption is the uniform noise model. I find that assumption explicitly stated and the derivation internally consistent. The more operationally load-bearing issue is the mismatch between the Gaussian PSF used in all theoretical calculations and the Airy PSF produced in the experiment. The paper compares experimental data to Gaussian-based curves without recomputing for Airy, so the experimental confirmation does not validate the theory as presented. This concern does not overturn the theoretical claim under its stated assumptions, but it strengthens the need for the CONDITIONAL verdict already given. I therefore recommend UNCHANGED while noting the specific missing calculation.","tokens_in":11539,"tokens_out":25446,"duration_ms":279451,"concrete_test":"Recompute the Chernoff exponents for the Airy PSF (or the measured experimental PSF) for both reduced SPADE (with the two lowest-order HG modes) and DI, using the same uniform noise model and parameters (σ, d, b/ν). Compare the ordering SPADE vs DI and the value of ξ_P against the experimental data in Fig. 4. If the Airy-based SPADE curve still exceeds the DI curve and matches the measured ξ_P within experimental uncertainty, the concern is resolved. If not, the paper's experimental support fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central numerical comparison (Fig. 1) and the theoretical curves used for comparison (blue dashed line) are computed for a Gaussian PSF (Eq. 9) and HG modes matched to that Gaussian. The experiment (Sec. 3) produces an Airy PSF with σ = 115 μm and uses a holographic decomposition onto the two lowest-order HG modes. For an Airy PSF, the modal weights p_{0}, p_{1} are not given by Eq. (11); the overlaps differ. The paper does not derive or compute the Chernoff exponent for the Airy PSF, nor does it justify that the Gaussian-based theory applies to the experimental conditions. The claimed agreement between the extracted ξ_P and the Gaussian-theory curve may be coincidental, and the central claim that a two-mode SPADE outperforms DI may not hold for the actual PSF. Thus the experimental demonstration does not actually test the theoretical model.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript addresses the binary hypothesis testing problem of discriminating a single incoherent point source from two closely spaced incoherent point sources in the presence of uniform background noise. It derives an analytical classical Chernoff exponent for spatial-mode demultiplexing (SPADE) under a Poisson-noise model, proves monotonic decay of the exponent with noise, computes the corresponding quantum Chernoff bound, and compares SPADE with direct imaging. A reduced SPADE scheme using only the two lowest-order Hermite-Gaussian modes is claimed to outperform direct imaging in the sub-Rayleigh regime and to approach the quantum limit. An experiment using a DMD-generated source, an Airy PSF, a holographic mode sorter, and a CMOS camera with LED background is presented as validation.","tokens_in":11768,"tokens_out":20786,"duration_ms":232006,"significance":"If the calculations and comparisons were correct, the paper would provide a practically important result: a two-mode SPADE measurement that is robust to background noise and near quantum-optimal for one-versus-two source discrimination. The theoretical framework (Eq. 6), the monotonicity proof (Appendix A), and the numerical quantum-limit evaluation (Appendix B) are clearly presented and internally consistent. The experimental demonstration with digital holography is thoughtful. However, the validation and the quantitative comparison currently contain load-bearing gaps: the theory is Gaussian-PSF while the experiment is Airy-PSF, and the reduced SPADE exponent is computed with a formula that assumes a complete POVM. These issues must be resolved before the central claims are supported.","major_comments":[{"comment":"The theoretical curves in Fig. 1 and the comparison in Fig. 4 are computed for the Gaussian PSF (Eq. 9) with HG modes (Eq. 10) and modal probabilities (Eq. 11). The experiment in Sec. 3 produces an Airy PSF (sigma = 115 um) and decomposes onto two HG modes. For an Airy PSF, the projections p_{I,q}, p_{II,q} are not given by Eq. (11), and the paper does not compute the corresponding Chernoff exponent or show that the Gaussian result is a good approximation. The claimed agreement in Fig. 4 is therefore not a test of the theory. Please provide the Airy-PSF predictions (or a quantitative justification for replacing the Airy PSF by a Gaussian with sigma = 115 um) and compare them with the extracted exponents.","section":"Sec. 3, Eqs. (9)-(11), Fig. 4"},{"comment":"The expression in Eq. (6) is derived for a complete POVM with sum_q p_{alpha,q}=1. For the reduced two-mode SPADE, sum_{q=0,1} p_{I,q}=1 but sum_{q=0,1} p_{II,q}=1-exp(-d^2/16sigma^2)(1+d^2/(16sigma^2))<1 (approximately 0.973 at d=2sigma). The correct exponent for the observed two-mode counts is max_s [s P_{I,S}+(1-s)P_{II,S}+2b/nu - sum_{q=0,1}(p_{I,q}+b/nu)^s(p_{II,q}+b/nu)^{1-s}] with P_{alpha,S}=sum_{q in S} p_{alpha,q}, not Eq. (6). Using Eq. (6) overestimates the reduced SPADE exponent by an amount of order (1-P_{II,S}) at the optimal s, which is material at the exponent scales shown in Fig. 1. Moreover, Sec. 3 defines the experimental u as nu+2b, suggesting nu is the two-mode signal count, whereas the theory uses nu as the total signal photon number; this inconsistency must be resolved.","section":"Sec. 2, Eq. (6), Fig. 1"},{"comment":"The comparison with direct imaging fixes the DI detector at 1000 pixels of size a=4.6 um, and the advantage largely follows from the resulting 1000-fold noise penalty. The manuscript does not discuss how the DI exponent depends on pixel number/size or whether the choice is optimal; a DI with fewer, appropriately sized pixels may partly recover the performance. The conclusion that SPADE consistently outperforms DI across all source separations requires either an optimization over DI parameters or a statement of the assumed detector-array model and a demonstration that the advantage is robust to that model.","section":"Sec. 2, Fig. 1 caption"}],"minor_comments":[{"comment":"The text says the extracted exponent agrees with the 'blue dash-dotted line' in Fig. 1, but the SPADE curve in Fig. 1 is a blue dashed line, not dash-dotted. Please correct the caption/text and clarify which curves are shown in the bottom panel of Fig. 4.","section":"Sec. 3, Fig. 4"},{"comment":"The manuscript should clarify whether the 10^3 independent samples used for each value of k are disjoint or overlapping, and how the error bars on the extracted exponents are obtained. As written, the fits for different k appear to use the same data in a way that could introduce correlations.","section":"Sec. 3, Eq. (14)-(15)"},{"comment":"The experiment uses a CMOS camera rather than ideal photon counters. The text should state whether readout noise, dark counts, and the spatial uniformity of the LED background are included in the single parameter b and how the measured b for the two SPADE pixels relates to the per-pixel b assumed in the DI model.","section":"Sec. 3, Eq. (4)"},{"comment":"The abstract and conclusion state that SPADE outperforms DI 'across all source separations,' while the numerical evidence in Fig. 1 is for two representative sub-Rayleigh separations. Either show the full d-dependence or qualify the statement as 'in the sub-Rayleigh regime for the specific DI pixelation considered.'","section":"Sec. 1 / Conclusion"},{"comment":"The experiment uses a laser (coherent) source and time-alternating DMD mirrors to emulate two thermal incoherent sources. The manuscript should justify that the Poisson-limit Chernoff exponent derived for thermal states applies to this coherent-state implementation, or state the necessary assumptions explicitly.","section":"Sec. 3, source preparation"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the journal's scope and the central idea is interesting. The main issues are correctable in principle, but they affect the quantitative validity of both the theoretical comparison and the experimental validation. I recommend major revision rather than rejection because the conceptual framework and appendices are sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe one thing to know: the theory here is a genuine step forward, the experiment is not yet a clean test of it.\n\nWhat's new: the authors extend the SPADE hypothesis-testing literature to include uniform Poissonian background noise, giving a closed-form Chernoff exponent (Eq. 6) that is easy to evaluate for any measurement, and they compare two-mode SPADE against the quantum limit. The monotonicity proof in Appendix A is sound, and Appendix B gives a concrete numerical recipe for the quantum Chernoff bound. The intuition—SPADE concentrates signal in a few modes so it pays background noise only for those modes, while DI pays it for every pixel—is clearly articulated and is the right way to think about the problem. The two-mode truncation is justified in the sub-Rayleigh regime by the 0.97 concentration argument, and the experiment does show error probabilities falling exponentially with photon number.\n\nThe soft spots are real. The experimental validation uses an Airy PSF but the theoretical curves, including the blue dashed line used for the 'agreement', are computed for a Gaussian PSF. The modal overlaps for an Airy PSF are not Eq. (11); the paper never computes the Chernoff exponent for the actual PSF. So the claimed agreement is not actually demonstrated. The stress-test note is on target. Second, the abstract and conclusion say 'across all source separations,' but the analysis and data are sub-Rayleigh; for large d the two-mode SPADE will not capture the separation information, so that overclaims. Third, the DI baseline is a fixed 1000-pixel array with no optimization or sensitivity study, and the experiment reports no error bars on the extracted Chernoff exponents. The uniform-noise assumption is an idealization; if the background is spatially varying, the DI penalty may shrink, but as a first treatment this is acceptable.\n\nBottom line: this deserves a serious referee. The theoretical framework is useful and likely correct within its stated model; the experimental section needs a major revision—either match the PSF to the theory or compute the Airy-PSF predictions—before the experimental claims can stand.","headline":"Strong theory on noise-robust SPADE, but the Airy-PSF experiment doesn't test the Gaussian-PSF theory.","tokens_in":12215,"tokens_out":5143,"would_cite":true,"duration_ms":56648,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A reduced two-mode SPADE measurement beats direct imaging for discriminating one versus two incoherent sources under uniform Poissonian background noise, approaching the quantum limit in the sub-Rayleigh regime.","keywords":["spatial-mode demultiplexing","hypothesis testing","Chernoff exponent","incoherent point sources","Rayleigh's curse","background noise","quantum limit","direct imaging"],"falsifier":"Measure the Chernoff exponent for the same one-versus-two task with structured background light whose per-pixel mean is higher on the two SPADE mode pixels than on the rest of the camera, keeping the total background fixed; if direct imaging then matches or exceeds SPADE, the claimed robustness is tied to the uniformity assumption.","tokens_in":11458,"feed_emoji":"🔭","tokens_out":7686,"duration_ms":85733,"temperature":0.7,"pith_summary":"This paper asks whether the advantage of spatial-mode demultiplexing survives background noise in the practical task of deciding whether a faint image contains one point source or two. The authors model uniform Poissonian noise on every detector and show, theoretically and experimentally, that a reduced SPADE measurement reading only the two lowest Hermite-Gaussian modes achieves a higher Chernoff exponent than direct imaging for sub-Rayleigh separations, approaching the quantum limit. The reason is compression: SPADE puts almost all source information into two modes, so it pays background noise on two detectors, while direct imaging spreads the signal over many pixels and accumulates noise from each one. The experiments confirm the predicted error exponents at a background-to-signal ratio of 0.11. If the model holds, this gives a practical, noise-tolerant detection scheme for astronomy and quantum sensing.","feed_headline":"Two spatial modes beat direct imaging in noisy source detection","feed_subtitle":"With two Hermite-Gaussian modes, hypothesis testing nears the quantum error limit at 11 percent background.","key_machinery":"The load-bearing object is the reduced SPADE measurement: a projection of the image-plane field onto only the two lowest-order Hermite-Gaussian modes, |φ0⟩ and |φ1⟩, whose detection probabilities are given by Eq. (11). Its role is to concentrate nearly all signal information (more than 97% of photons for d≤2σ) into two ports, so that uniform background noise enters the Chernoff exponent only twice, whereas direct imaging's 1000-pixel readout adds noise from every pixel. The Chernoff exponent, the exponential decay rate of the error probability with mean photon number, is the figure of merit that carries the comparison; the paper's Eq. (6) is the classical limit of this exponent in the presen","core_discovery":"The paper's central claim is that, under uniform Poissonian excess noise, two-mode SPADE—projecting onto the zeroth and first Hermite-Gaussian modes only—discriminates a single incoherent source from two separated incoherent sources with an asymptotic error exponent that exceeds direct imaging for separations in the sub-Rayleigh regime and stays close to the quantum Chernoff bound. The analytical classical Chernoff exponent follows from Eq. (6), with per-detector mean photon numbers u_{α,q}=ν_{α,q}+b and equal total signal ν under both hypotheses; for SPADE the detection probabilities are p_I,q=δ_{q0} and p_II,q=(1/q!)(d/4σ)^{2q} exp(−d^2/16σ^2). The experiment implements this two-mode proje","pith_inferences":["The compression mechanism suggests any measurement that funnels nearly all source information into a few orthogonal ports should inherit the same noise robustness; other mode bases or few-pixel detectors may show the same effect.","The uniform-noise model is an idealization; spatially structured background—brighter near the sources or on the SPADE ports—would weaken the cumulative-noise penalty and could change the ordering, so the claimed all-separations advantage should be re-tested under structured noise.","A practical extension is to replace the camera readout of the two sorted modes with two bucket detectors, which would keep the same error exponent while simplifying the apparatus for field use."],"forward_implications":["In the sub-Rayleigh regime (source separation up to roughly two PSF widths), the two lowest Hermite-Gaussian modes capture more than 97 percent of the signal photons, so near-quantum discrimination needs only two detection ports.","Under uniform Poissonian background, SPADE's Chernoff exponent remains above direct imaging's for all separations studied, so its error probability decays faster with observation time or photon number.","At a background-to-signal ratio of 0.11 per detector, the experimentally extracted exponents match the noisy-SPADE theory and approach the quantum Chernoff bound.","Because the advantage comes from avoiding per-pixel noise accumulation, the benefit over direct imaging grows with the number of pixels in the imaging array.","The same two-mode counts can be used for a generalized likelihood-ratio test when the source separation is unknown, since the SPADE counts are highly sensitive to d."],"supporting_citations":[{"why":"Introduces SPADE and shows that low-order Hermite-Gaussian modes carry most of the separation information in the sub-Rayleigh regime, motivating the two-mode truncation.","marker":"[7]"},{"why":"Establishes the quantum-optimal binary SPADE scheme for one-versus-two incoherent sources, the noiseless benchmark this paper extends.","marker":"[28]"},{"why":"Supplies the Poisson-limit formula for the classical Chernoff exponent and intensity-operator bound used in Eqs. (3) and (18).","marker":"[45]"},{"why":"Provides the quantum Chernoff bound that the paper uses as the ultimate limit for the noisy discrimination problem.","marker":"[46–50]"},{"why":"Gives the digital-holographic SPADE implementation and the LED-based uniform background noise technique used in the experiment.","marker":"[13,52–54]"},{"why":"Defines the Chernoff exponent and the likelihood-ratio decision rule that converts measured counts into hypothesis choices.","marker":"[43,44]"}],"fun_headline_variants":["Two-mode SPADE beats direct imaging under noise","Two Hermite-Gauss modes approach quantum error limit","Reduced SPADE discrimination robust to background noise","Spatial-mode demux separates sources despite high noise","Two-mode SPADE nears quantum limit in noisy detection"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The paper's comparison assumes the background noise is Poissonian and uniform, with the same mean count per detector in every pixel or mode and no dependence on the signal; if background light is brighter on the two SPADE detectors than on the rest of the array, the advantage shrinks and can reverse.","fun_headline_variants_meta":{"raw":{"variants":["Two-mode SPADE beats direct imaging under noise","Two Hermite-Gauss modes approach quantum error limit","Reduced SPADE discrimination robust to background noise","Spatial-mode demux separates sources despite high noise","Two-mode SPADE nears quantum limit in noisy detection"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000677,"raw_usage":{"total_tokens":2893,"prompt_tokens":699,"completion_tokens":2194,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":443,"completion_tokens_details":{"reasoning_tokens":2128}},"tokens_in":443,"tokens_out":2194,"duration_ms":17536,"temperature":1.0,"reasoning_tokens":2128,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T21:22:36.927963+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the Chernoff exponent for the same one-versus-two task with structured background light whose per-pixel mean is higher on the two SPADE mode pixels than on the rest of the camera, keeping the total background fixed; if direct imaging then matches or exceeds SPADE, the claimed robustness is tied to the uniformity assumption.","supporting_citations":[{"cited_title":"Quantum-optimal detection of one-versus-two incoherent optical sources with arbitrary separation,","cited_arxiv_id":null,"evidence_quote":"Establishes the quantum-optimal binary SPADE scheme for one-versus-two incoherent sources, the noiseless benchmark this paper extends."},{"cited_title":"Poisson quantum information,","cited_arxiv_id":null,"evidence_quote":"Supplies the Poisson-limit formula for the classical Chernoff exponent and intensity-operator bound used in Eqs. (3) and (18)."}],"review_version":1}