{"id":"48dc2a6c-8406-4d49-b7ea-34dcad9b8b82","arxiv_id":"2411.14154","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Using BOSS/eBOSS and DESI DR1 BAO data plus cosmic chronometers, the authors obtain Omega_K = -0.040 (+0.142 / -0.145) with Gaussian-process reconstruction, consistent with a flat universe.","lead":"This paper measures whether the universe is spatially flat using galaxy-clustering BAO data and galaxy-age (cosmic chronometer) data, avoiding assumptions about the sound horizon and the Hubble constant. It finds a flat universe within measurement uncertainty, though the precision is far lower than CMB-based estimates.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (8) covariance misses off-diagonal correlations in the GP/ANN reconstructed H(z); the quoted ΩK uncertainty and precision gain rest on this matrix, so a full-covariance rerun is required.","rationale":"The reader identified the incomplete covariance treatment as the weakest assumption, and I agree this is the most load-bearing concern for the paper's central quantitative claims. The method's novelty rests on the precision of ΩK and on the independence from rd and H0; the sd cancellation is algebraic, but the H0-independence claim, while under-supported, is unlikely to change the result substantially because the curvature constraint is weak enough that the H0-sensitivity is small relative to the quoted errors. The covariance issue, by contrast, directly controls whether the reported error bars and the claimed precision gain are real. If the off-diagonal correlations of the reconstructed H(z) are ignored, the effective number of independent data points is overestimated, typically leading to underestimated parameter uncertainties and potentially biased central values. The manuscript does not provide the full covariance matrix or a robustness check, so the current numbers in Table II cannot be taken at face value. That said, the central conclusion that the universe is spatially flat within uncertainties is likely to survive an enlarged error bar, which is why the conditional acceptance verdict remains appropriate rather than a rejection. The concrete test I propose—recomputing the likelihood with the full GP/ANN covariance—would settle the matter directly.","tokens_in":14051,"tokens_out":13216,"duration_ms":124198,"concrete_test":"For the GP+DESI+eBOSS case, reproduce the GaPP reconstruction and extract the full 9×9 predictive covariance matrix of H(z) at the nine BAO redshifts (including off-diagonal elements). Propagate this covariance through Eqs. (2), (5), and (7) to form the complete covariance of ΔD, including Cov(obs,obs), Cov(mod,mod), and the two cross terms, and rerun the emcee fit for ΩK. Compare the resulting 1σ interval and best fit with Table II; if the interval widens by more than 30% or the central value shifts by more than 0.05, the quoted precision is not robust. The same test should be repeated using the ANN posterior ensemble.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section II.C defines the likelihood in Eq. (8) with Cov = Covstat_ii + Covcorr_ij, but the covariance of the residuals ΔD_i = [DA]BAO+CC(zi) − [DA]CC(zi) is not fully specified. Both terms are constructed from the same reconstructed H_CC(z): [DA]BAO+CC uses H_CC at each BAO redshift via Eq. (2), and [DA]CC uses the integral DC(z) of H_CC via Eqs. (5) and (7). Therefore Cov(ΔD_i, ΔD_j) includes Cov([DA]obs_i, [DA]obs_j) and Cov([DA]mod_i, [DA]mod_j), which are nonzero because the GP/ANN posterior correlations of H_CC(z) across z are strong (the GP kernel has a finite length scale). The paper adds only diagonal statistical errors and a generic cross-term, without demonstrating that off-diagonal terms are negligible. If these correlations are positive, the data are less independent than assumed, so the reported 1σ intervals (ΩK = −0.040+0.142−0.145 for GP) are likely underestimated and the claim that the precision 'surpasses recent measurements' (Sec. IV) is not established. The central value could also shift when the full matrix is used, since the likelihood weighting changes. This is the load-bearing point for the quantitative headline, even if the qualitative flat-universe conclusion may survive.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a method to measure the cosmic curvature Omega_K using BAO observations from BOSS/eBOSS and DESI DR1 together with cosmic-chronometer H(z) data, without calibrating the BAO sound horizon and without assuming a specific cosmological model. The method reconstructs H(z) from cosmic chronometers with Gaussian process (GP) and artificial neural network (ANN) techniques, uses the BAO transverse and line-of-sight ratios to obtain absolute angular diameter distances via Eq. (2), and fits Omega_K through Eq. (8) with the FLRW distance relation. The authors report Omega_K = -0.040^{+0.142}_{-0.145} (GP) and Omega_K = -0.010^{+0.405}_{-0.424} (ANN) for the combined BOSS/eBOSS plus DESI DR1 sample, concluding that the late universe is spatially flat within 1 sigma and that the GP constraint is more precise than recent similar measurements.","tokens_in":14411,"tokens_out":9166,"duration_ms":85740,"significance":"If correct, this work provides a late-universe curvature measurement that is independent of the sound-horizon scale, H0, and any assumed dark-energy or cosmological model, which makes it a useful cross-check of Planck-based flatness conclusions. The central flatness result is likely robust: the reported central values are close to zero and the quoted uncertainties are large enough that even a moderate increase in the error budget would not move the constraint away from flatness at the 1-sigma level. The paper also makes a constructive comparison between GP and ANN reconstruction methods and shows that the two give consistent central values. The main limitation is that the quantitative precision claim rests on a covariance treatment that appears incomplete, so the specific improvement over previous work is not yet established.","major_comments":[{"comment":"The covariance matrix used in the likelihood does not propagate the off-diagonal correlations of the reconstructed H(z). Both [DA]BAO+CC in Eq. (2) and [DA]CC through DC(z) in Eqs. (5)-(7) are constructed from the same GP or ANN posterior for H_CC(z). Because the Matérn kernel in Eq. (4) has a finite correlation length, Cov([DA]BAO+CC_i, [DA]BAO+CC_j) and Cov([DA]CC_i, [DA]CC_j) are nonzero for i != j, and there are also cross-correlations between the two distance estimates. The paper adds only diagonal statistical errors (Covstat_ii) plus a local cross-covariance term Covcorr_ij, which as written is not the full covariance of the data vector Delta D. If the omitted terms are positive, the quoted 1-sigma intervals (e.g., Omega_K = -0.040^{+0.142}_{-0.145} in Sec. III) are underestimated, and the claim in Sec. IV that the precision surpasses recent measurements is not established. Please propagate the full GP/ANN posterior covariance into Eq. (8), including the covariance of the integrated DC(z), or provide a mock-based demonstration that the neglected off-diagonal correlations are negligible.","section":"II.C, Eq. (8)"},{"comment":"The statement that the result is independent of H0 is supported only by an undocumented test. Since [DA]BAO+CC in Eq. (2) has no explicit H0 dependence while [DA]CC in Eq. (7) depends on H0 through the distance formula, varying H0 as a free parameter while keeping the reconstructed H_CC(z) fixed will in general change Delta D and hence the inferred Omega_K. The cancellation the authors describe requires that H0 and the entire reconstructed H_CC(z) curve rescale together, which is not a property of the likelihood as written. Please provide either an explicit derivation of the invariance or a figure showing that the Omega_K posterior is stable over a wide range of adopted H0 priors.","section":"II.C (H0-independence claim)"},{"comment":"The BAO measurements are treated as independent: Table I quotes only diagonal errors for DM/rd and DH/rd. Published BOSS/eBOSS and DESI DR1 analyses provide full covariance matrices that include correlations between redshift bins and between the transverse and line-of-sight distance measurements. Because the precision claim in Sec. IV depends on the total error budget, the authors should either incorporate the full BAO covariance matrices or justify quantitatively that the off-diagonal elements are negligible for the combination considered here.","section":"II.A and Table I"}],"minor_comments":[{"comment":"The text in Sec. III says the precision 'is not exceptionally high compared to previous work,' but Sec. IV states that the GP precision 'surpasses recent measurements'; these statements should be reconciled with a direct comparison of the quoted error bars.","section":"Sec. III vs Sec. IV"},{"comment":"The ANN combined result is quoted with an upper error of 0.424 in the Abstract and 0.427 in Table II; the numbers should be made consistent.","section":"Abstract and Table II"},{"comment":"The text describes the kernel as the 'squared exponential form,' but Eq. (4) is the Matérn covariance with nu = 9/2; please correct the wording.","section":"Eq. (4) and surrounding text"},{"comment":"The phrase 'the distance duality relation relation' has a duplicated word and should be corrected.","section":"II.C"},{"comment":"The ANN architecture choice (single hidden layer with 4096 neurons) is justified only by reference to a code repository; a brief discussion of why this architecture is appropriate for the present data would improve reproducibility.","section":"II.B (ANN)"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of a cosmology journal and the core idea is sound, but the quantitative claims rest on a covariance treatment that is incomplete. The most important request is to rerun the analysis with the full covariance of the reconstructed H(z) propagated into Eq. (8), or to justify with mocks that the off-diagonal terms are negligible. The H0-independence claim also needs explicit support. If these points are addressed, the paper could be suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nQuick take: this is a competent, incremental application of an already-existing model-independent curvature test to the new DESI DR1 BAO data, with a GP/ANN comparison on top. The central result — Omega_K consistent with flat at about 0.14 precision — is probably right, but I would not trust the quoted error bars until the covariance treatment is fixed.\n\nWhat is genuinely new: the method itself (combining DM/DH from BAO with H(z) from cosmic chronometers to cancel the sound horizon) has been in the literature for years, as the authors acknowledge. The contribution here is the DESI DR1 numbers, the GP-versus-ANN comparison, and the observation that the two BAO samples give nearly identical precision. That is a useful data point, not a breakthrough.\n\nWhat the paper does well: the reconstruction procedure is standard and uses published code; they are careful not to extrapolate H(z) beyond the CC redshift range; and the central values are stable across methods and data splits. The flat-universe conclusion is robust to most plausible tweaks because the error bars are large.\n\nThe soft spots, in order. First, the covariance matrix in Eq. (8) is load-bearing, and it does not capture the off-diagonal correlations in the reconstructed H(z). Both [DA]_BAO+CC and the integrated [DC]_CC are built from the same GP/ANN curve, so their fluctuations are correlated across redshift. The paper adds only diagonal statistical errors plus a generic cross-term; nothing demonstrates the off-diagonals are negligible. Since the claim to surpass earlier work is a precision claim, this matters. The central value could shift too, though I doubt it moves much. Second, the statement that H0 has no effect is asserted but not shown. Looking at the equations, H0 only enters the model DA, not the observed BAO+CC distance, so it should matter unless marginalized; the paper provides no test or plot. It may be a small effect, but \"does not have any effect\" needs evidence. Third, the comparison of precision with refs [23,71-73,75] is made without accounting for different covariance treatments, so it may not be apples-to-apples.\n\nThe method is not circular in the sense of assuming flatness; the curvature is the fitted parameter. The covariance issue is real but fixable, and the qualitative conclusion (flat at 1-sigma) is likely to survive.\n\nBottom line: worth refereeing. Send it to a competent cosmologist with a request to see a full-covariance rerun and a proper H0 sensitivity check. I would not cite it in my own work, but it is a reasonable data extension for a reading group.\n\nBest,\n[You]","headline":"Incremental but honest application of an existing sound-horizon-free curvature test to DESI DR1; the central flat-universe result is plausible, but the quoted precision rests on an incomplete covariance treatment.","tokens_in":14905,"tokens_out":3590,"would_cite":false,"duration_ms":33024,"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 model-independent, sound-horizon-free and $H_0$-free curvature measurement finds the universe spatially flat within $1\\sigma$.","keywords":["cosmic curvature","baryon acoustic oscillations","cosmic chronometers","Gaussian process","artificial neural network","spatial flatness","model-independent cosmology","DESI DR1"],"falsifier":"Recompute the $\\Omega_K$ posterior propagating the full covariance of the reconstructed $H(z)$, including off-diagonal terms between different redshifts, rather than the diagonal-plus-local approximation in Eq. (8); if the error bars widen beyond roughly $0.14$ in the Gaussian-process case, the paper's precision claim would be undermined. A complementary check is to run the same pipeline on mock BAO and cosmic-chronometer catalogs with known nonzero $\\Omega_K$ and see whether the $1\\sigma$ intervals recover the input value at the expected rate.","tokens_in":13854,"feed_emoji":"🌌","tokens_out":16612,"duration_ms":129073,"temperature":0.7,"pith_summary":"The paper's goal is a measurement of the spatial curvature of the universe that does not lean on any cosmological model, on the sound-horizon ruler $r_d$, or on the Hubble constant $H_0$. It combines transverse and line-of-sight BAO measurements from BOSS/eBOSS and DESI DR1 with Hubble-parameter values from cosmic chronometers, reconstructing the expansion history twice: once with Gaussian processes and once with artificial neural networks. When all BAO data are combined, the Gaussian-process pipeline gives $\\Omega_K=-0.040^{+0.142}_{-0.145}$ and the ANN pipeline gives $\\Omega_K=-0.010^{+0.405}_{-0.424}$, both consistent with a spatially flat universe at $1\\sigma$. A sympathetic reader should care because this is one of the few curvature constraints that is independent of early-universe physics and of model-dependent assumptions, so it can test the standard cosmological model from late-universe geometry alone.","feed_headline":"BAO + cosmic-chronometer data: flat universe at 1 sigma, no H0","feed_subtitle":"Ω_K = −0.040 ± 0.14 from BOSS/eBOSS + DESI DR1 with GP, no sound-horizon or H0 prior.","key_machinery":"The central object is the ratio identity in Eq. (2): $[D_A(z)]_{\\mathrm{BAO+CC}} = \\frac{c}{(1+z)[H(z)]_{\\mathrm{CC}}} \\left(\\frac{D_M/r_d}{D_H/r_d}\\right)_{\\mathrm{BAO}}$, in which $r_d$ disappears. This converts BAO measurements into absolute distances using only the cosmic-chronometer $H(z)$ reconstruction. To build that reconstruction, the paper uses the 32 CC $H(z)$ points with two non-parametric methods: Gaussian processes with a Matérn($\\nu=9/2$) kernel, and an artificial neural network. The curvature then enters through the FLRW distance relation Eq. (7), and the fit is performed by maximizing the likelihood in Eq. (8), whose covariance matrix includes the reconstruction errors and a local cross-covariance term between the two distance estimators.","core_discovery":"On its own terms, the paper establishes that the sound horizon cancels exactly if one divides the transverse BAO distance $D_M/r_d$ by the line-of-sight BAO distance $D_H/r_d$, leaving $D_M/D_H$; multiplying by the cosmic-chronometer Hubble parameter converts this into an absolute angular diameter distance $[D_A]_{\\mathrm{BAO+CC}}$. Comparing these distances with the curved-geometry relation $D_A(z;\\Omega_K)$ obtained from the reconstructed comoving distance yields a one-parameter likelihood for $\\Omega_K$. The combined BOSS/eBOSS plus DESI DR1 sample gives $\\Omega_K=-0.040^{+0.142}_{-0.145}$ with Gaussian-process reconstruction, the tightest result in the paper ($\\Delta\\Omega_K\\simeq0.14$); the ANN reconstruction gives a consistent but much looser $-0.010^{+0.405}_{-0.424}$. The paper reads this as evidence that the late universe is spatially flat within $1\\sigma$, and that the two data sources and two reconstruction methods agree in their central values even though their precisions differ.","pith_inferences":["The paper does not propagate the full off-diagonal covariance of the reconstructed $H(z)$; if that covariance is sizable, the claimed precision advantage over earlier work could shrink.","The sign difference between the DESI (slightly positive) and BOSS/eBOSS (slightly negative) central values might reflect per-sample systematics rather than cosmic geometry; combining the samples could hide such systematics.","The same ratio construction could be applied to future high-redshift BAO measurements to test whether the flat conclusion persists beyond the current $z\\simeq2.3$ range.","An independent cross-check would be to replace the cosmic-chronometer $H(z)$ with model-independent $H(z)$ from other clocks, such as strong-lensing time delays, to test whether the reconstruction method drives the result."],"forward_implications":["If the result holds, a flat late-universe geometry is established without assuming $\\Lambda$CDM, using only BAO and clock measurements.","The cancellation of $r_d$ gives future BAO surveys a way to report absolute distances without waiting for a model-dependent sound-horizon calibration.","Because $H_0$ cancels out of the ratio, the curvature constraint would not shift if the Hubble tension is resolved by changing the early-universe $H_0$.","The roughly $\\sqrt{N}$ improvement in precision means ongoing DESI observations and more cosmic-chronometer data should tighten $\\Omega_K$ toward the few-percent level.","The method needs no distance-duality assumption, so any future inconsistency between this curvature value and one from standard candles would point to new physics or systematics."],"supporting_citations":[{"why":"Supplies the DESI DR1 BAO measurements of $D_M/r_d$ and $D_H/r_d$ that feed Eq. (2).","marker":"[47]"},{"why":"Supplies the BOSS/eBOSS galaxy BAO measurements used as the second data source.","marker":"[31]"},{"why":"Supplies the eBOSS Ly-$\\alpha$ auto-correlation BAO point at $z\\simeq2.33$ used in the combined sample.","marker":"[43]"},{"why":"Supplies the eBOSS Ly-$\\alpha$--quasar cross-correlation BAO point at $z\\simeq2.34$.","marker":"[44]"},{"why":"Provides the 32 cosmic-chronometer $H(z)$ measurements from which the GP and ANN reconstructions are built.","marker":"[52–57]"},{"why":"Provides the artificial neural network reconstruction algorithm used for the $H(z)$ curve.","marker":"[63]"},{"why":"Motivates the Matérn($\\nu=9/2$) covariance kernel used in the Gaussian-process reconstruction.","marker":"[70]"},{"why":"Provides the MCMC sampler used to explore the $\\Omega_K$ likelihood.","marker":"[69]"},{"why":"Gives a CMB-based flat-universe value used as a consistency comparison for the final $\\Omega_K$.","marker":"[3]"}],"fun_headline_variants":["Curvature from BAO without sound horizon: Ω_K= -0.04 ± 0.14","Flat universe at 1σ from BAO without sound horizon or H0","Ω_K = −0.04 ± 0.14: late universe flat without H0 prior","BAO+CC curvature: Ω_K= -0.04 ± 0.14, no sound horizon","Sound-horizon-free BAO: Ω_K= -0.04 ± 0.14, flat universe"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"That the reconstructed $H(z)$ curve from cosmic chronometers is accurate everywhere in the fitted redshift range, and that correlations between its error bars at different redshifts are small enough to ignore when computing the quoted $\\Omega_K$ uncertainties.","fun_headline_variants_meta":{"raw":{"variants":["Curvature from BAO without sound horizon: Ω_K= -0.04 ± 0.14","Flat universe at 1σ from BAO without sound horizon or H0","Ω_K = −0.04 ± 0.14: late universe flat without H0 prior","BAO+CC curvature: Ω_K= -0.04 ± 0.14, no sound horizon","Sound-horizon-free BAO: Ω_K= -0.04 ± 0.14, flat universe"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001043,"raw_usage":{"total_tokens":4453,"prompt_tokens":1080,"completion_tokens":3373,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":696,"completion_tokens_details":{"reasoning_tokens":3245}},"tokens_in":696,"tokens_out":3373,"duration_ms":22436,"temperature":1.0,"reasoning_tokens":3245,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:29:11.278187+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the $\\Omega_K$ posterior propagating the full covariance of the reconstructed $H(z)$, including off-diagonal terms between different redshifts, rather than the diagonal-plus-local approximation in Eq. (8); if the error bars widen beyond roughly $0.14$ in the Gaussian-process case, the paper's precision claim would be undermined. A complementary check is to run the same pipeline on mock BAO and cosmic-chronometer catalogs with known nonzero $\\Omega_K$ and see whether the $1\\sigma$ intervals recover the input value at the expected rate.","supporting_citations":[],"review_version":1}