{"id":"e6643164-7ec4-4fa8-b095-ea703f15d917","arxiv_id":"1909.00090","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"From Hubble parameter and Type Ia supernova data, Gaussian Process reconstruction places the cosmic deceleration-to-acceleration transition at z_t ≈ 0.59 and z_t ≈ 0.68, respectively.","lead":"Using a flexible statistical method called Gaussian Process regression, the authors reconstruct how fast the universe expanded at different times and find when cosmic acceleration began: around redshift 0.59 from Hubble parameter data and 0.68 from supernova data. The results support the standard picture that acceleration started relatively recently and can be estimated without assuming a particular cosmological model.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (28), used for the SNe Ia q(z) reconstruction, is algebraically inconsistent with the paper's own definitions and can produce a spurious transition redshift.","rationale":"The reader's weakest assumption concerned GP derivative bias; that is a legitimate issue, but Eq. (28) is a prior, purely algebraic problem. If the printed formula was actually used, the SNe result is invalid even with perfect GP derivatives. The H(z) result, based on Eqs. (13)-(14), is unaffected and may remain sound. The proposed analytic check on EdS is decisive because it requires no data or GP machinery, and the re-run with the corrected relation determines whether Table I changes materially. This does not impugn the authors' intent; it asks for a straightforward derivable correction before the SNe claim is accepted. If the code used the correct relation and Eq. (28) is only a typographical error, the concern reduces to a text correction, which the concrete test would reveal.","tokens_in":10972,"tokens_out":19677,"duration_ms":176106,"concrete_test":"Apply Eq. (28) to the analytic EdS luminosity distance D_L(z)=2(1+z)(1-1/sqrt(1+z)). The true deceleration parameter is q=+0.5 at every redshift, so any zero crossing of the printed q formula demonstrates that Eq. (28) is not the deceleration parameter. Then replace Eq. (28) in the analysis code with q_corr = 1 - (1+z)^2 D''_L / [(1+z)D'_L - D_L], using the same GaPP reconstructions and Pantheon covariance matrix, and compare the resulting z_t with Table I. If the shift exceeds the reported 1-sigma uncertainty, the published SNe result is driven by the algebraic error.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The SNe half of the central claim depends on Eq. (28), which does not follow from the definitions given in the paper. Starting from D_L=(1+z)D_C, E=1/D'_C, and q=-(1+z)D''_C/D'_C-1 (Eq. 22), one obtains q = 1 - (1+z)^2 D''_L / [(1+z)D'_L - D_L]. The printed Eq. (28), q = (1+z)^2 D''_L/D_L - (1+z)D'_L + 1, has the wrong denominator and an incorrect extra term. The same algebraic slip appears in the step from Eq. (23) to Eq. (24). Numerically, for a flat LambdaCDM luminosity distance at z=0.5, Eq. (28) gives q ~ +1.5, whereas the true q is ~ -0.11. The printed formula tends to cross zero where D_L is near 1 in units of c/H0, which is coincidental rather than a physical deceleration-to-acceleration transition. Applied to an Einstein-de Sitter universe, which has q=+0.5 at all z, Eq. (28) produces a spurious zero crossing near z~0.7. If the GaPP implementation used Eq. (28), the reported z_t=0.683 from Pantheon is not a measurement of the onset of acceleration, independent of any GP derivative bias.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses Gaussian Process (GP) reconstructions, implemented with the GaPP package, to determine the transition redshift z_t where cosmic expansion changes from deceleration to acceleration. Two data sets are used: 51 H(z) measurements and the Pantheon SNe Ia sample. From the H(z) reconstruction, the authors obtain z_t = 0.59^{+0.12}_{-0.11}; from the SNe Ia luminosity-distance reconstruction, assuming spatial flatness, they obtain z_t = 0.683^{+0.11}_{-0.082}. These results are reported for a squared-exponential kernel and are stated to be consistent with the Matern(5/2) and Matern(7/2) kernels. The paper claims that the GP method makes the analysis model-independent, apart from the flatness assumption for the SNe Ia part.","tokens_in":11270,"tokens_out":10374,"duration_ms":84335,"significance":"If correct, the paper would provide useful, model-independent cross-checks of the transition redshift from two independent cosmological probes, with the strength of using the full Pantheon covariance matrix and testing three kernel choices. The H(z) result is broadly consistent with earlier estimates in the literature. However, the SNe Ia half of the central claim rests on an algebraic relation for q(z) that is incorrect, and the lack of validation of the GP derivative reconstructions leaves the quoted uncertainties unsupported. The paper is therefore not acceptable in its present form, but the issues are local and fixable by re-deriving the formula and re-running the analysis.","major_comments":[{"comment":"The derivation of q(z) from the luminosity distance is algebraically incorrect. Starting from D_L=(1+z)D_C, E=1/D_C', and q=-(1+z)D_C''/D_C'-1 (Eq. 22), the correct relation is q = 1 - (1+z)^2 D_L'' / [(1+z)D_L' - D_L]. The printed Eq. (28), q = (1+z)^2 D_L''/D_L - (1+z)D_L' + 1, has the wrong denominator and an incorrect sign; the same slip appears in Eqs. (23) and (24). For a flat LambdaCDM model at z=0.5, Eq. (28) gives q approximately +0.3 while the true deceleration parameter is approximately -0.1, so the zero crossing of Eq. (28) is not a physical deceleration-to-acceleration transition. Since Section V.B states that the q(z) reconstruction is performed \"according to Eq. (28), the reported SNe Ia transition redshift in Table I is not a valid measurement as presented.","section":"III, Eqs. (23), (24), (28)"},{"comment":"The paper asserts that Monte Carlo sampling and analytic error propagation give \"negligible difference\" for the q(z) uncertainties, but it does not show the comparison, and the GP derivative reconstructions are not validated against simulations. Because z_t is read off as the zero crossing of q(z), a biased GP derivative H'(z) or D_L''(z) shifts z_t systematically, and the optimized length scale and sparse high-redshift data are known to affect GP derivatives. The authors should add a mock-data test (e.g., reconstructing q(z) from a flat LambdaCDM model with the same redshift sampling) to demonstrate that the reconstructed q(z) and its zero crossing are unbiased, and they should show the Monte Carlo versus error-propagation comparison for the H(z) analysis. This is necessary to justify the quoted 1-sigma confidence intervals.","section":"V.A and V.B"}],"minor_comments":[{"comment":"The sentence \"we found negligible difference with the error propagation (29)\" refers to Eq. (29), which is the SNe Ia uncertainty formula, not the H(z) uncertainty expression; the correct reference for the H(z) case is Eq. (14).","section":"V.A"},{"comment":"The Jacobian matrix in Eq. (27) is written as J = diag(sigma^2_DL), which is dimensionally inconsistent; the Jacobian should be J = diag(alpha D_Li). The final propagated variance formula is correct, but the notation should be fixed.","section":"III, Eq. (27)"},{"comment":"For the Matern(5/2) DL row, only one pair of upper/lower uncertainties is printed (0.83^{+0.25}_{-0.50}), whereas other rows list both 1-sigma and 2-sigma values; if 2-sigma values were intended, they should be provided.","section":"Table I"},{"comment":"There are small grammatical errors: \"decelerator parameter\" should be \"deceleration parameter,\" and \"module distance function\" should be \"luminosity distance function.\"","section":"Conclusion"},{"comment":"Expressions such as \"51 H(z) data\" should read \"51 H(z) data points\" for clarity.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The algebraic error in Eqs. (23), (24), and (28) is decisive and should be corrected before further consideration; if the corrected SNe Ia analysis changes z_t significantly, the paper will need a substantial re-analysis. The novelty is modest given existing GP-based transition-redshift estimates in the literature, but the H(z) result and the multi-kernel comparison are still of interest if the validation issues are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the score. The H(z) half of this paper is acceptable; the SNe Ia half is built on a wrong equation. Eq. (28), which the authors use to reconstruct q(z) from DL(z), does not follow from their own definitions. From DL=(1+z)DC and q=-(1+z)DC''/DC' - 1, the correct expression is q = 1 - (1+z)^2 D_L'' / [(1+z)D_L' - D_L]. The printed q = (1+z)^2 D_L''/D_L - (1+z)D_L' + 1 is different. For a flat ΛCDM model at z=0.5 the two give q ≈ -0.11 and q ≈ +1.5. The wrong one is what produces the claimed z_t = 0.683; it essentially crosses zero where D_L is near 1, an artifact. The same slip shows up in Eqs. (23) and (24). So I would not trust any of the Pantheon results in Table I.\n\nCredit where it's due: the GP implementation is clean, the three-kernel comparison is a good idea, and the H(z) result, z_t = 0.59, comes from the standard Eq. (13) and is consistent with Yu et al. and others. That's a small but valid update using the 51-point compilation.\n\nThe secondary issues: the authors claim Monte Carlo sampling and analytic error propagation give negligible differences but don't show it; this matters because second derivatives are unstable. They also treat the 51 H(z) points as independent without adjusting for known correlations in the BAO subset. And 'model-independent' is an overstatement for the SNe part, which assumes flatness. These are real but minor compared with the algebra.\n\nBottom line: this is a clear case for peer review, not desk rejection, because the H(z) result is worth keeping and the error is fixable. But the SNe section needs to be redone or dropped. I wouldn't cite it until that happens.","headline":"The SNe Ia half is built on an algebraic error in Eq. (28), producing a spurious transition redshift; the H(z) half is fine.","tokens_in":11804,"tokens_out":7659,"would_cite":false,"duration_ms":58202,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["98.80.-k","98.80.Es","95.36.+x"],"model":"deepseek-v4-flash","headline":"Reconstructing the deceleration parameter from Hubble and supernova data puts the transition redshift at 0.59–0.68 without assuming a cosmological model.","keywords":["transition redshift","deceleration parameter","Gaussian Process","H(z) data","Type Ia supernovae","model-independent cosmology","cosmic acceleration","kernel choice"],"falsifier":"Generate mock $H(z)$ and Type Ia supernova catalogs from a known flat Friedmann cosmology with a known transition redshift, apply the same Gaussian Process reconstruction, and check whether the recovered $z_t$ and its quoted uncertainty contain the input value; if the derivative reconstruction is biased at low redshift, the quoted model-independent $z_t$ would be systematically off.","tokens_in":10777,"feed_emoji":"🌌","tokens_out":11279,"duration_ms":93334,"temperature":0.7,"pith_summary":"The paper aims to measure the redshift at which the universe switched from deceleration to acceleration without assuming any cosmological model. It applies Gaussian Process regression, a non-parametric Bayesian technique, to 51 Hubble parameter measurements and to 1048 Type Ia supernova luminosity distances, reconstructing the deceleration parameter $q(z)$ in each case and locating its zero crossing. The H(z) data give $z_t = 0.59^{+0.12}_{-0.11}$; the supernova data, assuming spatial flatness, give $z_t = 0.683^{+0.11}_{-0.082}$. The two estimates agree and both are stable under alternative covariance kernels. These results matter because the transition redshift is a new cosmic parameter that a successful theory of cosmic acceleration must reproduce, and this is one of the few estimates that does not presuppose the theory.","feed_headline":"Two data sets pin cosmic acceleration's start near redshift 0.6","feed_subtitle":"Independent Hubble and supernova data agree on when expansion turned from deceleration to acceleration.","key_machinery":"The engine is Gaussian Process regression, a non-parametric Bayesian method that places a distribution over functions consistent with the data and can differentiate that distribution to yield covariances for $f'$, $f''$, etc. The paper uses three covariance kernels (a squared-exponential kernel and two smoother alternatives) to test kernel dependence, and propagates uncertainties through the derived expressions for $q(z)$: $q(z) = (1+z)H'(z)/H(z) - 1$ for the Hubble data and $q(z) = (1+z)^2 D''_L(z)/(D_L(z) - (1+z)D'_L(z)) + 1$ for the supernova distances. The zero-crossing condition $q(z_t) = 0$ defines the transition redshift.","core_discovery":"The central claim is that the transition redshift can be estimated directly from data. Reconstructing $H(z)$ and its first derivative from the 51-point H(z) compilation yields $q(z)$ through $q(z) = (1+z)H'(z)/H(z) - 1$, with a zero crossing at $z_t = 0.59^{+0.12}_{-0.11}$. Reconstructing the luminosity distance and its first two derivatives from 1048 supernovae, using $q(z) = (1+z)^2 D''_L(z)/(D_L(z) - (1+z)D'_L(z)) + 1$, gives $z_t = 0.683^{+0.11}_{-0.082}$ when spatial flatness is assumed. Both results are stable under the three covariance kernels, and the two probes agree within 1σ.","pith_inferences":["The quoted uncertainties are posterior errors from a single optimized kernel; systematic error from kernel choice and from the Gaussian Process prior mean is not included, so the full error budget is probably larger than reported.","The supernova reconstruction assumes spatial flatness. A joint Gaussian Process reconstruction of expansion and distance data could in principle relax that assumption and constrain curvature and $z_t$ simultaneously.","The consistency between the two probes could be sharpened with the distance-duality relation $D_L = (1+z)^2 D_A$: a future mismatch between H(z)-based and supernova-based $z_t$ would then point to new physics or systematics rather than merely different data."],"forward_implications":["The transition redshift is determined without reference to any dark-energy model, so a value near $z_t \\simeq 0.6$ becomes a direct empirical target that any theory of cosmic acceleration must reproduce.","The agreement between the H(z) and supernova results across three covariance kernels indicates that the estimate is not sensitive to the choice of kernel.","The supernova-based estimate inherits the spatial-flatness assumption; within a flat universe, the two probes together constrain when cosmic acceleration began.","The same Gaussian Process machinery extends to reconstruct other derived functions of $H(z)$ and $D_L(z)$, such as the dark-energy equation of state, in the same model-independent way."],"supporting_citations":[{"why":"Supplies the Gaussian Process regression method, including formulas for reconstructing a function and its derivatives with their covariances.","marker":"[30]"},{"why":"Provides the 51 H(z) measurements used for the H(z)-based transition redshift estimate.","marker":"[42]"},{"why":"Provides the 1048 Type Ia supernova distance measurements and their full covariance matrix used for the supernova-based estimate.","marker":"[59]"},{"why":"Gives an earlier Gaussian Process estimate of the transition redshift from H(z) data, used as a comparison.","marker":"[61]"},{"why":"Gives kinematic parametrization estimates of the transition redshift from combined supernova and H(z) data, used for comparison.","marker":"[27]"}],"fun_headline_variants":["Cosmic acceleration began at z≈0.6, says model-independent GP","GP on H(z) and SNe pins transition redshift at ~0.6","Two probes, one answer: cosmic acceleration started at z~0.6","Data-only GP finds cosmic deceleration-to-acceleration switch at z≈0.6"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The zero crossing of $q(z)$ is reliable only if the Gaussian Process derivative reconstructions $H'(z)$ and $D''_L(z)$ are unbiased, and the paper does not test this against simulated data; second derivatives from sparse supernova distances are especially sensitive to the kernel and the optimized length scale.","fun_headline_variants_meta":{"raw":{"variants":["Cosmic acceleration began at z≈0.6, says model-independent GP","GP on H(z) and SNe pins transition redshift at ~0.6","Two probes, one answer: cosmic acceleration started at z~0.6","Data-only GP finds cosmic deceleration-to-acceleration switch at z≈0.6"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000952,"raw_usage":{"total_tokens":3996,"prompt_tokens":819,"completion_tokens":3177,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":435,"completion_tokens_details":{"reasoning_tokens":3090}},"tokens_in":435,"tokens_out":3177,"duration_ms":492904,"temperature":1.0,"reasoning_tokens":3090,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T06:01:38.772971+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate mock $H(z)$ and Type Ia supernova catalogs from a known flat Friedmann cosmology with a known transition redshift, apply the same Gaussian Process reconstruction, and check whether the recovered $z_t$ and its quoted uncertainty contain the input value; if the derivative reconstruction is biased at low redshift, the quoted model-independent $z_t$ would be systematically off.","supporting_citations":[{"cited_title":"Seikel, C","cited_arxiv_id":null,"evidence_quote":"Supplies the Gaussian Process regression method, including formulas for reconstructing a function and its derivatives with their covariances."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives an earlier Gaussian Process estimate of the transition redshift from H(z) data, used as a comparison."},{"cited_title":"for DC(z),H(z) and q(z) parametrizations, respectively","cited_arxiv_id":null,"evidence_quote":"Gives kinematic parametrization estimates of the transition redshift from combined supernova and H(z) data, used for comparison."}],"review_version":1}