{"id":"fff3fa35-ba5a-4a21-aa5c-24bcbe9d9834","arxiv_id":"2412.09348","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Fitting a one-parameter modified gravity model to Pantheon+, DES, and BAO data gives mixed evidence: the n=6 version is preferred by information criteria in most combinations, but n=4 and n=10 are often disfavored, and the abstract overstates the result.","lead":"This paper tests a modified gravity theory in which matter and spacetime curvature couple directly, using the latest supernova and galaxy clustering data. It claims the data favor this theory over the standard cosmological model, but the evidence is weaker and more mixed than the headline suggests.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The abstract's evidence claim is contradicted by Table 2: only the selected n=6 variant is preferred, with weak-to-moderate BIC differences and mixed AIC signs; n=4 and n=10 are frequently disfavoured, so the conclusion rests on post-hoc model selection and exact Planck anchors.","rationale":"The paper performs a standard MCMC analysis with a conventional sampler, reports posteriors and fit statistics, and the numerical method follows the authors' prior work. Those are genuine independent supports for the parameter constraints themselves. The central claim, however, is a model-selection statement, and that is where the argument is weakest. The reader's stated weakest assumption (early-time decoupling) is not my primary concern: for the fitted Rn values, f2=(Rn/R)^n is extremely small at recombination, and the use of Planck-based early-time anchors is internally plausible. The decisive problem is the logical step from Tables 1-2 to the abstract. The tables show that one selected variant (n=6) is mildly preferred by BIC for all combinations, but not strongly for any combination, and the AIC evidence is mixed. Meanwhile two of the three considered variants are frequently disfavoured. The abstract's 'moderate to strong evidence ... for all dataset combinations' therefore appears to select the most favourable n after the fits and then present that as evidence for the broad NMC theory. A correct discrete-model comparison, or a Bayesian model average over n, would almost certainly reduce the evidence, and propagating Planck anchor uncertainties would further erode the BIC advantage. This does not invalidate the parameter constraints, but it does mean the headline conclusion is overstated. The recommended verdict remains CONDITIONAL, so I do not change the reader's verdict.","tokens_in":18020,"tokens_out":7417,"duration_ms":79721,"concrete_test":"Reproduce Tables 1-2 with n treated as a discrete model index: (a) include one additional model-selection penalty (e.g. ln(3) in BIC, or Bayesian model averaging over n=4, 6, 10), and (b) marginalize over the quoted Planck uncertainties on Omega*_m, H*_0 and rd in Eq. 9 and Section 3.3 instead of fixing them at their central values. If the maximum NMC evidence across all five dataset combinations is no longer DeltaBIC<-5 for every combination, the abstract's 'moderate to strong ... for all dataset combinations' claim should be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the inference from Table 2 to the abstract's conclusion. Table 2 does not show 'moderate to strong evidence ... for all dataset combinations' for the nonminimally coupled theory as such. For n=4, the NMC model is disfavoured by BIC relative to Lambda-CDM for PS+DESI (DeltaBIC=+7.12) and PS+eBOSS (+11.23); for n=10 it is disfavoured (DeltaBIC>+9) in every combination. Only the n=6 variant is preferred for all combinations, and its DeltaBIC values are -1.25 (PS), -8.84 (PS+DESI), -7.83 (PS+eBOSS), -5.31 (DESYR5+DESI), -5.62 (DESYR5+eBOSS): weak to moderate, never strong. The AIC signs for n=6 are +4.69, -3.40, -2.39, +0.13, -0.18, i.e. mixed. The preference thus rests on (i) the analysis singling out n=6 after comparing n=4, 6 and 10 without a multiple-model penalty, and (ii) the BIC parameter-count advantage obtained by fixing Omega*_m, H*_0 from Planck and rd=147.46 Mpc as exact inputs, without propagating their quoted uncertainties. If those anchors are treated as uncertain, the effective k difference between NMC and Lambda-CDM shrinks and the BIC advantage is reduced. The early-time decoupling assumption itself is plausible (for fitted Rn~3-5x10^-7 Mpc^-2, f2 is tiny at recombination), but even granting it, the published tables do not support the advertised strength or universality of the evidence.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper compares a nonminimally coupled curvature-matter gravity model, with f2(R)=(Rn/R)^n and f1=R, to flat ΛCDM using Pantheon+SH0ES supernovae (PS), DES-SN5YR supernovae, and DESI and eBOSS BAO data. The NMC model is initialized with Planck ΛCDM parameters at high redshift, leaving Rn (and, for PS, MB) as the fitted parameter; n is fixed to 4, 6, or 10. The authors compute χ2, AIC, and BIC model-comparison statistics and report best-fit H0 and Ωm values. They conclude that current cosmological data suggest a preference for the NMC theory, with 'moderate to strong evidence' for all dataset combinations, while acknowledging an unresolved BAO-related tension.","tokens_in":18455,"tokens_out":6690,"duration_ms":67171,"significance":"If the statistical claim were supported, the paper would be significant: a one-parameter extension of GR would simultaneously remove the cosmological constant, reproduce SNIa distances, and relax the Hubble tension. The work is also commendably transparent: it uses recent public datasets, calibrates the ΛCDM baseline against published Pantheon+ results, quotes convergence criteria, and provides parameter tables. However, the headline conclusion is not supported by the paper's own Table 2, and the evidence for the NMC model is at most weak-to-moderate and only for the post-hoc selected n=6 variant. The core value lies in the parameter constraints and in demonstrating that n=10 is strongly ruled out, which gives the framework falsifiable content.","major_comments":[{"comment":"The abstract and conclusions state that there is 'moderate to strong evidence' for a preference of the nonminimally coupled theory 'for all dataset combinations.' Table 2 contradicts this. For n=4, the NMC model is disfavoured by both AIC and BIC for PS+DESI (ΔAIC=12.56, ΔBIC=7.12) and PS+eBOSS (ΔAIC=16.67, ΔBIC=11.23); for n=10, it is strongly disfavoured in every combination (ΔBIC > 9). Only the n=6 variant is preferred, with ΔBIC values between -1.25 and -8.84 and mixed ΔAIC signs (+4.69, -3.40, -2.39, +0.13, -0.18). The data therefore support, at most, a weak-to-moderate preference for the n=6 variant, not for the nonminimally coupled theory as a whole, and not with consistent evidence from both information criteria.","section":"Section 4.2 and Section 5, Table 2"},{"comment":"Table 2 lists for the PS+eBOSS rows the same χ2 values as the PS+DESI rows for n=4 (1575.92) and n=10 (1581.60), but Table 1 gives χ2=1574.26 and χ2=1570.45 for those models, respectively. The reported ΔAIC and ΔBIC values for these two rows are therefore internally inconsistent with Table 1. For example, using Table 1, the PS+eBOSS n=10 χ2 difference from ΛCDM is 13.20, which would give ΔBIC ≈ 5.76 rather than the reported 16.91. Since these rows are used to argue that n=4 is disfavoured when BAO data are added, the corrected values must be recomputed and may materially weaken that specific conclusion.","section":"Table 2 versus Table 1"},{"comment":"The authors compare n=4, 6, and 10 and then single out n=6 as the preferred representative of the NMC model, but n is not included in the model-comparison statistics. No trials factor, prior over integer n, or marginalization over n is applied. This makes the reported evidence for the NMC framework a post-hoc selection from three exponents. A proper treatment of this model-selection step is required before the 'preference for the NMC theory' conclusion can be drawn; the raw ΔBIC values for n=6 overstate the evidence for the framework as a whole.","section":"Section 4.2, choice of n=6"},{"comment":"The NMC model is initialized with Planck ΛCDM values H*0=67.4 km/s/Mpc and Ω*m=0.315, and the Planck-calibrated sound horizon rd=147.46 Mpc is used as an exact input. The quoted uncertainties on these anchors (δH0=0.5 km/s/Mpc, δΩm=0.007, δrd=0.28 Mpc) are never propagated into the analysis. This is load-bearing for the BIC comparison because the NMC model is assigned k=1 while the anchors effectively carry information that should either be marginalized over or included as informative priors, both of which would reduce the parsimony advantage that drives much of the reported BIC preference. The paper should also demonstrate quantitatively that f2(R) is negligible at recombination for the fitted Rn values, for example by evaluating (Rn/R)^n at z≈1100, since the claimed consistency with Planck rests on this decoupling assumption.","section":"Section 2.3, Section 3.3, Section 4.2"}],"minor_comments":[{"comment":"There are several typographical issues, including 'E ffectively' at the start of Section 4 and 'overperforming' in Section 4.2; these should be corrected.","section":"Throughout"},{"comment":"The sentence 'this model’s fit to the PS sample is weakly disfavoured ... when considering their respective AIC values' understates the sign inconsistency in Table 2: for PS, n=6 has ΔAIC=+4.69, which is a weak-to-moderate disfavour by the cited criteria, while the BIC gives no clear preference; the text should be more precise about this mixed evidence.","section":"Section 4.2"},{"comment":"The paper does not state the number of data points N used in each BIC calculation. Since the BIC penalty scales with ln N and the PS and DESYR5 samples have different sizes, listing N for each dataset combination would improve reproducibility.","section":"Section 3"},{"comment":"The comparison of the NMC model with the DESI BAO constraints in Ref. [31] could be made more explicit, since that reference also studies a nonminimally coupled gravity model against DESI data and would help contextualize the present results.","section":"Section 4.1"}],"recommendation":"major_revision","confidential_remarks":"This is a borderline case. The underlying numerical pipeline and the use of current datasets are credible, and the paper contains useful parameter constraints. However, the central statistical claim is overstated, and Table 2 contains internally inconsistent entries that affect the conclusions. The manuscript could become publishable after correcting the tables, addressing the model-selection issue for n, propagating or justifying the Planck anchors, and substantially revising the abstract and conclusions to match the actual strength of the evidence. If the authors can show that the n=6 preference survives these corrections, the paper would be a useful contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my read of the Varela–Bertolami NMC paper.\n\nThe genuinely new content: they run their established nonminimally coupled curvature-matter model against DES-SN5YR, DESI BAO and eBOSS, and they add the n=6 case. None of that appeared in their earlier work, and the resulting parameter constraints for Rn and derived H0, Omega_m are useful. The analysis is well done: cobaya MCMC, proper covariance matrices, and a sensible calibration check (their LCDM fit to Pantheon+ matches the original). They are also upfront that the model does not resolve the BAO tension, which is honest.\n\nThe soft spot is the headline. The abstract and conclusions claim 'moderate to strong evidence for a preference of the nonminimally coupled theory ... for all dataset combinations.' That is not what Table 2 shows. n=10 is strongly disfavored by both AIC and BIC in every combination. n=4 is disfavored by BIC for PS+DESI and PS+eBOSS. Only n=6 is preferred, and its ΔBIC values are between -1.25 and -8.84, i.e. weak to moderate, with mixed signs in ΔAIC. So the evidence is neither strong nor universal. It rests on choosing n=6 after comparing three exponents, without a multiple-model penalty, and on treating Planck-derived anchors (Omega_m*, H0*, rd) as exact numbers. If the uncertainties on rd and the Planck parameters were propagated, the effective parameter count of the NMC model would rise and the BIC advantage would shrink. The authors should either soften the abstract or do the extra work of marginalizing over the anchors.\n\nNone of this is fatal. The model is what it is, the data handling looks clean, and the constraints are a legitimate contribution. What needs fixing is the inference from Table 2 to the conclusion. I'd send it to peer review—the referee should push for a revised abstract and a discussion of the post-hoc n=6 selection and the fixed-anchor assumption, but the paper deserves engagement rather than desk rejection.\n\nFor a reading group, it's a decent example of how model-selection claims can outrun the tables. I'd bring it if we want to talk about BIC vs AIC and fixed external parameters.","headline":"New constraints from DESY5 and DESI/eBOSS are useful and the analysis is clean, but the abstract's claim of 'moderate to strong evidence ... for all dataset combinations' is contradicted by the paper's own Table 2.","tokens_in":18992,"tokens_out":3104,"would_cite":true,"duration_ms":32619,"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":"This paper claims that current cosmological data favor a nonminimal coupling between matter and curvature over flat $\\Lambda$CDM.","keywords":["Modified gravity","Nonminimally coupled gravity","Cosmological data","Hubble tension","dark energy","baryon acoustic oscillations","type Ia supernovae","model selection"],"falsifier":"Measure the BAO sound horizon from early-universe physics without assuming any late-time cosmology and check whether it equals $r_d = 147.46 \\pm 0.28$ Mpc; a significant deviation would remove the anchor on which the NMC model's BAO predictions rest.","tokens_in":17785,"feed_emoji":"🌌","tokens_out":13931,"duration_ms":118702,"temperature":0.7,"pith_summary":"This paper asks whether current cosmological data point to a modified theory of gravity in which matter is coupled directly to spacetime curvature, rather than coupled minimally as in general relativity. It compares a one-parameter nonminimal coupling model, with no cosmological constant, against flat $\\Lambda$CDM using type Ia supernova distances, Cepheid-calibrated distance anchors, and baryon acoustic oscillation measurements from several surveys. The authors report moderate to strong statistical evidence, by AIC and BIC criteria, that at least one version of the nonminimally coupled theory is preferred over $\\Lambda$CDM for every dataset combination they consider. If the result holds, a single extra interaction term in the gravitational action could account for the late-time acceleration attributed to dark energy and ease the Hubble tension, although a mismatch with BAO data remains.","feed_headline":"Cosmological data prefer matter-curvature coupling over ΛCDM","feed_subtitle":"A one-parameter theory with no cosmological constant beats ΛCDM on several independent dataset combinations.","key_machinery":"The central object is the action $S = \\int d^4x \\, \\sqrt{-g} \\left[(1/2\\kappa^2)R + (1+f_2(R))\\mathcal{L}_m\\right]$, with $\\mathcal{L}_m=-\\rho$ and $f_2(R)=(R_n/R)^n$. The inverse power law makes the coupling negligible at high curvature and increasingly significant at low curvature, so CMB-era $\\Lambda$CDM values can be used as fixed initial conditions and the theory has a single free parameter, $R_n$. The modified Friedmann and Raychaudhuri equations are integrated numerically using $F_2 \\tilde{\\rho}$ as the dynamical variable, and the resulting expansion history $H(z)$ is scored against the data with $\\chi^2$, AIC, and BIC statistics.","core_discovery":"The central claim is that current cosmological data suggest the presence of a nonminimal coupling of matter and curvature over the minimally coupled standard theory. In the paper's own terms, the model with $f_2(R)=(R_n/R)^n$ and $n=4,6,10$ fits Cepheid-calibrated supernova distances as well as or better than $\\Lambda$CDM while returning a Hubble constant consistent with both CMB-era initial conditions and local distance-ladder measurements, and when BAO data are added the $n=6$ version is preferred over $\\Lambda$CDM with moderate to strong evidence. The $n=10$ version is strongly disfavoured by both information criteria, and the $n=4$ version fits supernovae alone well but degrades when BAO data are included. The authors conclude that the data as a whole favour the nonminimally coupled theory, while explicitly acknowledging an unresolved BAO tension.","pith_inferences":["A natural extension the paper does not carry out is treating $n$ as a free parameter: the results suggest $n=6$ is the compromise between the smooth $n=4$ and sharp $n=10$ behaviours, so allowing $n$ to vary could sharpen the comparison at the cost of a second parameter.","Because the NMC model is anchored to $\\Lambda$CDM before recombination, a direct search for modified-gravity effects in CMB temperature and polarization spectra would test the foundation of the comparison; none is performed here.","The remaining BAO tension might be relieved by adding positive powers of $R$ to $f_2$, which act at early times; the paper calls this remote given how well CMB data match $\\Lambda$CDM, but it is a concrete modification to test.","Independent late-time probes of the expansion rate, such as gravitational-wave standard sirens, could confirm or challenge the predicted $H_0$ without relying on the Cepheid or BAO assumptions used here."],"forward_implications":["If the preference holds, late-time cosmic acceleration can be reproduced without a cosmological constant, because the inverse-power coupling mimics dark energy at low curvature.","The model bridges part of the Hubble tension: with CMB-anchored early conditions and a late-time deviation, its fitted $H_0$ stays compatible with Cepheid-calibrated supernova distances.","A BAO tension persists: adding baryon acoustic oscillation data pulls $H_0$ down toward $69$ km/s/Mpc, several sigma away from the supernova-only value, so the theory is not yet coherent across all probes.","The $n=10$ version is strongly ruled out by both AIC and BIC, while the $n=6$ version is the most consistently preferred variant when BAO data are included.","Ongoing and future BAO and supernova surveys can discriminate the model from $\\Lambda$CDM because its distance-redshift relation is fixed by a single parameter."],"supporting_citations":[{"why":"Introduces the inverse power-law coupling $f_2=(R_n/R)^n$ and motivates the $n=4$ and $n=10$ versions.","marker":"[29]"},{"why":"Establishes the numerical evolution method and the earlier Hubble-tension analysis that this work extends.","marker":"[30]"},{"why":"Supplies the CMB-era $\\Lambda$CDM parameters used to fix the model's high-redshift initial conditions.","marker":"[32]"},{"why":"Supplies the sound-horizon value $r_d=147.46 \\pm 0.28$ Mpc used to anchor all BAO predictions.","marker":"[52]"},{"why":"Supplies the Cepheid-calibrated supernova distances and covariance used for the first dataset combination.","marker":"[33]"},{"why":"Supplies the Cepheid distance calibration used to anchor the supernova absolute magnitude in the first dataset combination.","marker":"[15]"},{"why":"Provides the model-comparison methodology, including the $\\chi^2$ marginalization over absolute magnitude, adopted here.","marker":"[8]"},{"why":"Provides the BAO measurements used in the first spectroscopic survey dataset combination.","marker":"[11]"},{"why":"Provides the BAO likelihood used in the second spectroscopic survey dataset combination.","marker":"[10]"},{"why":"Supplies the supernova sample and covariance used in the later five-year supernova dataset combinations.","marker":"[50]"}],"fun_headline_variants":["Cosmic data favor matter–curvature coupling over ΛCDM","Nonminimal matter–gravity coupling gains ground in surveys","A simple gravity tweak outperforms ΛCDM in data fit","Survey evidence supports a matter–curvature interaction","Matter–curvature coupling gets a boost from cosmic data"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison assumes the nonminimal coupling fully switches off at high redshift, so early-universe $\\Lambda$CDM values for the matter density, expansion rate, and BAO sound horizon can be used as fixed anchors; if the coupling still matters at the CMB epoch, the model is fitted with biased inputs.","fun_headline_variants_meta":{"raw":{"variants":["Cosmic data favor matter–curvature coupling over ΛCDM","Nonminimal matter–gravity coupling gains ground in surveys","A simple gravity tweak outperforms ΛCDM in data fit","Survey evidence supports a matter–curvature interaction","Matter–curvature coupling gets a boost from cosmic data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001655,"raw_usage":{"total_tokens":6546,"prompt_tokens":895,"completion_tokens":5651,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":511,"completion_tokens_details":{"reasoning_tokens":5564}},"tokens_in":511,"tokens_out":5651,"duration_ms":44759,"temperature":1.0,"reasoning_tokens":5564,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T17:05:04.334813+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the BAO sound horizon from early-universe physics without assuming any late-time cosmology and check whether it equals $r_d = 147.46 \\pm 0.28$ Mpc; a significant deviation would remove the anchor on which the NMC model's BAO predictions rest.","supporting_citations":[],"review_version":1}