{"id":"e305d339-3d9d-412e-acfa-61e7b6bd5ab3","arxiv_id":"2411.16046","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Under a one-parameter thawing dark energy model, DESI BAO data barely moves the equation of state and the preference over ΛCDM is weaker than in the CPL model.","lead":"This paper tests whether the DESI survey's hint that dark energy changes over time depends on the model cosmologists use to describe dark energy. It finds that a slowly changing dark energy model shows much weaker evidence for change, suggesting the DESI conclusion is not robust.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Compressed CMB likelihood is validated only for ΛCDM and CPL, not for SSLCPL, so the SSLCPL constraints that drive the model-dependence conclusion are not yet protected against CMB-compression bias.","rationale":"The reader identified two concerns: BAO/H(z) double counting and compressed Planck not validated for SSLCPL. I think the second is the load-bearing one. The 26 radial BAO H(z) points (refs. 39–49) come from pre-DESI surveys (BOSS/WiggleZ, etc.), while the 'BAO' dataset is DESI DR1 (ref. 6), so the specific double-counting scenario is not supported by the reference list. The compressed-CMB gap, by contrast, is an internal hole: the paper itself validates P18 versus full Planck only in Table I for ΛCDM and CPL, not SSLCPL. Since the headline comparison is between CPL and SSLCPL, an unvalidated likelihood for SSLCPL is the weakest load-bearing point. A full-Planck rerun for SSLCPL would settle whether the reduced tension and ΔAIC values are robust. My recommended verdict remains CONDITIONAL, matching the reader, so no verdict adjustment is needed.","tokens_in":12559,"tokens_out":11273,"duration_ms":109996,"concrete_test":"Re-run the flat and non-flat SSLCPL fits for the datasets that drive the claim (BAO+P18+H+D5, BAO+P18+H+U3, and the corresponding BAO− variants) replacing the compressed P18 likelihood with the full Planck 2018 likelihood (TTTEEE+lowE, e.g., via Cobaya or MontePython), keeping H(z), SNe Ia, and DESI BAO likelihoods unchanged. Compare the resulting w0 marginalized means and 68% intervals and ΔAIC values with Table III; if w0 shifts by more than ~0.05 or ΔAIC changes by more than ~2 units, the compressed-P18 approximation is not safe for SSLCPL and the model-dependence claim would need re-analysis.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on the SSLCPL rows of Tables II/III and Figures 3–5. Those rows use the compressed Planck likelihood P18 (shift parameter R, acoustic scale θ*, ωb; Section II A). Section III validates P18 against full Planck spectra only for ΛCDM, flat CPL, and CPL+Ωk (Table I); the sentence 'These results confirm that we can use the three compressed data points to represent the full power spectra' is not supported for SSLCPL. In SSLCPL the early-time equation of state is not Λ: w(a)→w0+wa at high z, with wa tied to w0 and Ωφ0 by Eq. (15). Distance priors calibrated on Planck data under ΛCDM/CPL-like expansion may be biased for this one-parameter family, and the bias would feed directly into the w0 posteriors and ΔAIC values (e.g., −6 for BAO+P18+H+D5) that constitute the paper's evidence for reduced tension. Because no external check of the compressed approximation is provided for the very model used for the headline comparison, the model-dependence conclusion is conditional on the compressed CMB approximation being accurate for SSLCPL.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper assesses whether the DESI BAO evidence for dynamical dark energy is robust to the choice of dark-energy parameterization. Using MCMC fits to DESI BAO (with and without the z=0.51 point), compressed Planck distance priors, H(z) data, and three SN Ia compilations, the authors compare flat and non-flat CPL models with the single-parameter SSLCPL thawing model. They find that adding DESI BAO has a small effect on the SSLCPL w0 constraint, that the tension with ΛCDM is reduced relative to CPL, and that spatial curvature has little impact. The central claim is that the DESI dynamical-dark-energy signal is model-dependent rather than a robust model-independent result.","tokens_in":12793,"tokens_out":4744,"duration_ms":47107,"significance":"The question addressed is timely and relevant: the DESI 2024 BAO results have been interpreted as evidence for evolving dark energy, and this paper argues that this evidence weakens in a physically motivated one-parameter thawing model. If the result holds, it is a useful counterweight to CPL-based claims and would strengthen the case that DESI BAO data alone do not yet single out dynamical dark energy. The analysis uses standard likelihoods and public MCMC machinery, and the paper reports complete posterior tables and ΔAIC values for many data combinations, which is helpful. However, the two load-bearing data-handling assumptions described in the major comments—the independence of H(z) radial BAO points from DESI BAO, and the validity of the compressed Planck likelihood for the SSLCPL model—are neither justified nor tested. Because both assumptions enter directly into the SSLCPL rows of Tables II and III that drive the conclusion, the paper in its current form is conditional.","major_comments":[{"comment":"The H(z) compilation used in every combined dataset includes 26 radial BAO measurements (refs. [39–49], covering z ~ 0.07–2.36). These are combined with the DESI BAO data in Section II B without any discussion of possible double counting or of correlations between the two BAO datasets. If the radial-BAO H(z) points at overlapping redshifts and the DESI BAO measurements are not independent, the reported χ2_min and ΔAIC values in Tables II and III (e.g., ΔAIC = −6 for SSLCPL with BAO+P18+H+D5) would be biased, and the central CPL-versus-SSLCPL comparison could shift. Please test this explicitly, for example by dropping the 26 radial BAO points from H(z) or by constructing a joint covariance, and report how ΔAIC and the w0 posteriors change.","section":"Section II A / II B"},{"comment":"The compressed Planck likelihood P18 is validated against the full Planck spectra only for ΛCDM, flat CPL, and CPL+Ωk. The SSLCPL model that carries the paper's main conclusion has a different early-time equation of state: Eq. (15) combined with w(a)=w0+wa(1−a) gives w(a→0)=w0+wa, which is not −1 at the best-fit values. Therefore the statement in Section III that 'These results confirm that we can use the three compressed data points to represent the full power spectra' is not supported for SSLCPL, and the SSLCPL entries in Tables II and III and Figures 3–5 rest on an unvalidated approximation. Please validate the compressed P18 likelihood for SSLCPL (for instance by comparing with a Planck full-likelihood fit for a representative SSLCPL case) or, if that is not feasible, explicitly present the conclusions as conditional on the compressed-CMB approximation being unbiased for SSLCPL.","section":"Section III / Table I"}],"minor_comments":[{"comment":"The word 'dengeneracy' should be 'degeneracy' in the sentence describing the CPL parameterization.","section":"Section I"},{"comment":"The phrase 'transverse comoving angular diameter distance' should be simplified to 'transverse comoving distance' DM(z).","section":"Section II A"},{"comment":"The AIC criterion sentence 'If −6 ≤ ∆AIC, then the evidence in favor of the model is positive' is missing an upper bound; as written it overlaps with the following 'If −10 ≤ ∆AIC' sentence and is internally inconsistent.","section":"Section II B"},{"comment":"The rows labeled 'DESI' are copied from Ref. [6], not fits performed in this paper; the caption should state this more explicitly so that the comparison between compressed and full-spectra results is not confused with a comparison between datasets.","section":"Table I"},{"comment":"The phrase 'non-flat flat CPL model' appears twice in the conclusion and should be corrected to 'non-flat CPL model'.","section":"Section IV"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of the journal and addresses a currently debated question. The two major issues—the untreated overlap between radial-BAO H(z) points and DESI BAO, and the unvalidated compressed-CMB likelihood for SSLCPL—are technical rather than philosophical, and both can be addressed with additional robustness checks. If those checks confirm the reported ΔAIC values, I would be happy to see the paper accepted; in its present form the headline model-dependence conclusion is conditional on assumptions that need to be stated and tested."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe short version: this paper does a clean, useful thing—it re-weights the DESI dynamical dark energy evidence using the SSLCPL parametrization, a one-parameter thawing model, and shows that the tension with ΛCDM is reduced compared to CPL. That message is not brand new; model-dependence of the DESI claim is already in the literature (e.g., Cortês & Liddle, Shlivko & Steinhardt, Park et al.). What's new is the specific application to DESI BAO data, across three SNe samples, with AIC comparisons and posterior shifts. That is a legitimate extension.\n\nThe analysis is standard: emcee MCMC with compressed Planck (R, θ*, ωb), DESI BAO, H(z), and three SNe compilations. The tables are internally consistent, and the paper checks the influence of the z=0.51 point and of curvature. The SSLCPL model is physically motivated, and the self-citation is appropriate since the model is their own.\n\nThe soft spots, in order of importance:\n\n1. The compressed Planck likelihood is validated against full Planck only for ΛCDM, flat CPL, and CPL+Ωk. The SSLCPL rows that drive the conclusion are not validated. In SSLCPL the early-time equation of state can move away from Λ, and while dark energy is negligible at recombination so the effect may be small, it has to be shown. This is a real caveat, not a fatal one.\n\n2. The H(z) compilation includes 26 radial BAO points, and the DESI BAO set is added without any discussion of independence or possible overlap. They are different surveys, so the issue may be minor, but it should be stated.\n\n3. Sampling details (priors, convergence, chain lengths) are missing, which makes reproduction harder but does not weaken the qualitative conclusion.\n\nNone of this kills the paper. The central claim—that the DESI dynamical dark energy evidence is model-dependent—holds, as a conditional statement. It is a modest, honest extension of work that is already out there.\n\nThis is for cosmologists interpreting DESI results and for people working on dark energy parametrizations. I'd send it to peer review; the referee should ask for a compressed-CMB validation for SSLCPL and a note on BAO overlap. Worth a serious read, though not an urgent one.","headline":"A useful model-dependence check on DESI's dynamical dark energy evidence, but the headline comparison rests on a compressed CMB likelihood not validated for the model that drives the conclusion.","tokens_in":13349,"tokens_out":4474,"would_cite":false,"duration_ms":41006,"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":"Fitting DESI BAO data with a one-parameter thawing-field model largely removes the preference for dynamical dark energy that appears in the two-parameter CPL fit.","keywords":["dynamical dark energy","DESI BAO","CPL parameterization","SSLCPL model","thawing scalar fields","Akaike information criterion","model selection","spatial curvature"],"falsifier":"Recompute the CPL and SSLCPL fits with the 26 radial BAO points removed from the $H(z)$ set whenever DESI BAO is included in the likelihood; if the SSLCPL $\\Delta\\mathrm{AIC}$ then drops below $-10$ or the DESI-induced shift in $w_0$ becomes as large as in CPL, the conclusion that DESI BAO has little influence in SSLCPL would be refuted.","tokens_in":12351,"feed_emoji":"🌌","tokens_out":12692,"duration_ms":99059,"temperature":0.7,"pith_summary":"The paper asks whether the DESI baryon acoustic oscillation (BAO) data really support a dark energy that changes over time, or whether that impression depends on the dark-energy model chosen for the fit. It compares the standard two-parameter Chevallier–Polarski–Linder (CPL) parametrization with the one-parameter SSLCPL parametrization, which is designed to approximate slowly rolling (thawing) scalar fields. The central finding is that in the SSLCPL model, adding DESI BAO data barely shifts the inferred present-day equation of state $w_0$, and the tension with the cosmological-constant model $\\Lambda$CDM is reduced relative to CPL. The authors conclude that the evidence for dynamical dark energy from DESI BAO is model-dependent, and that allowing spatial curvature does not change that conclusion.","feed_headline":"Dark-energy 'evidence' from DESI fades in a one-parameter model","feed_subtitle":"In the SSLCPL thawing-field model, DESI BAO barely moves w0 and the tension with ΛCDM drops from strong to weak.","key_machinery":"The machinery that carries the argument is the SSLCPL parametrization, a one-parameter version of the CPL model built from the slow-roll approximation to thawing scalar fields. In CPL, the dark-energy equation of state is $w(z)=w_0+w_a z/(1+z)$; in SSLCPL, $w_a$ is not free but is fixed by the degeneracy relation (the paper's equation 15) involving $w_0$ and the dark-energy density parameter $\\Omega_{\\phi 0}$, leaving only $w_0$ as a dark-energy parameter. This relation is what redistributes the DESI BAO constraint: with only one free parameter, the data no longer push $w_0$ away from $-1$ as strongly as they do in the two-parameter CPL fit. The statistical comparison uses the Akaike information criterion, $\\mathrm{AIC}=\\chi^2_{\\min}+2m$, to quantify the evidence difference between CPL, SSLCPL and $\\Lambda$CDM across nine data combinations.","core_discovery":"The paper's central claim is that the DESI BAO data by themselves do not robustly establish dynamical dark energy, because the apparent preference for $w_0>-1$ and $w_a<0$ is not stable when the dark-energy parametrization is changed. In the flat CPL model, adding DESI BAO to a combination of Planck compressed data, cosmic-chronometer $H(z)$ data, and supernova samples tightens the constraints on the parameters and deepens the tension with $\\Lambda$CDM, with the Akaike information criterion difference reaching $\\Delta\\mathrm{AIC}=-10$ when the DES five-year supernovae are used. Repeating the identical fits with the SSLCPL model, where the second CPL parameter $w_a$ is fixed by a thawing-field degeneracy relation rather than fitted freely, the DESI BAO data have only a small effect on $w_0$, and the best fit stays near $w_0\\simeq-0.90$; the tension with $\\Lambda$CDM is correspondingly weaker, with $\\Delta\\mathrm{AIC}=-6$ for the same supernova sample. Opening the curvature parameter $\\Omega_k$ changes the results only marginally. The conclusion is that the evidence for dynamical dark energy from DESI BAO is contingent on the assumed cosmological model.","pith_inferences":["A direct next step would be to repeat the analysis with the full Planck likelihood rather than the three compressed parameters; the paper validates the compressed approximation only for $\\Lambda$CDM and CPL, not for SSLCPL, so the SSLCPL posteriors could shift in a full-likelihood fit.","The same degeneracy argument implies that the DESI 'dynamical dark energy' significance is likely to shrink further in any model with a theoretically motivated relation between $w_0$ and $w_a$, such as thawing or freezing quintessence, so claims of evolving dark energy should be reported together with the model prior that produced them.","A robust resolution would come from model-independent reconstructions of $w(z)$ from BAO+SNe; if the reconstructed $w(z)$ is consistent with a constant $w=-1$ at low redshift, the CPL-based evidence would be exposed as a parametrization artifact."],"forward_implications":["If the SSLCPL model is the right description of dark energy, the DESI BAO data are consistent with $\\Lambda$CDM at about the $1\\sigma$ level for Pantheon+ and about the $2\\sigma$ level for DES and Union3 supernovae, so the widely quoted preference for evolving dark energy does not persist.","Model-comparison measures like $\\Delta\\mathrm{AIC}$ penalize extra parameters, so a two-parameter model will always show more 'evidence' for dynamics than a one-parameter model fitted to the same data; the SSLCPL result quantifies how much of the DESI hint is just parameter freedom.","The outlying DESI BAO point at $z=0.51$ is not the driver: excluding it (the BAO$^-$ datasets) leaves the model-dependence conclusion intact.","Allowing spatial curvature does not rescue the dynamical dark energy preference; the constraints and $\\Delta\\mathrm{AIC}$ values barely move when $\\Omega_k$ is freed."],"supporting_citations":[{"why":"Supplies the DESI BAO measurements (and the CPL constraints from the DESI collaboration) that the paper re-analyses to test model dependence.","marker":"[6]"},{"why":"Introduces the CPL parametrization $w(z)=w_0+w_a(1-a)$ used as the two-parameter comparison model.","marker":"[7]"},{"why":"Popularizes the CPL form used here and provides the theoretical basis for the two-parameter fit.","marker":"[8]"},{"why":"Derives the SSLCPL approximation from slow-roll thawing scalar fields, giving the one-parameter model under test.","marker":"[23]"},{"why":"Fixes the degeneracy relation between $w_0$ and $w_a$ in the SSLCPL model, the central identity of the analysis.","marker":"[24]"},{"why":"Provides the Pantheon+ supernova sample (PP) that gives the most $\\Lambda$CDM-consistent constraints in both models.","marker":"[9]"},{"why":"Provides the Union3 supernova sample and the compressed Planck parameters (Table F1) used in the combined datasets.","marker":"[10]"},{"why":"Supplies the DES 5-year supernova sample (D5) that yields the strongest CPL-versus-$\\Lambda$CDM evidence ($\\Delta\\mathrm{AIC}=-10$).","marker":"[11]"},{"why":"Gives the $\\Delta\\mathrm{AIC}$ thresholds that the paper uses to translate model-comparison values into evidence statements.","marker":"[15]"}],"fun_headline_variants":["DESI dark energy evidence hangs on model assumptions","Thawing-field model blunts DESI's dark energy push","Model choice flips DESI dark energy evidence","SSLCPL model dims DESI's dynamical dark energy hint","DESI BAO: dark energy dynamism not robust to model"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that the 26 radial BAO measurements included inside the $H(z)$ compilation are statistically independent of the DESI BAO measurements when both are used in the same combined likelihood; if they share information, the reported $\\chi^2$ and $\\Delta\\mathrm{AIC}$ values would be biased and the model comparison could shift.","fun_headline_variants_meta":{"raw":{"variants":["DESI dark energy evidence hangs on model assumptions","Thawing-field model blunts DESI's dark energy push","Model choice flips DESI dark energy evidence","SSLCPL model dims DESI's dynamical dark energy hint","DESI BAO: dark energy dynamism not robust to model"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000261,"raw_usage":{"total_tokens":1614,"prompt_tokens":987,"completion_tokens":627,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":544}},"tokens_in":603,"tokens_out":627,"duration_ms":5829,"temperature":1.0,"reasoning_tokens":544,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:36:59.997708+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the CPL and SSLCPL fits with the 26 radial BAO points removed from the $H(z)$ set whenever DESI BAO is included in the likelihood; if the SSLCPL $\\Delta\\mathrm{AIC}$ then drops below $-10$ or the DESI-induced shift in $w_0$ becomes as large as in CPL, the conclusion that DESI BAO has little influence in SSLCPL would be refuted.","supporting_citations":[{"cited_title":"Chevallier and D","cited_arxiv_id":null,"evidence_quote":"Introduces the CPL parametrization $w(z)=w_0+w_a(1-a)$ used as the two-parameter comparison model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Popularizes the CPL form used here and provides the theoretical basis for the two-parameter fit."},{"cited_title":"Gao and Y","cited_arxiv_id":null,"evidence_quote":"Derives the SSLCPL approximation from slow-roll thawing scalar fields, giving the one-parameter model under test."},{"cited_title":"Gong and Q","cited_arxiv_id":null,"evidence_quote":"Fixes the degeneracy relation between $w_0$ and $w_a$ in the SSLCPL model, the central identity of the analysis."},{"cited_title":"Scolnic et al., The Pantheon+ Analysis: The Full Data Set and Light-curve Release, Astrophys","cited_arxiv_id":null,"evidence_quote":"Provides the Pantheon+ supernova sample (PP) that gives the most $\\Lambda$CDM-consistent constraints in both models."},{"cited_title":"de Cruz Perez, C.-G","cited_arxiv_id":null,"evidence_quote":"Gives the $\\Delta\\mathrm{AIC}$ thresholds that the paper uses to translate model-comparison values into evidence statements."}],"review_version":1}