{"id":"340b98b6-88d1-4fb0-a6ed-08ec7a661d18","arxiv_id":"2504.20664","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 step transition in dark energy's equation of state to Planck+DESI DR2+DESY5 data gives a reported ~5 sigma preference for a phantom-to-quintessence transition at z about 0.49.","lead":"This cosmology paper fits a model where dark energy's equation of state switches from phantom-like at high redshift to quintessence-like at low redshift, and reports up to about 5 sigma preference for the switch using Planck, DESI DR2 and the DESY5 supernova sample. The claim is important if real, but it rests on a supernova sample with a known 0.04 magnitude offset and does not reduce the Hubble tension.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 5 sigma headline rests on a chi-square approximation (Sec. IV A, Eqs. 12-13) that is invalid because z_dagger is unidentified under the null Delta=0; the reported significance is therefore not calibrated.","rationale":"The reader's weakest_assumption precisely identifies the same issue: the chi-square difference test in Sec. IV A assumes 2 degrees of freedom, while under the null Delta=0 the transition redshift z_dagger is unidentifiable, so the regularity conditions fail. This is genuinely the most load-bearing concern because the entire headline '>5 sigma' is derived from that test, and it is not rescued by the more conservative Bayes factor. The paper itself does not acknowledge this identifiability problem, although it does acknowledge the DESY5 sample offset (Sec. V, citing [162]). The direction of the effect, Delta<0 and z_dagger ~ 0.5, is consistent across datasets, and the VCDM embedding provides a stable theoretical interpretation, but neither addresses the miscalibrated significance. The recommended verdict remains CONDITIONAL, unchanged from the reader: the paper's central claim should be re-evaluated with a boundary-corrected significance test or a bootstrap calibration, alongside a systematic treatment of the SN sample offset and code release. I see no reason to upgrade or downgrade the verdict based on this stress-test pass; the existing CONDITIONAL verdict already captures the required action.","tokens_in":24144,"tokens_out":3679,"duration_ms":45501,"concrete_test":"Run a parametric bootstrap under the null: generate synthetic Planck+DESI+DESY5 data from the best-fit LambdaCDM model using the same likelihood code and covariance matrices, then fit both the 6-parameter LambdaCDM and the 8-parameter w-dagger model to each realization and record Delta_chi^2_min. With 1000 realizations, compare the observed Delta_chi^2 = -27.34 to the empirical null distribution. If fewer than about 5% of simulated values exceed 27.34, the naive chi-square_2 p-value is confirmed; if the frequency is substantially larger, the reported 5 sigma is inflated and should be replaced with the empirically calibrated significance. A faster analytic cross-check is to apply Davies' (1987) bound for a chi-square_1 process maximized over the nuisance z_dagger and compare the resulting corrected p-value.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim of '5 sigma evidence' for a dynamical dark-energy transition is computed from Delta_chi^2_min = -27.34 (-29.24 for wVCDM) using the assumption in Sec. IV A that Delta_chi^2 follows a chi-square distribution with 2 degrees of freedom. This assumption is the load-bearing step for the headline significance. It is not justified, because under the null hypothesis Delta = 0, the transition redshift z_dagger is completely unidentifiable: the model prediction is independent of z_dagger when Delta = 0. The likelihood-ratio statistic therefore violates the regularity conditions that guarantee an asymptotic chi-square null distribution. In this non-regular setting, the correct null distribution is not chi-square_2; it is typically a distribution with a heavier tail than chi-square_1 (because the profile likelihood is maximized over z_dagger) and with a point mass contribution, and its exact form depends on the data covariance and the prior range of z_dagger. The conversion through Eq. (13) into sigma units is thus uncalibrated, and the stated >5 sigma cannot be claimed without a boundary-corrected or bootstrap-calibrated p-value. This concern is separate from, and prior to, the DESY5/PantheonPlus 0.04 mag offset, which the paper acknowledges; it directly affects every significance value in Table III, including the 4.9/5.1 sigma entry that motivates the abstract. The Bayes factor evidence (ln B_ij ~ -9) is a separate, more robust indicator, but it does not rescue the specific >5 sigma claim as stated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a two-parameter extension of ΛCDM in which the dark-energy equation of state switches discontinuously (w†CDM) or smoothly (w†VCDM) between w=-1+Δ and w=-1-Δ at a transition redshift z†. The authors fit this model to Planck CMB data, DESI DR2 BAO measurements, and the PantheonPlus, Union3, and DESY5 supernova samples, using CLASS and MontePython. They report a consistent preference for Δ<0 (phantom-to-quintessence transition) with z† around 0.5, and claim the strongest evidence for Planck+DESI+DESY5, with a significance exceeding ~5σ obtained from a chi-square difference test. They also report AIC improvements and Bayes factors favoring the extended model, while explicitly stating that the transition does not resolve the Hubble tension.","tokens_in":24389,"tokens_out":5531,"duration_ms":58750,"significance":"If the claimed ~5σ significance were reliable, this would be an important result: it would substantially challenge ΛCDM and point to a specific late-time transition in the dark-energy sector, consistent with the DESI DR2 preference for evolving dark energy. The paper's strengths include the use of current public datasets, the two-model (phenomenological and VCDM) cross-check, and the explicit caveat that H0 tension is not resolved. The consistency of the negative Δ preference across dataset combinations is suggestive. However, the headline significance is computed under an invalid asymptotic approximation, and the strongest signal comes from a supernova sample with a known internal offset relative to PantheonPlus. The central claim therefore needs recalibration before the evidence can be considered established.","major_comments":[{"comment":"The assumption that Δχ² follows a chi-square distribution with 2 degrees of freedom is not valid for this model. Under the null hypothesis Δ=0, the transition redshift z† is completely unidentifiable: the predicted observables are independent of z† when Δ=0. The likelihood-ratio statistic therefore violates the regularity conditions required for the asymptotic chi-square approximation. The correct null distribution is not χ²_2; it typically includes a contribution from profiling over the unidentifiable parameter, producing heavier tails than χ²_1 and a prior-dependent mixture. Since every significance in Table III (including the 4.9/5.1σ entries) is derived from Eqs. (12)-(13), the headline 'exceeding ~5σ' reported in the abstract is uncalibrated and likely overstated. The authors should provide a null distribution calibrated by simulation from the best-fit ΛCDM model, or a proper analytic treatment of the non-identifiability, and then re-evaluate all p-values and sigmas. Until this is done, the central statistical claim cannot be supported.","section":"IV A, Eqs. (12)-(13) and Table III"},{"comment":"The strongest evidence is obtained exclusively with the DESY5 sample, and the text acknowledges a ~0.04 mag offset between DESY5 and PantheonPlus, citing Ref. [162]. The reported Δχ² and the derived 5σ significance do not propagate this systematic difference into the significance calculation, so the headline result is conditional on the DESY5 calibration being correct. The authors state that the offset 'may partly explain the enhanced deviation,' but they do not quantify how the significance would change if the offset were treated as a systematic uncertainty. Given that the 5σ claim rests entirely on this dataset combination, the paper should either incorporate the offset as a nuisance parameter, report a shifted-magnitude sensitivity test, or demote the DESY5 combination in the abstract in favor of the more robust (though lower-significance) combinations. This is load-bearing for the central claim.","section":"V, Planck+DESI+DESY5 (Table II)"}],"minor_comments":[{"comment":"The conversion from p-value to sigma uses a two-tailed Gaussian quantile, σ = Φ^{-1}(1-p/2), but the chi-square difference test is one-sided: only negative Δχ² values count as evidence for the extended model. Using p/2 inflates the reported significance, albeit slightly. The standard one-sided conversion is σ = Φ^{-1}(1-p), or the authors should explain why a two-tailed interpretation is appropriate here.","section":"IV A, Eq. (13)"},{"comment":"The smoothness parameter ζ in the VCDM model is fixed to 10^1.5 for all runs, and the paper does not demonstrate that the main results are insensitive to this choice. A brief test with ζ varied over a range (or a statement of why this value is theoretically motivated) would strengthen the robustness of the VCDM results.","section":"III and IV"},{"comment":"The test is described as 'agnostic', but the parametrization imposes a specific functional form with a single sign-change transition. The paper could avoid overstatement by referring to the model as 'minimally parametrized' rather than fully agnostic, since the form of w(a) is fixed in advance.","section":"Abstract and Introduction"},{"comment":"In Figure 4 the labels 'w CDM' and 'w VCDM' are partially overwritten, and some numerical values of ln B are not clearly visible; the figure should be redrawn for legibility. Several language issues also appear in the text, such as 'in relative good agreement' and 'the under consideration in this work'; these should be corrected.","section":"Figure 4 and general text"}],"recommendation":"major_revision","confidential_remarks":"The paper has a genuine and interesting observational result, but the headline 5σ claim is not statistically valid as presented because of the unidentifiable z† under the null. This is fixable by a simulation-based calibration, and the consistency of the negative Δ across datasets gives the authors a reasonable route to a defensible paper. I recommend major revision rather than rejection. I also note that the v2 manuscript still contains the invalid chi-square assumption, so the revision should specifically address this point."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know about this paper. First, it is a competent and readable fit of a very simple late-time dark energy transition — a sign switch in w(a), plus a smooth VCDM version — to Planck, DESI DR2, and three supernova compilations. The genuinely new piece is the dataset combination and the VCDM embedding, and the result is coherent: every combination prefers Delta<0 with z_dag ~ 0.5, and the authors are careful to say the transition does not resolve H0 and that DESY5 has a known 0.04 mag offset relative to PantheonPlus. That is honest and useful.\n\nThe problem is the 5 sigma headline. It comes from Sec. IV A, where Delta_chi^2 between the 8-parameter model and 6-parameter Lambda-CDM is treated as chi-square with 2 degrees of freedom. That approximation is not valid here because, when Delta=0, z_dag is completely unidentifiable: the model's predictions do not depend on the transition redshift. The likelihood-ratio statistic therefore has a non-regular null distribution with a heavier tail and a point mass, and the p-values and sigmas in Table III are uncalibrated. This is not a nitpick; it is the load-bearing step for the abstract's '>5 sigma' claim. The direction of the effect is robust, and the Bayes factor also points toward the transition model, but a calibrated significance is likely lower, and 'discovery' language is premature.\n\nTwo smaller points. The strongest signal uses DESY5, which is exactly the sample with the known offset; the authors flag this but do not fold it into the significance. And the VCDM embedding is post-hoc — not circular, but not a prediction either.\n\nI would send this to a serious referee. The requested revision should be concrete: recalculate significance with a boundary-corrected or bootstrap-calibrated test, treat the DESY5/PantheonPlus offset systematically, and tone down the abstract until that is done. If that lands, the paper is a solid contribution to the late-time-transition literature.","headline":"A competent late-time transition analysis whose 5-sigma headline is uncalibrated because z_dag is unidentified under the null; worth a serious referee, but the significance needs a boundary-corrected test.","tokens_in":25015,"tokens_out":3851,"would_cite":true,"duration_ms":42143,"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":"The paper claims about 5-sigma evidence that the dark energy equation of state crossed the phantom divide near z≈0.5, switching from phantom-like behavior at high redshift to quintessence-like behavior at low redshift, without resolving…","keywords":["dark energy equation of state","phantom crossing","quintessence","late-time transition","DESI BAO","Hubble tension","VCDM","model comparison"],"falsifier":"Generate mock catalogs that mimic Planck+DESI+DESY5 under a true $\\Lambda$CDM cosmology, fit both the transition model and $\\Lambda$CDM to each mock, and compare the observed $\\Delta\\chi^2\\approx -27$ to the simulated null distribution; if the observed improvement is not as rare as a two-tailed $5\\sigma$ fluctuation, the central claim collapses.","tokens_in":23882,"feed_emoji":"🔭","tokens_out":8998,"duration_ms":78131,"temperature":0.7,"pith_summary":"The paper argues that the combined cosmic microwave background, baryon acoustic oscillation, and supernova data favor a dark energy equation of state that changes behavior at late times: phantom-like (w < −1) at redshifts above about 0.5 and quintessence-like (w > −1) below. The strongest signal comes from combining Planck CMB data, DESI DR2 BAO measurements, and the DES Year 5 supernova sample, giving roughly 5-sigma significance over the cosmological-constant model. The authors stress that the transition does not resolve the Hubble tension, with the inferred H0 remaining around 66 km/s/Mpc. If the evidence holds, the cosmological constant would not be the full explanation of late-time acceleration, and a dynamical dark energy sector—embedded here in a stable modified-gravity theory called VCDM—would be required.","feed_headline":"5-sigma evidence dark energy switched near z=0.5","feed_subtitle":"Planck+DESI+DES-Y5 data prefer a phantom-to-quintessence transition, but the Hubble tension remains","key_machinery":"The load-bearing object is the sign-switching equation-of-state parameterization $w(a) = -1 + \\Delta\\,\\mathrm{sgn}(a_\\dagger - a)$, whose two extra parameters ($\\Delta$, the deviation magnitude, and $a_\\dagger$ or $z_\\dagger$, the transition epoch) let the data pick both the size and the direction of a late-time transition without prior bias. A smooth variant $w(N) = -1 + \\Delta\\tanh[\\zeta(N_\\dagger - N)]$ is embedded in the VCDM theory (a minimal modified gravity with no new propagating degrees of freedom) to show that the switch need not trigger instabilities. The statistical engine is the $\\chi^2$-difference test with two additional degrees of freedom, supplemented by AIC and Bayes-factor comparisons, applied to MCMC chains over the full parameter set.","core_discovery":"The central discovery claim is that the dark energy equation of state $w$ crosses the phantom divide $w=-1$ near redshift $z_\\dagger\\approx 0.49$--$0.53$, in the direction from phantom at high redshift to quintessence at low redshift. This is captured by the two-parameter extension $w(a) = -1 + \\Delta\\,\\mathrm{sgn}(a_\\dagger - a)$, where $\\Delta<0$ means $w<-1$ before the transition and $w>-1$ after it; the smooth version uses $w(N) = -1 + \\Delta\\tanh[\\zeta(N_\\dagger - N)]$ and is embedded in the VCDM modified-gravity theory to keep perturbations stable. With Planck+DESI+DESY5, the fit improves by $\\Delta\\chi^2 \\approx -27$ to $-29$ and the AIC by about $-23$, translating to a significance of $\\sim 5\\sigma$ ($4.9\\sigma$ for the abrupt model, $5.1\\sigma$ for the smooth VCDM model). The same negative-$\\Delta$ signature appears, at lower significance, in every other dataset combination tested. The authors also find that the transition does not fully resolve the Hubble tension.","pith_inferences":["The reported significance may be overstated: when $\\Delta=0$ the transition redshift is unidentifiable, so the $\\chi^2$ difference is not guaranteed to follow a $\\chi^2$ distribution with 2 degrees of freedom; calibrating the null distribution via simulations could shift the claimed evidence.","Part of the DESY5-driven signal could be a supernova systematics artifact: the paper notes a ~0.04 magnitude offset between low- and high-redshift SN Ia, and a joint reanalysis of all three SN samples with a unified systematic model might reduce the significance.","A falsifiable consequence: a real transition at $z\\approx 0.5$ should also affect the growth of cosmic structure, so redshift-space distortion and lensing measurements at $0.5\\lesssim z\\lesssim 1$ could either corroborate or rule out the mechanism."],"forward_implications":["If the signal is real, the ΛCDM model is disfavored at about 5σ by the Planck+DESI+DESY5 combination, and by 2.7–3.6σ in other combinations, strengthening the case for dynamical dark energy.","The preferred cosmology has $w<-1$ at $z\\gtrsim 0.5$ and $w>-1$ today, consistent with the trend reported by the DESI BAO data themselves.","The transition does not resolve the $H_0$ tension: the inferred Hubble constant remains near 66 km/s/Mpc, still in tension with local measurements.","The same negative-$\\Delta$ preference appears in both the abrupt and the smooth VCDM versions, so the qualitative conclusion is independent of transition shape."],"supporting_citations":[{"why":"Supplies the DESI DR2 BAO measurements in nine redshift bins that drive the low-redshift preference for a transition.","marker":"[70]"},{"why":"Provides the Planck 2018 CMB temperature and polarization likelihoods used in every dataset combination.","marker":"[2]"},{"why":"PantheonPlus supernova sample; with Planck+DESI gives a 2.9σ preference for the transition.","marker":"[151]"},{"why":"Union3 supernova compilation; yields a 3.6σ preference and supports Δ<0.","marker":"[152]"},{"why":"DES Year 5 supernova sample that produces the strongest ~5σ signal.","marker":"[153]"},{"why":"Introduces the VCDM modified-gravity theory used to embed the smooth transition without instabilities.","marker":"[133]"},{"why":"The parametrized post-Friedmann (PPF) framework used to evolve dark energy perturbations in the analysis.","marker":"[135]"},{"why":"Cited for the ~0.04 magnitude offset between low- and high-redshift SN Ia, used to interpret the DESY5 enhancement.","marker":"[162]"}],"fun_headline_variants":["5-sigma cosmic flip: dark energy shifts phase","Dark energy phantom-to-quintessence swing at z~0.5","Hubble tension persists as dark energy pivots","Late-time dark energy shift: phantom to quintessence, 5-sigma"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The headline significance assumes that the $\\chi^2$ difference between the 8-parameter transition model and the 6-parameter $\\Lambda$CDM follows a $\\chi^2$ distribution with 2 degrees of freedom; this standard likelihood-ratio regularity condition fails at $\\Delta=0$ because the transition redshift is fully unidentifiable, so the reported $\\sim 5\\sigma$ may be inflated.","fun_headline_variants_meta":{"raw":{"variants":["5-sigma cosmic flip: dark energy shifts phase","Dark energy phantom-to-quintessence swing at z~0.5","Hubble tension persists as dark energy pivots","Late-time dark energy shift: phantom to quintessence, 5-sigma"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000645,"raw_usage":{"total_tokens":3056,"prompt_tokens":1132,"completion_tokens":1924,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":748,"completion_tokens_details":{"reasoning_tokens":1852}},"tokens_in":748,"tokens_out":1924,"duration_ms":13507,"temperature":1.0,"reasoning_tokens":1852,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:23:33.277125+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate mock catalogs that mimic Planck+DESI+DESY5 under a true $\\Lambda$CDM cosmology, fit both the transition model and $\\Lambda$CDM to each mock, and compare the observed $\\Delta\\chi^2\\approx -27$ to the simulated null distribution; if the observed improvement is not as rare as a two-tailed $5\\sigma$ fluctuation, the central claim collapses.","supporting_citations":[],"review_version":1}