Pith. sign in

REVIEW 1 major objections 4 minor 51 references

DESI's DR2 preference for dynamical dark energy is a fragile, single-bin, specification-dependent signal, not evidence that the cosmological constant is wrong.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-03 17:13 UTC pith:K4JSAEA5

load-bearing objection A well-scoped reanalysis that convincingly shows the DESI DR2 dynamical dark energy signal is fragile and LRG2-localized, but the anytime-validity guarantee rests on a plausible yet unverified year-scaling model. the 1 major comments →

arxiv 2607.28918 v1 pith:K4JSAEA5 submitted 2026-07-31 astro-ph.CO stat.ME

A Sequentially-Valid Reanalysis of DESI's Dynamical Dark Energy Signal

Jinyoung Kim (1 , 2) , David F. Mota (3) , Andrius Tamosiunas (3 , 4) ((1) Stanford University , (2) Two Sigma Investments , (3) University of Oslo , (4) Case Western Reserve University) This is my paper
classification astro-ph.CO stat.ME
keywords dark energybaryon acoustic oscillationsDESI DR2dynamical dark energye-processanytime-valid inferencecosmological constantsequential testing
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper reanalyzes the DESI DR1-to-DR2 baryon acoustic oscillation data with an e-process, a likelihood-ratio measure of evidence that stays valid when the same hypothesis is tested repeatedly across data releases. It tries to establish that the DESI DR2 preference for a time-varying dark energy equation of state over the cosmological constant is not robust: a running e-value reaches 33.97 under a pre-specified alternative aligned with the data's preferred direction, but removing the LRG2 redshift bin collapses it to 0.49, and look-elsewhere-corrected statistics fail to reject. A sympathetic reader would care because the method addresses the growing worry that repeatedly checking a cosmological signal across releases inflates false-positive rates, and because the result undercuts headline claims of a 3–4 sigma detection.

Core claim

On the paper's own terms, the central discovery is that the DESI BAO evidence for w0waCDM carries an anytime-valid false-positive guarantee only under the right test specification, and that the signal is localised almost entirely in one redshift bin. The running mixture e-value M_DR2 = 33.97 crosses the illustrative 5% threshold of 20, giving a Markov p-value of 0.029, but this rejection requires a narrow prior concentrated near the DR2-preferred direction and a signal shared coherently across the bins. Leave-one-out analysis shows LRG2 carries 78.6% of the summed evidence; recomputing on the remaining six bins gives M = 0.49, mildly favouring the cosmological constant. Allowing each of the

What carries the argument

The central object is the mixture e-value, defined as a Bayes factor with a fixed, pre-specified prior on the dark energy parameters (w0, wa): M_t = ∫ [L(D≤t | θ) / L(D≤t | H0)] π(θ)dθ. Under the null, each e-value has expectation at most one, and the running maximum of an e-process obeys Ville's inequality, bounding the false-positive rate at any stopping time without a trials-factor penalty. Because DESI's releases are nested, the paper models the shared noise with a year-scaling decomposition — DR2 carries one-third of DR1's year-one noise plus two-thirds of new years' noise — which turns the release sequence into an exact martingale and makes the anytime-valid guarantee apply. Leave-one-

Load-bearing premise

The whole anytime-valid guarantee depends on assuming that the noise shared between DR1 and DR2 scales exactly with observing time, with DR2 carrying one-third of DR1's year-one noise and two-thirds of new noise, so that the release sequence becomes an exact martingale.

What would settle it

Measure the actual noise overlap between DR1 and DR2 from a re-reduction of the same sky or from mocks with injected systematics. If the true year-one noise share falls outside [0.25, 0.40], or if E[M_DR2 | F_1] does not equal M_DR1 in closed form, the martingale property fails and the e-value no longer carries its time-uniform guarantee; the headline M=33.97 would then not justify the 'no trials-factor penalty' claim.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the analysis is right, DR2's reported 3–4 sigma preference for evolving dark energy should not be read as evidence that w ≠ −1; the same data, tested sequentially, reject only under one concentrated prior and only while the LRG2 bin is included.
  • Future DESI, Euclid, Roman, and LSST releases should report an e-process value with its test specification stated alongside conventional sigma significances, so evidence accumulation across releases stays transparent and false-positive control holds.
  • Under the year-scaling model, DR3 and later data can be folded into the same running process at no additional multiplicity cost, but the decisive improvement will be sharper measurements of the existing LRG2 bin, not just new high-redshift bins.
  • A physically motivated thawing-quintessence alternative rejects more strongly than any agnostic specification, while freezing-quintessence priors do not reject; the choice of test hypothesis matters substantially for the conclusion.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The martingale model's dependence on a precise year-one noise share (assumed 1/3, admissible 0.25–0.40) is testable: if future DESI re-reductions measure a different overlap, the headline 'no trials-factor penalty' guarantee would need revision and the rejection should be re-evaluated.
  • The LRG2-specific concentration suggests a bin-local systematic rather than smooth dark energy; a direct test would cross-correlate LRG2's distance residuals with imaging depth, seeing, or redshift-success maps, or split the bin by angular region and check whether the e-value survives.
  • The strong dependence on which Planck background column is used, with e-values spanning a factor of 163 across self-consistent columns, implies that a fully marginalised analysis over h, Omega_m, and r_d could move the headline value substantially; the paper expects fixed-background values to be inflated, so a marginalised e-process is a natural next calculation.
  • If adopted by other multi-release surveys, this approach would shift reporting norms from sigma values alone to pre-registered e-processes — a change that would materially alter how future anomalies are interpreted, since e-values can be multiplied across independent streams such as supernova and CMB datasets only if the full likelihoods are shared.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

1 major / 4 minor

Summary. The paper applies e-process methodology to DESI DR1→DR2 BAO data to test ΛCDM against w0waCDM under repeated looks. It constructs a running mixture e-value M_t, reports M_DR2 = 33.97 under a Default prior (crossing an illustrative 5% threshold of 20), but shows this rejection is fragile: it is concentrated in LRG2, collapses to M = 0.49 when LRG2 is removed, fails look-elsewhere corrections (product 4.46, mean 1.52), depends on prior width, and fails when compressed Planck CMB is added (E = 2.19). To obtain a sequential guarantee, the authors introduce a year-scaling noise model (App. B.2) under which the joint process is a martingale, giving anytime-valid p = 0.015 and the claim that future releases can be folded in at no further cost. The paper concludes that the DESI DR2 signal is fragile and specification-dependent rather than robust evidence for dynamical dark energy, and recommends that sequential surveys report e-process values with stated test specifications.

Significance. If the fragility claim holds, this is an important contribution to the ongoing debate over DESI DR2, providing a formal statistical demonstration that the reported 3–4σ preference for w0waCDM is not robust. The e-process framework is well suited to multi-release surveys, and the paper's stress tests (bin removal, look-elsewhere correction, prior sensitivity, background cosmology) are thorough and transparent. The paper ships analysis code and data, and the e-value validity proofs are standard and correct. The bootstrapped concentration analysis (Table 4) is a nice addition that distinguishes a localised anomaly from a smooth signal. However, the paper's headline sequential-validity guarantee depends on a specific year-scaling model that is not empirically validated, and the claim that future releases can be added 'at no further cost' requires qualification.

major comments (1)
  1. [Section 4.1 / Appendix B.2] The anytime-validity guarantee (Ville bound p = 0.015, 'DR3+ at no further cost') rests entirely on the year-scaling decomposition DR1 = μ + ε_y1, DR2 = μ + (1/3)ε_y1 + (2/3)ε_y23 (Eq. B.1). The paper shows the snapshot process is not a martingale (E[M_DR2|F1] = 0.750 vs M_DR1 = 1.054), so the sequential claim is model-dependent. The sensitivity analysis varies α1 in [0.25,0.40] and checks the observed M_joint, but not the null false-positive rate. At α1 = 0.25, within the published covariance ratio range (0.22–0.49), E[M_joint_DR2|H0] = 162, so the supermartingale property and the 5% false-positive control fail. Please either qualify the anytime-validity claim as conditional on the exact α1 = 1/3 model, or provide a false-positive sensitivity analysis / robust e-process construction.
minor comments (4)
  1. [Section 5 / Table 2] Describing M = 0.49 as 'mildly favouring ΛCDM' overstates what an e-value below one means; e-values are not calibrated for evidence in favour of the null. Please soften the wording or provide the null distribution of the running mixture without LRG2.
  2. [Section 2] The Default prior is described as pre-specified, but the box is explicitly sized to contain external SN+CMB MAP points and also contains the DR2 MLE. Please clarify the chronology: was the prior fixed before the DESI DR2 result, and did the DR2 MLE play any role in the prior choice?
  3. [Section 4.4] The DR3 forecast assumes that the published DR2 covariance scales with exposure. This assumption is stated in App. B.2 but should be repeated in the main text of Section 4.4, where the forecast is presented without that caveat.
  4. [Appendix B.6] The sentence 'we expect the fixed-background e-values to be inflated relative to the marginalised case' is a speculation that is not tested in the paper. Please phrase it as a concern and note that the net sign of marginalisation is not guaranteed, as the text itself acknowledges.

Circularity Check

0 steps flagged

No significant circularity: the e-value construction and all robustness checks are self-contained, with no load-bearing self-citation.

full rationale

The paper's central derivation is self-contained. The e-value is a proper-prior Bayes factor (Eq. 2.1) whose null expectation is one by Proposition 3; the Type-I control comes from Markov/Ville bounds proved in Appendix A. The only nonstandard ingredient, the joint DR1→DR2 process, is built explicitly in Appendix B.2: under the stated year-scaling law (Eq. B.1), the conditional mean identity E[M_joint_DR2|F1] = M_DR1 holds by construction, and the paper verifies rather than assumes the martingale property (it reports explicitly that the snapshot sequence is neither a super- nor a submartingale). The headline 'fragile, specification-dependent' conclusion is supported by independent leave-one-out, look-elsewhere, and prior-width checks in Sections 4.2–4.3, not by definition. The Default prior being sized to contain the published SN+CMB MAP points is a specification choice, not a fitted parameter, and it is stress-tested; the δ-window of Appendix B.5 is descriptive geometry, not a step in the inference. Reference [17] shares an author but is used only as corroborating context; the paper's own computations carry the argument. No equation reduces to another by construction, and no fitted input is renamed as a prediction.

Axiom & Free-Parameter Ledger

4 free parameters · 6 axioms · 0 invented entities

The analysis introduces no new physical entities — no new particles, forces, or fields. The free parameters are all test-specification or modeling choices: the prior box, the REGROW effect-size δ, the fixed Planck background triple, and the year-scaling noise share. The correctness of the headline rejection depends on the alignment of these choices with the data; the strength of the paper is that it tables this dependence explicitly (Tables 1 and 3, Appendices B.2–B.6). The 'innovation' y = 3/2 DR2 − 1/2 DR1 of Eq. (B.3) is a statistical construction, not an invented entity.

free parameters (4)
  • Default prior box bounds [w0, wa] = [−1.5,−0.5] × [−2.0,1.0] (30×30 grid) = box bounds; yields M_DR2 = 33.97 under the Default specification
    Chosen by hand, not derived; the paper states the box is 'sized to be wide enough to contain the published joint SN+CMB MAP points' of Pantheon+, DES-Y5, Union3 and the DR2 MLE. Whether the threshold is crossed depends directly on this choice (Table 1: Narrow rejects at 144, Wide does not at 10.7).
  • REGROW minimum effect size δ (Fisher-σ) = δ = 2 and δ = 3 reject (72.5, 297); δ = 1 does not (7.58)
    The rejecting δ values lie inside the data-supported window δ ∈ [0.87, 4] derived from the same DR2 likelihood (Appendix B.5, Eq. B.4), with δ_MLE = 4.06. Choosing δ is equivalent to choosing whether the test rejects; the paper openly reports this sensitivity.
  • Planck-fixed background triple (h, Ωm, rd) = (0.6766, 0.3111, 147.05 Mpc) = mixed Planck 2018 columns: (h, Ωm) from +lensing+BAO chain, rd from CMB-only
    Fixed rather than marginalized; Appendix B.3 shows the headline e-value spans a factor of 163 across self-consistent Planck columns, with the self-consistent +lensing+BAO column giving M_DR2 = 8.70 (no rejection). The body quotes the mid-range mixed triple (a).
  • Year-one noise share α1 = 1/3 in the martingale decomposition = 1/3 (survey-design assumption)
    Not fitted to data but a modeling choice that the anytime-valid claim depends on: the joint e-value ranges M_joint ∈ [65, 131] over admissible α1 ∈ [0.25, 0.40] and grows without bound as the 'true share' approaches independence (α1 = 0.25 → E = 162). The rejection survives this range but the quoted anytime-valid p = 0.015 does not.
axioms (6)
  • domain assumption Gaussian BAO likelihood with the published 13×13 (DR2) and 12-measurement (DR1) covariances, block-diagonal by bin
    Section 3: the entire e-value computation uses the published covariance as exact; Type-I control 'still assumes the Gaussian BAO likelihood with the published covariance, the same model the χ2 analysis uses' (Section 2).
  • domain assumption CPL parametrization w(a) = w0 + wa(1−a)
    Section 2 and B.6: 'Every number conditions on the CPL parametrisation'; the paper names 'whether the LRG2 signal survives outside CPL' as the most direct open robustness test.
  • ad hoc to paper Year-scaling decomposition DR1 = μ + ε_y1, DR2 = μ + (1/3)ε_y1 + (2/3)ε_y23, with shared mean μ and Gaussian noise
    Appendix B.2, Eqs. (B.1)–(B.3): introduced for this analysis to make the nested release sequence an exact martingale; the paper's own Monte Carlo shows the snapshot sequence is neither super- nor submartingale.
  • domain assumption Planck 2018 background held fixed, with (h, Ωm, rd) treated as inputs
    Appendix B.3: standard for BAO-only analyses but suppresses the wa–Ωm and w0–Ωm degeneracies; the paper flags the mixed-column inconsistency and the factor-163 sensitivity its own Table 3 demonstrates.
  • standard math Ville's inequality and optional stopping for non-negative supermartingales
    Appendix A, Theorem 1: the foundational bound; standard probability theory applied to the constructed martingale.
  • domain assumption Independence across the 7 redshift bins in the per-bin product construction
    Section 4.2 and Appendix B.4: the per-bin-independent product Q_k M_k is valid only because the published covariance is claimed to be exactly block-diagonal across bins.

pith-pipeline@v1.3.0-daily-deepseek · 5378 in / 5765 out tokens · 228210 ms · 2026-08-03T17:13:52.428194+00:00 · methodology

0 comments
read the original abstract

Surveys such as DESI release data in stages, and each release invites a fresh assessment of whether dark energy is consistent with a cosmological constant. Standard Wilks-based $\sigma$ values are interpreted as if the test were performed only once, yet revisiting the same question at DR1, DR2, DR3, and beyond inflates the chance of an apparently significant fluctuation. We reanalyse the DESI BAO evidence for dynamical dark energy using an e-process, a likelihood-ratio-based measure of evidence that remains valid under repeated looks at accumulating data: it controls the false-detection probability under continued testing across releases with no trials-factor penalty. Applied to the DESI DR1$\to$DR2 BAO sequence, the result depends strongly on which departures from $\Lambda$CDM the test is designed to detect. For a pre-specified alternative aligned with the DR2-preferred direction, the running evidence reaches $M_{DR2} = 33.97$, crossing an illustrative 5% threshold of 20. The evidence is concentrated almost entirely in the LRG2 redshift bin, and removing that bin reduces the running e-value to $M = 0.49$, mildly favouring $\Lambda$CDM. When allowing any of the seven bins to have produced the excess, the signal does not survive a look-elsewhere correction. Other physically agnostic alternatives also fail to reject $\Lambda$CDM, while a physically motivated thawing-quintessence alternative rejects more strongly than any agnostic choice. The concentration of evidence in LRG2 specifically is difficult to reconcile with a smoothly evolving equation of state. The data therefore points to a fragile, single-bin, specification-dependent signal rather than robust evidence for dynamical dark energy. We recommend that future DESI, Euclid, Roman, and LSST data releases report an anytime-valid e-process value with its test specification stated alongside conventional $\sigma$ significances.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

51 extracted references · 1 canonical work pages

  1. [1]

    Aghanim, Y

    Planck Collaboration, N. Aghanim, Y. Akrami, M. Ashdown, J. Aumont, C. Baccigalupi et al., Planck 2018 results. VI. Cosmological parameters, A&A641(2020) A6 [1807.06209]

    Pith/arXiv arXiv 2018

  2. [2]

    Louis, A

    T. Louis, A. La Posta, Z. Atkins, H.T. Jense, I. Abril-Cabezas, G.E. Addison et al.,The Atacama Cosmology Telescope: DR6 power spectra, likelihoods andΛCDM parameters, J. Cosmology Astropart. Phys.2025(2025) 062 [2503.14452]

    Pith/arXiv arXiv 2025

  3. [3]

    Riess, A.V

    A.G. Riess, A.V. Filippenko, P. Challis, A. Clocchiatti, A. Diercks, P.M. Garnavich et al., Observational Evidence from Supernovae for an Accelerating Universe and a Cosmological Constant, AJ116(1998) 1009 [astro-ph/9805201]

    Pith/arXiv arXiv 1998

  4. [4]

    Perlmutter, G

    S. Perlmutter, G. Aldering, G. Goldhaber, R.A. Knop, P. Nugent, P.G. Castro et al., Measurements ofΩandΛfrom 42 High-Redshift Supernovae, ApJ517(1999) 565 [astro-ph/9812133]

    Pith/arXiv arXiv 1999

  5. [5]

    Bacon, A.R

    D.J. Bacon, A.R. Refregier and R.S. Ellis,Detection of weak gravitational lensing by large-scale structure, MNRAS318(2000) 625 [astro-ph/0003008]

    Pith/arXiv arXiv 2000

  6. [6]

    Kilbinger,Cosmology with cosmic shear observations: a review,Reports on Progress in Physics78(2015) 086901 [1411.0115]

    M. Kilbinger,Cosmology with cosmic shear observations: a review,Reports on Progress in Physics78(2015) 086901 [1411.0115]

    Pith/arXiv arXiv 2015

  7. [7]

    Peacock, S

    J.A. Peacock, S. Cole, P. Norberg, C.M. Baugh, J. Bland-Hawthorn, T. Bridges et al.,A measurement of the cosmological mass density from clustering in the 2dF Galaxy Redshift Survey, Nature410(2001) 169 [astro-ph/0103143]

    Pith/arXiv arXiv 2001

  8. [8]

    Tegmark, M.R

    M. Tegmark, M.R. Blanton, M.A. Strauss, F. Hoyle, D. Schlegel, R. Scoccimarro et al.,The Three-Dimensional Power Spectrum of Galaxies from the Sloan Digital Sky Survey, ApJ606 (2004) 702 [astro-ph/0310725]

    Pith/arXiv arXiv 2004

  9. [9]

    Eisenstein, I

    D.J. Eisenstein, I. Zehavi, D.W. Hogg, R. Scoccimarro, M.R. Blanton, R.C. Nichol et al., Detection of the Baryon Acoustic Peak in the Large-Scale Correlation Function of SDSS Luminous Red Galaxies, ApJ633(2005) 560 [astro-ph/0501171]

    Pith/arXiv arXiv 2005

  10. [10]

    Anderson, É

    L. Anderson, É. Aubourg, S. Bailey, F. Beutler, V. Bhardwaj, M. Blanton et al.,The clustering of galaxies in the SDSS-III Baryon Oscillation Spectroscopic Survey: baryon acoustic oscillations in the Data Releases 10 and 11 Galaxy samples, MNRAS441(2014) 24 [1312.4877]

    Pith/arXiv arXiv 2014

  11. [11]

    Di Valentino, L.A

    E. Di Valentino, L.A. Anchordoqui, Ö. Akarsu, Y. Ali-Haimoud, L. Amendola, N. Arendse et al.,Cosmology Intertwined II: The hubble constant tension,Astroparticle Physics131(2021) 102605 [2008.11284]

    Pith/arXiv arXiv 2021

  12. [12]

    Di Valentino, J.L

    E. Di Valentino, J.L. Said, A. Riess, A. Pollo, V. Poulin, A. Gómez-Valent et al.,The CosmoVerse White Paper: Addressing observational tensions in cosmology with systematics and fundamental physics,Physics of the Dark Universe49(2025) 101965 [2504.01669]. – 20 –

    Pith/arXiv arXiv 2025

  13. [13]

    Poulin, J.L

    V. Poulin, J.L. Bernal, E.D. Kovetz and M. Kamionkowski,Sigma-8 tension is a drag, Phys. Rev. D107(2023) 123538 [2209.06217]

    Pith/arXiv arXiv 2023

  14. [14]

    Pantos and L

    I. Pantos and L. Perivolaropoulos,Status of the S8 tension: A 2026 review of probe discrepancies,Physics of the Dark Universe52(2026) 102286 [2602.12238]

    arXiv 2026

  15. [15]

    Abdul Karim, J

    M. Abdul Karim, J. Aguilar, S. Ahlen, S. Alam, L. Allen, C. Allende Prieto et al.,DESI DR2 results. II. Measurements of baryon acoustic oscillations and cosmological constraints, Phys. Rev. D112(2025) 083515 [2503.14738]

    Pith/arXiv arXiv 2025

  16. [16]

    Cortês and A.R

    M. Cortês and A.R. Liddle,Interpreting DESI’s evidence for evolving dark energy, J. Cosmology Astropart. Phys.2024(2024) 007 [2404.08056]

    Pith/arXiv arXiv 2024

  17. [17]

    Wang and D

    D. Wang and D. Mota,Did DESI DR2 truly reveal dynamical dark energy?,Eur. Phys. J. C 85(2025) 1356 [2504.15222]

    arXiv 2025

  18. [18]

    Ó Colgáin and M.M

    E. Ó Colgáin and M.M. Sheikh-Jabbari,DESI and SNe: dynamical dark energy,Ωm tension or systematics?, MNRAS542(2025) L24 [2412.12905]

    Pith/arXiv arXiv 2025

  19. [19]

    Efstathiou,Evolving dark energy or supernovae systematics?, MNRAS538(2025) 875 [2408.07175]

    G. Efstathiou,Evolving dark energy or supernovae systematics?, MNRAS538(2025) 875 [2408.07175]

    Pith/arXiv arXiv 2025

  20. [20]

    Ormondroyd, W.J

    A.N. Ormondroyd, W.J. Handley, M.P. Hobson, A.N. Lasenby and D. Yallup,Dynamic or Systematic? Bayesian model selection between dark energy and supernova biases, MNRAS (2026) [2509.13220]

    Pith/arXiv arXiv 2026

  21. [21]

    D.D.Y. Ong, D. Yallup and W. Handley,A Bayesian Perspective on Evidence for Evolving Dark Energy,2511.10631

    Pith/arXiv arXiv

  22. [22]

    G. Liu, Y. Wang and W. Zhao,Impact of LRG1 and LRG2 in DESI 2024 BAO data on dark energy evolution,2407.04385

    Pith/arXiv arXiv 2024

  23. [23]

    Z. Wang, S. Lin, Z. Ding and B. Hu,The role of LRG1 and LRG2’s monopole in inferring the DESI 2024 BAO cosmology,Mon. Not. Roy. Astron. Soc.534(2024) 3869 [2405.02168]

    Pith/arXiv arXiv 2024

  24. [24]

    Ó Colgáin, S

    E. Ó Colgáin, S. Pourojaghi, M.M. Sheikh-Jabbari and L. Yin,How much has DESI dark energy evolved since DR1?,2504.04417

    arXiv

  25. [25]

    Chevallier and D

    M. Chevallier and D. Polarski,Accelerating universes with scaling dark matter,Int. J. Mod. Phys. D10(2001) 213 [gr-qc/0009008]

    Pith/arXiv arXiv 2001

  26. [26]

    Linder,Exploring the expansion history of the universe,Phys

    E.V. Linder,Exploring the expansion history of the universe,Phys. Rev. Lett.90(2003) 091301

    2003

  27. [27]

    Nesseris, Y

    S. Nesseris, Y. Akrami and G.D. Starkman,To CPL, or not to CPL? What we have not learned about the dark energy equation of state,arXiv e-prints(2025) arXiv:2503.22529 [2503.22529]

    Pith/arXiv arXiv 2025

  28. [28]

    Artola, I

    M. Artola, I. Ayuso, R. Lazkoz and V. Salzano,Is Chevallier-Polarski-Linder dark energy a mirage?, Phys. Rev. D113(2026) 023513 [2510.04191]

    arXiv 2026

  29. [29]

    Alam and M

    S. Alam and M. Wali Hossain,Beyond CPL: Evidence for dynamical dark energy in three-parameter models, J. Cosmology Astropart. Phys.2026(2026) 042 [2510.03779]

    Pith/arXiv arXiv 2026

  30. [30]

    Lee,Assessing the robustness of the CPL parametrization to basis and prior variations: insights from DESI DR2 BAO data,European Physical Journal C86(2026) 191 [2506.18230]

    S. Lee,Assessing the robustness of the CPL parametrization to basis and prior variations: insights from DESI DR2 BAO data,European Physical Journal C86(2026) 191 [2506.18230]

    arXiv 2026

  31. [31]

    BICEP2 Collaboration,Detection of B-Mode Polarization at Degree Angular Scales by BICEP2,Phys. Rev. Lett.112(2014) 241101 [1403.3985]

    Pith/arXiv arXiv 2014

  32. [32]

    BICEP2/Keck and Planck Collaborations,Joint Analysis of BICEP2/Keck Array and Planck Data,Phys. Rev. Lett.114(2015) 101301 [1502.00612]

    Pith/arXiv arXiv 2015

  33. [33]

    Bowman, A.E.E

    J.D. Bowman, A.E.E. Rogers, R.A. Monsalve, T.J. Mozdzen and N. Mahesh,An absorption profile centred at 78 megahertz in the sky-averaged spectrum,Nature555(2018) 67

    2018

  34. [34]

    Singh et al.,On the detection of a cosmic dawn signal in the radio background,Nature Astronomy6(2022) 607 [2112.06778]

    S. Singh et al.,On the detection of a cosmic dawn signal in the radio background,Nature Astronomy6(2022) 607 [2112.06778]. – 21 –

    Pith/arXiv arXiv 2022

  35. [35]

    Vovk and R

    V. Vovk and R. Wang,E-values: Calibration, combination and applications,The Annals of Statistics49(2021) 1736

    2021

  36. [36]

    Shafer,Testing by betting: A strategy for statistical and scientific communication,Journal of the Royal Statistical Society Series A: Statistics in Society184(2021) 407

    G. Shafer,Testing by betting: A strategy for statistical and scientific communication,Journal of the Royal Statistical Society Series A: Statistics in Society184(2021) 407

    2021

  37. [37]

    Ramdas, P

    A. Ramdas, P. Grünwald, V. Vovk and G. Shafer,Game-Theoretic Statistics and Safe Anytime-Valid Inference,Statistical Science38(2023) 576

    2023

  38. [38]

    P.D. Grünwald,Beyond neyman–pearson: E-values enable hypothesis testing with a data-driven alpha,Proceedings of the National Academy of Sciences121(2024) e2302098121 [https://www.pnas.org/doi/pdf/10.1073/pnas.2302098121]

    work page doi:10.1073/pnas.2302098121 2024

  39. [39]

    Grünwald, R

    P. Grünwald, R. {de Heide} and W. Koolen,Safe testing,Journal of the Royal Statistical Society. Series B: Statistical Methodology86(2024) 1091

    2024

  40. [40]

    Howard, A

    S.R. Howard, A. Ramdas, J. McAuliffe and J. Sekhon,Time-uniform chernoff bounds via nonnegative supermartingales,Probab. Surveys17(2020) 257 [1808.03204]

    arXiv 2020

  41. [41]

    Brout et al.,The Pantheon+ Analysis: Cosmological Constraints,Astrophys

    D. Brout et al.,The Pantheon+ Analysis: Cosmological Constraints,Astrophys. J.938(2022) 110 [2202.04077]

    Pith/arXiv arXiv 2022

  42. [42]

    Collaboration, T.M.C

    D. Collaboration, T.M.C. Abbott et al.,The Dark Energy Survey: Cosmology Results With ∼1500 New High-redshift Type Ia Supernovae Using The Full 5-year Dataset,Astrophys. J. Lett.973(2024) L14 [2401.02929]

    Pith/arXiv arXiv 2024

  43. [43]

    Rubin et al.,Union Through UNITY: Cosmology with 2,000 SNe Using a Unified Bayesian Framework,2311.12098

    D. Rubin et al.,Union Through UNITY: Cosmology with 2,000 SNe Using a Unified Bayesian Framework,2311.12098

    Pith/arXiv arXiv

  44. [44]

    DESI Collaboration,DESI 2024 VI: Cosmological Constraints from the Measurements of Baryon Acoustic Oscillations,JCAP02(2025) 021 [2404.03002]

    Pith/arXiv arXiv 2024

  45. [45]

    Ville,Étude critique de la notion de collectif, Monographies des Probabilités, fasc

    J. Ville,Étude critique de la notion de collectif, Monographies des Probabilités, fasc. 3, Gauthier-Villars, Paris (1939)

    1939

  46. [46]

    Wasserman, A

    L. Wasserman, A. Ramdas and S. Balakrishnan,Universal inference,Proceedings of the National Academy of Science117(2020) 16880 [1912.11436]

    Pith/arXiv arXiv 2020

  47. [47]

    Caldwell and E.V

    R.R. Caldwell and E.V. Linder,Limits of quintessence,Phys. Rev. Lett.95(2005) 141301 [astro-ph/0505494]

    Pith/arXiv arXiv 2005

  48. [48]

    Scherrer and A.A

    R.J. Scherrer and A.A. Sen,Thawing quintessence with a nearly flat potential,Phys. Rev. D77 (2008) 083515 [0712.3450]

    Pith/arXiv arXiv 2008

  49. [49]

    Shlivko and P.J

    D. Shlivko and P.J. Steinhardt,Assessing observational constraints on dark energy,Phys. Lett. B855(2024) 138826 [2405.03933]

    Pith/arXiv arXiv 2024

  50. [50]

    Nesseris and J

    S. Nesseris and J. García-Bellido,Comparative analysis of model-independent methods for exploring the nature of dark energy,Phys. Rev. D88(2013) 063521 [1306.4885]

    Pith/arXiv arXiv 2013

  51. [51]

    Chen, Q.-G

    L. Chen, Q.-G. Huang and K. Wang,Distance Priors from Planck Final Release,JCAP02 (2019) 028 [1808.05724]. – 22 –

    Pith/arXiv arXiv 2019