Pith. sign in

REVIEW 3 major objections 4 minor 4 cited by

Aletheia claims that the non-linear matter power spectrum can be predicted accurately from the linear spectrum's shape plus a single clustering-amplitude number, σ12, with only a small growth-history correction, and validates this on dark-e

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 21:42 UTC pith:IYOOHYWT

load-bearing objection Solid emulator with credible sub-percent claims in-range; the k=2 Mpc^-1 extension rests on an untested shape-independence assumption for the resolution correction. the 3 major comments →

arxiv 2511.13826 v1 pith:IYOOHYWT submitted 2025-11-17 astro-ph.CO

Aletheia: Emulating the non-linear matter power spectrum in the context of evolution mapping

Ariel G. Sanchez , Andr\'es N. Ruiz , Facundo Rodriguez , Carlos Correa , Andrea Fiorilli , Matteo Esposito , Jenny Gonzalez-Jara , Nelson D. Padilla
show 2 more authors
Alejandro P\'erez-Fern\'andez Sofia Contarini
This is my paper
classification astro-ph.CO
keywords matter power spectrumnon-linear structure growthemulatorevolution mappingGaussian process regressiondark energy equation of statecosmological simulationssigma12
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.

The paper tries to establish that the full non-linear matter power spectrum can be emulated from far fewer inputs than conventional emulators require. It separates cosmological parameters into shape parameters, which fix the form of the linear spectrum, and evolution parameters, which only change its amplitude, and then compresses all evolution-parameter and redshift effects into a single amplitude number, σ12. A second, small correction accounts for differences in growth history through an integrated parameter, x~. The payoff is sub-percent predictions across a parameter range that includes dynamic dark-energy models outside most existing emulators' training boxes, validated against N-body simulations.

Core claim

Aletheia demonstrates that the evolution mapping principle can be turned into a practical emulator: for fixed reference evolution parameters, the non-linear boost factor B(k) — the ratio of the non-linear power spectrum to its de-wiggled linear counterpart — is a smooth function only of the shape parameters and σ12. Deviations when evolution parameters differ are captured by a first-order Taylor expansion in the integrated growth-history parameter x~, whose slope is emulated separately. The final prediction is the de-wiggled linear spectrum of the target cosmology times the emulated boost times a linear growth-history correction. Validated on independent simulations, this two-stage construct

What carries the argument

The central object is the evolution-mapping parametrisation: h-independent shape parameters Θs = (ωb, ωc, ns) are separated from evolution parameters, and all evolution/redshift dependence is folded into σ12, the RMS linear fluctuation smoothed over spheres of radius 12 Mpc. The small residual physics is encoded in x~, a memory-weighted average of the standard growth factor combination Ωm/f^2 over past history, with a Gaussian kernel of width η = 0.12 in ln σ12. Two Gaussian-process emulators with Matérn kernels map (Θs, σ12, ln k) to the boost factor and to the derivative ∂R/∂x~; a two-dimensional spline correction C(k, σ12), calibrated on a high-resolution simulation suite, extends predict

Load-bearing premise

The emulator's high-wavenumber accuracy rests on the assumption that the numerical resolution correction measured for one fiducial cosmology applies unchanged to every cosmology, and only a single non-fiducial test was used to check it.

What would settle it

Run high-resolution N-body pairs (for example 2048^3 particles in a 700 Mpc box) for several training-set cosmologies with differing ωb, ωc, and ns at the same σ12 < 0.83, then compare Aletheia's corrected prediction to the high-resolution spectrum near k = 1–2 Mpc^-1; if the residual varies systematically with shape parameters by more than about 0.5%, the shape-independent resolution correction is not generally valid.

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

If this is right

  • Non-linear P(k) predictions become available for dynamic dark-energy models, including equation-of-state parameters outside conventional emulators' training ranges, without running new N-body simulations.
  • Predictions cover k = 0.006–2 Mpc^-1 at arbitrary query wavenumbers, with high-k accuracy extended by the resolution correction.
  • Because only σ12 and x~ describe the evolution side, no specific dark-energy equation-of-state form is assumed; any smooth dark-energy model with negligible perturbations and scale-independent linear growth is covered.
  • The reduced dimensionality should let future emulators cover wider parameter spaces with the same number of simulations, or reach higher accuracy for a fixed training set.

Where Pith is reading between the lines

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

  • Extension: if the resolution correction truly depends only on k and σ12, the same calibration could be reused across emulator generations, so extending to higher wavenumbers does not multiply the simulation cost by the number of cosmologies emulated.
  • Extension: a cheap validation would re-derive C(k, σ12) from a second high-resolution simulation with different ωb, ωc, and ns; agreement within a few tenths of a percent would close the main gap in the proof, while disagreement would indicate the correction needs shape dependence.
  • Extension: the σ12/x~ compression suggests the same two-step scheme could carry over to other clustering statistics, such as redshift-space galaxy spectra or the bispectrum, effectively reducing the description of non-linear cosmic structure to a low-dimensional manifold.

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

3 major / 4 minor

Summary. The paper presents Aletheia, a Gaussian-process emulator for the non-linear matter power spectrum built on the evolution-mapping framework of Sánchez et al. (2022). Shape parameters (ωb, ωc, ns) and the clustering amplitude σ12 are the primary inputs: a first emulator E_B predicts the boost relative to the de-wiggled linear spectrum for a reference evolution history, while a second emulator E_∂R/∂x̃ predicts the small linear correction arising from the integrated growth-history variable x̃. A resolution-correction spline C_s(k, σ12), derived from high-resolution fiducial simulations, extends the final predictions to k = 2 Mpc⁻¹. The authors validate the component emulators on 50 held-out cosmologies, compare the full emulator against EuclidEmulator2 on ten dynamic-dark-energy N-body runs, and test the final corrected emulator on a high-resolution DESI best-fit cosmology.

Significance. The evolution-mapping reduction is physically motivated and yields a low-dimensional, smooth emulation problem; the paper provides independent held-out test sets, a public PyPI package, and a clear validation protocol. If the high-k correction and the head-to-head comparison are properly supported, Aletheia would be a timely and wider-range P(k) emulator for w0-wa and more general dark-energy models outside EuclidEmulator2's training range. The main significance therefore hinges on the empirical status of the resolution correction and on whether the variance comparison is pipeline-fair.

major comments (3)
  1. [§3.4, Eq. (15); §4.3, Table 1] The resolution correction C(k, σ12) is computed from the fiducial AletheiaMass cosmology and explicitly assumed to be independent of the shape parameters Θs. The only non-fiducial validation, the DESI row of Table 1, has ωb = 0.02236, ωc = 0.12021, ns = 0.9648 — essentially the Planck fiducial shape — so it does not exercise the shape dependence of C. Because E_final applies this correction uniformly over the full ±5σ training range up to k = 2 Mpc⁻¹, the claimed sub-percent high-k accuracy for non-fiducial shapes is not yet established. I recommend high-resolution runs for at least two corner shapes (e.g., high ωc/low ns and low ωc/high ns) or a conservative restriction of the validated k-range.
  2. [§4.2, Fig. 7] The headline factor-of-five variance comparison uses AletheiaDE simulations produced with the same GADGET-4/2LPTIC pipeline that generated the AletheiaEmu training data. EuclidEmulator2 was calibrated on different codes, resolutions, and initial-condition generators; part of its ~1% variance may therefore reflect pipeline systematics rather than emulator inaccuracy. The comparison would be conclusive only if both emulators were benchmarked against an external simulation suite independent of Aletheia's training pipeline, or if the two pipelines were shown to agree at the ~0.2% level. This is a fairness concern, not a circularity charge: the held-out cosmologies are genuinely independent.
  3. [§4.2 vs §4.3; abstract] The variance comparison and the sub-percent accuracy discussion in §4.2 refer to the uncorrected emulator E_P; the corrected final emulator E_final is validated only against the single DESI high-resolution cosmology in §4.3. The abstract does not distinguish these. Please state explicitly which quantity the headline claims refer to and provide a composite accuracy statement for E_final across the shape range, or revise the claims accordingly.
minor comments (4)
  1. [Fig. 10 caption] Typo: 'Secton' should be 'Section'.
  2. [§2.2, Eq. (4)] The choice η = 0.12 is justified only by 'we find'; a sensitivity test over a plausible η range would strengthen the claim that the x̃ definition is robust.
  3. [§3.4 and Fig. 5] The uncertainty in the smoothing spline C_s is not propagated into E_final. A one-sigma band in Fig. 5, or a note on its size, would clarify the precision of the high-k correction.
  4. [Eq. (12)] The notation E_∂R/∂x̃ is awkward in text; using a symbol such as E_deriv or E_R consistently would improve readability.

Circularity Check

0 steps flagged

No significant circularity: Aletheia's accuracy claims are validated on independent N-body simulations, not reduced to fitted inputs.

full rationale

The core derivation is self-contained. The two-stage emulators E_B and E_dR/dx are trained on 100 AletheiaEmu nodes and then checked against 50 test cosmologies excluded from training; the full emulator is further tested on the AletheiaDE dynamic-dark-energy suite and on a separate DESI best-fit high-resolution run. Equation (5) is an empirically tested first-order Taylor expansion: the linearity of R(k,x~) is shown directly from measured simulation ratios in Fig. 2, and the derivative is measured from the same simulations rather than imposed. Equation (12) multiplies the GP-predicted boost and derivative by target-cosmology linear-theory quantities (P_DW, sigma12, x~), so no prediction is equal by construction to a fitted parameter. The resolution correction C(k,sigma12) in Eq. (15) is calibrated at a fiducial cosmology and assumed shape-independent; this is an acknowledged correctness limitation (Section 3.4, and its Section 4.3 check uses a near-fiducial DESI shape), but it is an application assumption, not a circular reduction, because the corrected result is still compared to independent high-resolution simulations. The self-citations to Sanchez et al. (2022) and Esposito et al. (2024) provide lineage and simulation data, but the paper re-establishes the evolution-mapping relations in its own figures; eta=0.12 is a tuned kernel width, not a predicted quantity. Overall, no load-bearing circular step was found.

Axiom & Free-Parameter Ledger

3 free parameters · 6 axioms · 1 invented entities

The paper contributes an empirical emulator, so its load-bearing assumptions are mostly domain assumptions about the validity of the evolution-mapping compression and the first-order growth-history correction, plus two fitted elements: η and the GP/spline hyperparameters. No fundamentally new physical entities are introduced; x̃ is a derived summary variable with predictive validation.

free parameters (3)
  • Kernel width η of x̃ memory kernel = 0.12
    Tuned in §2.2 to provide a good fit to deviations from evolution mapping; no uncertainty or cross-validation given.
  • Gaussian Process hyperparameters (σ_f² and length scales ℓ_i) = not reported
    Maximized marginal likelihood for E_B and E_dR/dx̃; standard but free.
  • Resolution-correction spline C_s(k,σ12) = not reported
    Smoothing spline fit to high-resolution AletheiaMass ratios in §3.4, assumed independent of shape parameters.
axioms (6)
  • domain assumption Non-linear evolution mapping: P(k|Θs,z,Θe) is approximately P(k|Θs,σ12) for fixed shape parameters.
    Central premise, inherited from Sánchez et al. 2022 and re-shown in Fig. 1; empirical, not derived.
  • domain assumption The first-order Taylor expansion of the P(k) ratio in x̃ (Eq. 5) is accurate.
    Empirically validated in Fig. 2; underpins the second-stage correction.
  • domain assumption Growth-history effects are captured by x = Ωm(z)/f²(z) smoothed over τ=ln σ12 with width η.
    Motivated by SPT; the kernel width η is tuned ad hoc.
  • ad hoc to paper Resolution correction depends only on k and σ12, not on shape parameters.
    Explicit assumption in §3.4; only one non-fiducial cosmology tested in §4.3.
  • domain assumption Dark energy perturbations are negligible and linear growth is scale-independent.
    Limit of applicability stated in §4.2 and §5; excludes models with DE clustering or scale-dependent growth.
  • domain assumption GADGET-4 plus 2LPTIC simulations provide unbiased estimates of the true non-linear P(k).
    All training/testing power spectra come from this toolchain; any shared systematics are absorbed into the emulator.
invented entities (1)
  • Integrated growth-history parameter x̃ with Gaussian memory kernel independent evidence
    purpose: Quantify deviations from exact evolution mapping due to recent growth history; input for second-stage linear correction.
    Introduced in Eq. (4); it is a derived summary variable rather than a physical entity, but it is validated by predicting held-out simulation ratios.

pith-pipeline@v1.3.0-alltime-deepseek · 26404 in / 14021 out tokens · 134205 ms · 2026-08-03T21:42:54.216436+00:00 · methodology

0 comments
read the original abstract

We present Aletheia, a new emulator of the non-linear matter power spectrum, $P(k)$, built upon the evolution mapping framework. This framework addresses the limitations of traditional emulation by focusing on $h$-independent cosmological parameters, which can be separated into those defining the linear power spectrum shape ($\mathbf{\Theta}_{\mathrm{s}}$) and those affecting only its amplitude evolution ($\mathbf{\Theta}_{\mathrm{e}}$). The combined impact of evolution parameters and redshift is compressed into a single amplitude parameter, $\sigma_{12}$. Aletheia uses a two-stage Gaussian Process emulation: a primary emulator predicts the non-linear boost factor as a function of ($\mathbf{\Theta}_{\mathrm{s}}$) and $\sigma_{12}$ for fixed evolution parameters, while a second one applies a small linear correction based on the integrated growth history. The emulator is trained on shape parameters spanning $\pm$5$\sigma$ of Planck constraints and a wide clustering range $0.2 < \sigma_{12} < 1.0$, providing predictions for $0.006\,{\rm Mpc}^{-1} < k < 2\,{\rm Mpc}^{-1}$. We validate Aletheia against N-body simulations, demonstrating sub-percent accuracy. When tested on a suite of dynamic dark energy models, the full emulator's predictions show a variance of approximately 0.2%, a factor of five smaller than that of the state-of-the-art EuclidEmulator2 (around 1% variance). Furthermore, Aletheia maintains sub-percent accuracy for the best-fit dynamic dark energy cosmology from recent DESI data, a model whose parameters lie outside the training ranges of most conventional emulators. This demonstrates the power of the evolution mapping approach, providing a robust and extensible tool for precision cosmology.

Figures

Figures reproduced from arXiv: 2511.13826 by Alejandro P\'erez-Fern\'andez, Andrea Fiorilli, Andr\'es N. Ruiz, Ariel G. Sanchez, Carlos Correa, Facundo Rodriguez, Jenny Gonzalez-Jara, Matteo Esposito, Nelson D. Padilla, Sofia Contarini.

Figure 1
Figure 1. Figure 1: The left panel shows the ratio of non-linear power spectra for a set of simulations with identical shape parameters but widely varying evolution parameters, all evaluated at redshifts that correspond to the same values of 𝜎12. The specific cosmological parameters for each model are detailed in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The ratio of power spectra, 𝑅(𝑘), as a function of the integrated growth history parameter 𝑥˜ for four choices of 𝑘 and 𝜎12, as indicated in the legend. The points show the measurements from the Aletheia simulations, where each colour represents a different cosmology (with parameters detailed in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of cosmological parameters for the training (blue points) and testing (orange points) sets of simulations used for E𝐵 (𝑘). The panels show 2D projections of the parameter space (𝜔b, 𝜔c, 𝑛s , 𝜎12 ). The grey ellipses represent the 1𝜎 and 2𝜎 confidence regions derived from Planck 2018 data for the shape parameters (𝜔b, 𝜔c, 𝑛s). Our sampling strat￾egy uses the eigenvector directions of the Planck… view at source ↗
Figure 4
Figure 4. Figure 4: Raw training data for the Aletheia emulators. The left panel shows the measured boost factor, 𝐵(𝑘), for the 100 simulations to train E𝐵 (𝑘), colour￾coded by the clustering amplitude 𝜎12 at which they were evaluated. This panel illustrates the strong, smooth dependence of the boost factor on 𝜎12. The right panel displays the measured derivative term 𝜕𝑅/𝜕𝑥˜ for the same cosmologies, determined from the deriv… view at source ↗
Figure 5
Figure 5. Figure 5: Correction of the Aletheia emulator for resolution effects. The solid lines show the measured ratio of the uncorrected emulator prediction, E𝑃 (𝑘), to the high-resolution power spectrum, 𝑃HR (𝑘), derived from the AletheiaMass simulations. The lines are colour-coded by the value of 𝜎12. The fluctuations are caused by cosmic variance due to the smaller box size of the high-resolution runs. The dashed lines r… view at source ↗
Figure 6
Figure 6. Figure 6: Performance of the individual emulator components of Aletheia. The left panel shows the relative error of the E𝐵 (𝑘) emulator, (𝑃emu/𝑃sim − 1), with the shaded region indicating the 1𝜎 variance across the test set. The right panel displays the absolute error of the Ed𝑅/d ˜𝑥 (𝑘) emulator. Both emulators demonstrate high accuracy across the relevant wavenumber range, with the error of E𝐵 (𝑘) being consistent… view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of the performance of the full Aletheia emulator against N-body simulations and EuclidEmulator2. Each panel shows the mean and variance of the ratio E𝑃 (𝑘)/𝑃(𝑘) across ten test cosmologies for various redshifts computed using Aletheia (red) and EuclidEmulator2 (blue). Both emulators show excellent agreement with the simulations, with deviations largely driven by cosmic variance. Aletheia exhibit… view at source ↗
Figure 8
Figure 8. Figure 8: Non-linear matter power spectrum for the best-fitting dynamic dark energy cosmology of DESI Collaboration et al. (2025b). N-body simulation results (black solid lines) at six redshifts are compared against the predictions of Aletheia (red dashed lines). The linear-theory predictions (grey dotted lines) illustrate the impact of non-linear evolution. The evolution mapping de￾sign allows Aletheia to provide r… view at source ↗
Figure 10
Figure 10. Figure 10: The ratio of the Aletheia predictions to the power spectra in￾ferred from high-resolution simulations at varying redshifts. The dot-dashed lines show the results of the uncorrected emulator prediction E𝑃 (𝑘), which underestimates the power at high 𝑘 for low clustering amplitudes. The solid lines show the results of the final emulator prediction Efinal(𝑘), including the resolution correction described in S… view at source ↗

discussion (0)

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

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Improved recipes for peculiar velocity power spectra using Evolution Mapping

    astro-ph.CO 2026-05 accept novelty 7.0

    Improved fitting functions for P_θθ(k) and P_δθ(k) parametrized by σ12 achieve 1-2% accuracy across cosmologies and outperform existing prescriptions.

  2. Evolution mapping III: A new recipe for the halo mass function

    astro-ph.CO 2025-11 conditional novelty 6.0

    A new 9-parameter fitting function for the halo mass function, using the integrated growth history (x-tilde) and effective spectral index (n_eff), reaches 1–5% accuracy across a wide range of cosmologies and overdensi...

  3. Dark Quest II: A Wide-Coverage Neural Network Emulator of the Nonlinear Matter Power Spectrum Across Extended Cosmologies

    astro-ph.CO 2026-05 unverdicted novelty 5.0

    A neural network emulator trained on multi-resolution N-body simulations reproduces the nonlinear matter power spectrum to subpercent accuracy up to the Nyquist scale across an extended nine-dimensional cosmological p...

  4. GINKAKU: Scalable Cosmological Structure Formation Simulation Code and Post-processing Pipeline

    astro-ph.CO 2026-05 unverdicted novelty 5.0

    GINKAKU is a scalable TreePM N-body code with linear-response external sources, validated to ~1% power-spectrum agreement with GADGET/PKDGRAV3/RAMSES and applied to DQ2 production runs with improved halo post-processing.

Reference graph

Works this paper leans on

58 extracted references · 5 canonical work pages · cited by 4 Pith papers

  1. [1]

    Abbott T. M. C., et al., 2022, @doi [ ] 10.1103/PhysRevD.105.023520 , https://ui.adsabs.harvard.edu/abs/2022PhRvD.105b3520A 105, 023520

    doi:10.1103/physrevd.105.023520 2022

  2. [2]

    Alam S., et al., 2017, @doi [ ] 10.1093/mnras/stx721 , 470, 2617–2652

    doi:10.1093/mnras/stx721 2017

  3. [3]

    Alam S., et al., 2021, @doi [ ] 10.1103/PhysRevD.103.083533 , https://ui.adsabs.harvard.edu/abs/2021PhRvD.103h3533A 103, 083533

    doi:10.1103/physrevd.103.083533 2021

  4. [4]

    Anderson L., et al., 2014, @doi [ ] 10.1093/mnras/stu523 , http://adsabs.harvard.edu/abs/2014MNRAS.441...24A 441, 24

    doi:10.1093/mnras/stu523 2014

  5. [5]

    E., Pontzen A., 2016, @doi [ ] 10.1093/mnrasl/slw098 , https://ui.adsabs.harvard.edu/abs/2016MNRAS.462L...1A 462, L1

    Angulo R. E., Pontzen A., 2016, @doi [ ] 10.1093/mnrasl/slw098 , https://ui.adsabs.harvard.edu/abs/2016MNRAS.462L...1A 462, L1

    doi:10.1093/mnrasl/slw098 2016

  6. [6]

    E., Zennaro M., Contreras S., Aric \'o G., Pellejero-Iba \ n ez M., St \"u cker J., 2021, @doi [ ] 10.1093/mnras/stab2018 , https://ui.adsabs.harvard.edu/abs/2021MNRAS.tmp.1931A

    Angulo R. E., Zennaro M., Contreras S., Aric \'o G., Pellejero-Iba \ n ez M., St \"u cker J., 2021, @doi [ ] 10.1093/mnras/stab2018 , https://ui.adsabs.harvard.edu/abs/2021MNRAS.tmp.1931A

    doi:10.1093/mnras/stab2018 2021

  7. [7]

    Bocquet S., Heitmann K., Habib S., Lawrence E., Uram T., Frontiere N., Pope A., Finkel H., 2020, @doi [ ] 10.3847/1538-4357/abac5c , https://ui.adsabs.harvard.edu/abs/2020ApJ...901....5B 901, 5

    doi:10.3847/1538-4357/abac5c 2020

  8. [8]

    Brout D., et al., 2022, @doi [ ] 10.3847/1538-4357/ac8e04 , https://ui.adsabs.harvard.edu/abs/2022ApJ...938..110B 938, 110

    doi:10.3847/1538-4357/ac8e04 2022

  9. [9]

    Chevallier M., Polarski D., 2001, @doi [International Journal of Modern Physics D] 10.1142/S0218271801000822 , http://adsabs.harvard.edu/abs/2001IJMPD..10..213C 10, 213

    doi:10.1142/s0218271801000822 2001

  10. [11]

    Crocce M., Pueblas S., Scoccimarro R., 2006, @doi [ ] 10.1111/j.1365-2966.2006.11040.x , https://ui.adsabs.harvard.edu/abs/2006MNRAS.373..369C 373, 369

    arXiv 2006

  11. [12]

    Crocce M., Pueblas S., Scoccimarro R., 2012, 2LPTIC: 2nd-order Lagrangian Perturbation Theory Initial Conditions , Astrophysics Source Code Library, record ascl:1201.005

    2012

  12. [13]

    arXiv:2503.14738

    DESI Collaboration et al., 2025a, @doi [arXiv e-prints] 10.48550/arXiv.2503.14738 , https://ui.adsabs.harvard.edu/abs/2025arXiv250314738D p. arXiv:2503.14738

    Pith/arXiv arXiv doi:10.48550/arxiv.2503.14738

  13. [14]

    DESI Collaboration et al., 2025b, @doi [ ] 10.1088/1475-7516/2025/02/021 , https://ui.adsabs.harvard.edu/abs/2025JCAP...02..021A 2025, 021

    doi:10.1088/1475-7516/2025/02/021 2025

  14. [15]

    DESI Collaboration et al., 2025c, @doi [ ] 10.1088/1475-7516/2025/07/028 , https://ui.adsabs.harvard.edu/abs/2025JCAP...07..028A 2025, 028

    doi:10.1088/1475-7516/2025/07/028 2025

  15. [16]

    Dalal R., et al., 2023, @doi [ ] 10.1103/PhysRevD.108.123519 , https://ui.adsabs.harvard.edu/abs/2023PhRvD.108l3519D 108, 123519

    doi:10.1103/physrevd.108.123519 2023

  16. [17]

    DeRose J., et al., 2019, @doi [ ] 10.3847/1538-4357/ab1085 , https://ui.adsabs.harvard.edu/abs/2019ApJ...875...69D 875, 69

    doi:10.3847/1538-4357/ab1085 2019

  17. [18]

    G., 2023, @doi [ ] 10.1093/mnras/stac3667 , https://ui.adsabs.harvard.edu/abs/2023MNRAS.519.2962E 519, 2962

    Eggemeier A., Camacho-Quevedo B., Pezzotta A., Crocce M., Scoccimarro R., S \'a nchez A. G., 2023, @doi [ ] 10.1093/mnras/stac3667 , https://ui.adsabs.harvard.edu/abs/2023MNRAS.519.2962E 519, 2962

    doi:10.1093/mnras/stac3667 2023

  18. [19]

    J., et al., 2005, @doi [ ] 10.1086/466512 , https://ui.adsabs.harvard.edu/abs/2005ApJ...633..560E 633, 560

    Eisenstein D. J., et al., 2005, @doi [ ] 10.1086/466512 , https://ui.adsabs.harvard.edu/abs/2005ApJ...633..560E 633, 560

    doi:10.1086/466512 2005

  19. [20]

    G., Bel J., Ruiz A

    Esposito M., S \'a nchez A. G., Bel J., Ruiz A. N., 2024, @doi [ ] 10.1093/mnras/stae2351 , https://ui.adsabs.harvard.edu/abs/2024MNRAS.534.3906E 534, 3906

    doi:10.1093/mnras/stae2351 2024

  20. [21]

    Euclid Collaboration et al., 2019, @doi [ ] 10.1093/mnras/stz197 , https://ui.adsabs.harvard.edu/abs/2019MNRAS.484.5509E 484, 5509

    doi:10.1093/mnras/stz197 2019

  21. [22]

    Euclid Collaboration et al., 2021, @doi [ ] 10.1093/mnras/stab1366 , https://ui.adsabs.harvard.edu/abs/2021MNRAS.505.2840E 505, 2840

    doi:10.1093/mnras/stab1366 2021

  22. [23]

    Euclid Collaboration et al., 2025, @doi [ ] 10.1051/0004-6361/202450810 , https://ui.adsabs.harvard.edu/abs/2025A&A...697A...1E 697, A1

    doi:10.1051/0004-6361/202450810 2025

  23. [24]

    Garny M., Taule P., 2021, @doi [ ] 10.1088/1475-7516/2021/01/020 , https://ui.adsabs.harvard.edu/abs/2021JCAP...01..020G 2021, 020

    doi:10.1088/1475-7516/2021/01/020 2021

  24. [25]

    H., Eisenstein D

    Garrison L. H., Eisenstein D. J., Ferrer D., Tinker J. L., Pinto P. A., Weinberg D. H., 2018, @doi [The Astrophysical Journal Supplement Series] 10.3847/1538-4365/aabfd3 , https://ui.adsabs.harvard.edu/abs/2018ApJS..236...43G 236, 43

    work page doi:10.3847/1538-4365/aabfd3 2018

  25. [26]

    N., et al., 2017, @doi [ ] 10.1093/mnras/stw3384 , http://adsabs.harvard.edu/abs/2017MNRAS.467.2085G 467, 2085

    Grieb J. N., et al., 2017, @doi [ ] 10.1093/mnras/stw3384 , http://adsabs.harvard.edu/abs/2017MNRAS.467.2085G 467, 2085

    work page doi:10.1093/mnras/stw3384 2017

  26. [27]

    Hamann J., Hannestad S., Lesgourgues J., Rampf C., Wong Y. Y. Y., 2010, @doi [ ] 10.1088/1475-7516/2010/07/022 , https://ui.adsabs.harvard.edu/abs/2010JCAP...07..022H 2010, 022

    doi:10.1088/1475-7516/2010/07/022 2010

  27. [28]

    Hamilton A. J. S., Kumar P., Lu E., Matthews A., 1991, @doi [ ] 10.1086/186057 , https://ui.adsabs.harvard.edu/abs/1991ApJ...374L...1H 374, L1

    doi:10.1086/186057 1991

  28. [29]

    Heitmann K., White M., Wagner C., Habib S., Higdon D., 2010, @doi [ ] 10.1088/0004-637X/715/1/104 , https://ui.adsabs.harvard.edu/abs/2010ApJ...715..104H 715, 104

    doi:10.1088/0004-637x/715/1/104 2010

  29. [30]

    Heitmann K., et al., 2016, @doi [ ] 10.3847/0004-637X/820/2/108 , https://ui.adsabs.harvard.edu/abs/2016ApJ...820..108H 820, 108

    doi:10.3847/0004-637x/820/2/108 2016

  30. [31]

    W., Eastwood J

    Hockney R. W., Eastwood J. W., 1988, Computer Simulation Using Particles, 1 edn. CRC Press, Boca Raton, FL, @doi 10.1201/9780367806934 , https://doi.org/10.1201/9780367806934

    doi:10.1201/9780367806934 1988

  31. [32]

    M., Simonovi \'c M., Zaldarriaga M., 2020, @doi [ ] 10.1088/1475-7516/2020/05/042 , https://ui.adsabs.harvard.edu/abs/2020JCAP...05..042I 2020, 042

    Ivanov M. M., Simonovi \'c M., Zaldarriaga M., 2020, @doi [ ] 10.1088/1475-7516/2020/05/042 , https://ui.adsabs.harvard.edu/abs/2020JCAP...05..042I 2020, 042

    doi:10.1088/1475-7516/2020/05/042 2020

  32. [33]

    Li X., et al., 2023, @doi [ ] 10.1103/PhysRevD.108.123518 , https://ui.adsabs.harvard.edu/abs/2023PhRvD.108l3518L 108, 123518

    doi:10.1103/physrevd.108.123518 2023

  33. [34]

    V., 2003, @doi [Physical Review Letters] 10.1103/PhysRevLett.90.091301 , http://adsabs.harvard.edu/abs/2003PhRvL..90i1301L 90, 091301

    Linder E. V., 2003, @doi [Physical Review Letters] 10.1103/PhysRevLett.90.091301 , http://adsabs.harvard.edu/abs/2003PhRvL..90i1301L 90, 091301

    doi:10.1103/physrevlett.90.091301 2003

  34. [35]

    arXiv:2503.14452

    Louis T., et al., 2025, @doi [arXiv e-prints] 10.48550/arXiv.2503.14452 , https://ui.adsabs.harvard.edu/abs/2025arXiv250314452L p. arXiv:2503.14452

    Pith/arXiv arXiv doi:10.48550/arxiv.2503.14452 2025

  35. [36]

    D., Beckman R

    McKay M. D., Beckman R. J., Conover W. J., 1979, Technometrics, 21, 239

    1979

  36. [37]

    A., Dodds S

    Peacock J. A., Dodds S. J., 1996, @doi [ ] 10.1093/mnras/280.3.L19 , https://ui.adsabs.harvard.edu/abs/1996MNRAS.280L..19P 280, L19

    work page doi:10.1093/mnras/280.3.l19 1996

  37. [38]

    Pedregosa F., et al., 2011, @doi [Journal of Machine Learning Research] 10.48550/arXiv.1201.0490 , https://ui.adsabs.harvard.edu/abs/2011JMLR...12.2825P 12, 2825

    Pith/arXiv arXiv doi:10.48550/arxiv.1201.0490 2011

  38. [39]

    Pezzotta A., et al., 2025, @doi [ ] 10.1103/vy3h-p92n , https://ui.adsabs.harvard.edu/abs/2025PhRvD.112b3520P 112, 023520

    doi:10.1103/vy3h-p92n 2025

  39. [40]

    Planck Collaboration et al., 2020, @doi [ ] 10.1051/0004-6361/201833910 , https://ui.adsabs.harvard.edu/abs/2020A&A...641A...6P 641, A6

    doi:10.1051/0004-6361/201833910 2020

  40. [41]

    E., Williams C

    Rasmussen C. E., Williams C. K. I., 2006, Gaussian Processes for Machine Learning . The MIT Press

    2006

  41. [42]

    G., 2020, @doi [ ] 10.1103/PhysRevD.102.123511 , https://ui.adsabs.harvard.edu/abs/2020PhRvD.102l3511S 102, 123511

    S \'a nchez A. G., 2020, @doi [ ] 10.1103/PhysRevD.102.123511 , https://ui.adsabs.harvard.edu/abs/2020PhRvD.102l3511S 102, 123511

    doi:10.1103/physrevd.102.123511 2020

  42. [43]

    G., Baugh C

    S \'a nchez A. G., Baugh C. M., Percival W. J., Peacock J. A., Padilla N. D., Cole S., Frenk C. S., Norberg P., 2006, @doi [ ] 10.1111/j.1365-2966.2005.09833.x , https://ui.adsabs.harvard.edu/abs/2006MNRAS.366..189S 366, 189

    arXiv 2006

  43. [44]

    G., et al., 2017, @doi [ ] 10.1093/mnras/stw2495 , http://adsabs.harvard.edu/abs/2017MNRAS.464.1493S 464, 1493

    S \'a nchez A. G., et al., 2017, @doi [ ] 10.1093/mnras/stw2495 , http://adsabs.harvard.edu/abs/2017MNRAS.464.1493S 464, 1493

    doi:10.1093/mnras/stw2495 2017

  44. [45]

    G., Ruiz A

    S \'a nchez A. G., Ruiz A. N., Jara J. G., Padilla N. D., 2022, @doi [ ] 10.1093/mnras/stac1656 , https://ui.adsabs.harvard.edu/abs/2022MNRAS.514.5673S 514, 5673

    doi:10.1093/mnras/stac1656 2022

  45. [46]

    N., Frieman J

    Scoccimarro R., Colombi S., Fry J. N., Frieman J. A., Hivon E., Melott A., 1998, @doi [ ] 10.1086/305399 , https://ui.adsabs.harvard.edu/abs/1998ApJ...496..586S 496, 586

    doi:10.1086/305399 1998

  46. [47]

    M., et al., 2018, @doi [ ] 10.3847/1538-4357/aab9bb , https://ui.adsabs.harvard.edu/abs/2018ApJ...859..101S 859, 101

    Scolnic D. M., et al., 2018, @doi [ ] 10.3847/1538-4357/aab9bb , https://ui.adsabs.harvard.edu/abs/2018ApJ...859..101S 859, 101

    doi:10.3847/1538-4357/aab9bb 2018

  47. [48]

    Scolnic D., et al., 2022, @doi [ ] 10.3847/1538-4357/ac8b7a , https://ui.adsabs.harvard.edu/abs/2022ApJ...938..113S 938, 113

    doi:10.3847/1538-4357/ac8b7a 2022

  48. [49]

    Sefusatti E., Crocce M., Scoccimarro R., Couchman H. M. P., 2016, @doi [ ] 10.1093/mnras/stw1229 , https://ui.adsabs.harvard.edu/abs/2016MNRAS.460.3624S 460, 3624

    doi:10.1093/mnras/stw1229 2016

  49. [50]

    E., et al., 2003, @doi [ ] 10.1046/j.1365-8711.2003.06503.x , https://ui.adsabs.harvard.edu/abs/2003MNRAS.341.1311S 341, 1311

    Smith R. E., et al., 2003, @doi [ ] 10.1046/j.1365-8711.2003.06503.x , https://ui.adsabs.harvard.edu/abs/2003MNRAS.341.1311S 341, 1311

    arXiv 2003

  50. [51]

    Springel V., Pakmor R., Zier O., Reinecke M., 2021, @doi [ ] 10.1093/mnras/stab1855 , https://ui.adsabs.harvard.edu/abs/2021MNRAS.506.2871S 506, 2871

    doi:10.1093/mnras/stab1855 2021

  51. [52]

    Stein M., 1987, Technometrics, 29, 143151

    1987

  52. [53]

    Takahashi R., 2008, @doi [Progress of Theoretical Physics] 10.1143/PTP.120.549 , https://ui.adsabs.harvard.edu/abs/2008PThPh.120..549T 120, 549

    work page doi:10.1143/ptp.120.549 2008

  53. [54]

    Taruya A., 2016, @doi [ ] 10.1103/PhysRevD.94.023504 , https://ui.adsabs.harvard.edu/abs/2016PhRvD..94b3504T 94, 023504

    doi:10.1103/physrevd.94.023504 2016

  54. [55]

    Tristram M., et al., 2024, @doi [ ] 10.1051/0004-6361/202348015 , https://ui.adsabs.harvard.edu/abs/2024A&A...682A..37T 682, A37

    doi:10.1051/0004-6361/202348015 2024

  55. [56]

    Tr \"o ster T., et al., 2020, @doi [ ] 10.1051/0004-6361/201936772 , https://ui.adsabs.harvard.edu/abs/2020A&A...633L..10T 633, L10

    doi:10.1051/0004-6361/201936772 2020

  56. [57]

    H., et al., 2025, @doi [arXiv e-prints] 10.48550/arXiv.2503.19441 , https://ui.adsabs.harvard.edu/abs/2025arXiv250319441W p

    Wright A. H., et al., 2025, @doi [arXiv e-prints] 10.48550/arXiv.2503.19441 , https://ui.adsabs.harvard.edu/abs/2025arXiv250319441W p. arXiv:2503.19441

    Pith/arXiv arXiv doi:10.48550/arxiv.2503.19441 2025

  57. [58]

    Yang Y., Bird S., Ho M.-F., 2025, @doi [ ] 10.1103/PhysRevD.111.083529 , https://ui.adsabs.harvard.edu/abs/2025PhRvD.111h3529Y 111, 083529

    work page doi:10.1103/physrevd.111.083529 2025

  58. [59]

    d'Amico G., Gleyzes J., Kokron N., Markovic K., Senatore L., Zhang P., Beutler F., Gil-Mar \' n H., 2020, @doi [ ] 10.1088/1475-7516/2020/05/005 , https://ui.adsabs.harvard.edu/abs/2020JCAP...05..005D 2020, 005

    doi:10.1088/1475-7516/2020/05/005 2020