pith. machine review for the scientific record. sign in

arxiv: 2604.19746 · v2 · submitted 2026-04-21 · 🌌 astro-ph.CO

Recognition: unknown

Calibration-Induced Systematics in SALT3 Training and Their Impact on Dark Energy Constraints from Stage IV Supernova Surveys

Kene Anumba , David O. Jones , Richard Kessler , Daniel Scolnic , W. D'Arcy Kenworthy , Rebecca C. Chen , Bastien Carreres , Maria Vincenzi
show 8 more authors
Erik R. Peterson Maria Acevedo Ben Rose Dillon Brout Jillian Paulin Rujuta A. Purohit Rebekah Hounsell The Roman Supernova Cosmology Project Infrastructure Team
Authors on Pith no claims yet

Pith reviewed 2026-05-10 01:14 UTC · model grok-4.3

classification 🌌 astro-ph.CO
keywords supernovaeSALT3calibrationdark energysystematicsLSSTRomanfigure of merit
0
0 comments X

The pith

Small calibration errors during light-curve fitting reduce the dark energy figure of merit by 50 percent for next-generation surveys

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

This paper uses simulations of data from the Vera Rubin Observatory and the Roman Space Telescope to test how photometric calibration uncertainties propagate through the SALT3 model for Type Ia supernovae. It applies shifts to zero points and filter wavelengths and tracks their separate impacts when occurring during model training and when occurring during the fitting of light curves to estimate distances. The key finding is that uncertainties in the fitting stage cause about 50% loss in the dark energy figure of merit while the same uncertainties in training cause only 13% loss, and that the fitting effects are hard to mitigate because they correlate with cosmological parameters. This distinction is important for allocating resources to improve calibration in large surveys that will observe over a million supernovae.

Core claim

Zero-point shifts of 5 mmag and filter mean wavelength shifts of 5 angstrom lead to a ∼50% decrease in the FoM relative to a statistical-only case when calibration uncertainties are propagated only through light-curve fitting. The same calibration shifts applied only during model training produce a smaller ∼13% degradation. Contrary to previous analyses, calibration uncertainties in light-curve fitting dominate over those from model training. Their effect during light-curve fitting varies smoothly with redshift and is nearly degenerate with cosmology, preventing mitigation through self-calibration. Finally, the FoM dependence on the size of the calibration uncertainties is roughly linear.

What carries the argument

SALT3 spectro-photometric model with calibration perturbations (zero-point and filter wavelength shifts) propagated separately through training and light-curve fitting stages

If this is right

  • Calibration uncertainties in light-curve fitting dominate the systematic error budget for dark energy measurements.
  • The smooth redshift variation of fitting errors makes them nearly degenerate with dark energy equation-of-state parameters.
  • The degradation of the figure of merit scales roughly linearly with the amplitude of calibration shifts.
  • Self-calibration methods are ineffective against these fitting-stage systematics.

Where Pith is reading between the lines

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

  • Efforts to improve on-sky calibration monitoring could disproportionately benefit cosmological constraints by targeting the fitting stage.
  • The linear dependence implies that incremental calibration improvements will yield proportional gains in precision.
  • This separation of training and fitting effects could be tested with other models like SALT2 to see if the dominance of fitting holds generally.
  • Combining supernova data with other probes might help break the degeneracy between calibration errors and cosmology.

Load-bearing premise

The sizes of the applied zero-point and wavelength shifts and the characteristics of the simulated survey data accurately represent the real-world calibration uncertainties without introducing unaccounted biases.

What would settle it

Reanalyzing the problem with real calibration data from current or future surveys or with varied shift sizes to check whether the reported 50% and 13% FoM degradations persist.

Figures

Figures reproduced from arXiv: 2604.19746 by Bastien Carreres, Ben Rose, Daniel Scolnic, David O. Jones, Dillon Brout, Erik R. Peterson, Jillian Paulin, Kene Anumba, Maria Acevedo, Maria Vincenzi, Rebecca C. Chen, Rebekah Hounsell, Richard Kessler, Rujuta A. Purohit, The Roman Supernova Cosmology Project Infrastructure Team, W. D'Arcy Kenworthy.

Figure 1
Figure 1. Figure 1: Comparison of LSST and Roman Space Telescope filter transmission curves overlaid on SN Ia spectra at low and high redshift. The figure illustrates the wavelength coverage of each filter set and how they sample the SN Ia spectra at different redshifts, highlighting Roman’s extended near-infrared sensitivity which is essential for observing high-redshift SNe Ia. 6. Systematic accounting: For each calibration… view at source ↗
Figure 2
Figure 2. Figure 2: The gray-colored distributions illustrate the number of SNe Ia with respect to redshift that we simulated for both LSST (left) and Roman (right). The y-axis is in log scale. To ensure high-quality simulated data for training, we applied several selection cuts on the light curves and spectra. The final training set consisted of 1039 LSST SNe, 2491 Roman SNe and 4388 Roman prism spectra. Our chosen redshift … view at source ↗
Figure 3
Figure 3. Figure 3: Number of light curves from all surveys used in model training to constrain each phase/wavelength bin. Photometric and spectral coverage are shown in the top and bottom panels, respectively. For each plot, the broad lines are due to filter transitions. core-collapse supernovae such as IIP, IIL, Ib, Ic (see K25 for details on the source models for these simu￾lations). The only selection applied to these sam… view at source ↗
Figure 4
Figure 4. Figure 4: The top panels illustrate the accuracy of recovering the flux surfaces, M0 (left) and M1 (right), which correspond to the spectral energy distribution (SED) of a fiducial SN Ia and its first-order correction respectively while the bottom panels show the color law (left) and its difference relative to the input model, SALT3.NIREXT (right). where the true underlying model is known, we use SALT3.NIREXT for th… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of trained model components illus￾trating the effect of random calibration variations shown in different colors. (a) and (b) are M0 (mean spectral compo￾nent) and M1 (first variability component) at peak bright￾ness for different calibration realizations. (c) Fractional change in M0 relative to the nominal model. Fnom refers to the flux for the surface without any systematics. ⟨σF /Fnom⟩ gives t… view at source ↗
Figure 6
Figure 6. Figure 6: Redshift trend of distance modulus residuals relative to the nominal (no-systematics) case for all the analysis configurations. The y-axis label is as defined in Equation 12. In each panel, the light blue curves correspond to individual systematic realizations while the black points show the mean residual in redshift bins. The labels R, L, LR, Fixed and Random are as defined in the beginning of this sectio… view at source ↗
Figure 7
Figure 7. Figure 7: Mean χ 2 from quadratic fits to ∆µsyst among the analysis variants in [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Contour plots for Flatw0waCDM model from the simulation. We show statistical uncertainties and calibration systematics for Train+Fit(LR) analysis. the Roman WIDE and DEEP fields results in compara￾ble FoM reduction. These trends show the importance of each survey field in constraining cosmological param￾eter with Rubin-LSST DDF playing a critical role. 4. CONCLUSION We analyzed simulated HLTDS + Rubin-LSST… view at source ↗
Figure 10
Figure 10. Figure 10: ⟨FoM⟩ versus systematic shift amplitude. The statistical-only Train+Fit is shown in black, while Cal SALT3 systematics are applied in Train(LR), Fit(LR), and Train+Fit(LR) analysis variants. A shift amplitude of 2 means that a zero-point offset of 2 mmag and filter mean wavelength of 2˚A were applied simultaneously. samples (Betoule et al. 2014; Brout et al. 2022a). We have demonstrated that the reduced s… view at source ↗
Figure 9
Figure 9. Figure 9: Cal SALT3 RFoM showing the resulting degrada￾tion due to perturbations applied to individual filters. Fixed shifts of 5 mmag in zero-point and 5 ˚A in mean wavelength are applied. Train, Fit, and Train+Fit denote perturba￾tions applied during the SALT3 training, light curve fitting, or both stages respectively. Left 3 columns are for zero-point shifts; right 3 columns are for filter-wavelength shifts. vato… view at source ↗
Figure 11
Figure 11. Figure 11: Comparison of ⟨FoM⟩ for Random (upper-left) and Fixed (bottom-left) systematic shifts, and each panel shows Train, Fit, Train+Fit. The right panels show the corresponding average uncertainties on w0 and wa multiplied by 1000 respectively [PITH_FULL_IMAGE:figures/full_fig_p015_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Impact of selectively excluding SNe from spe￾cific survey fields on the ⟨FoM⟩, derived from Train+Fit (LR) analysis. ‘noDDF’ excludes all Rubin-LSST Deep Drilling Field SNe; ‘noWide’ and ‘noDeep’ exclude Roman Wide and Deep field SNe, respectively; ‘All data’ retains the full sample with no exclusions. ‘NOSYS’ denotes statistical uncertain￾ties only, with no systematic errors included. ber 80GSFC24M0006. … view at source ↗
Figure 13
Figure 13. Figure 13: Comparison of fitted light curve parameters and their associated uncertainties for the two models: the trained model - KA25 and the input model - SALT3 NIREXT. Each panel shows the distribution of SN parameter for both models using the same binning. Distributions are plotted on a logarithmic y-axis to highlight differences across wide ranges. This comparison assesses how well the trained model recovers th… view at source ↗
Figure 14
Figure 14. Figure 14: The dependence of FoM on the number model realizations used to sample the Cal SALT3 calibration errors. Increasing the number of realizations improves the characterization of calibration uncertainties until convergence is reached around 60 model realizations, beyond which additional realizations yield diminishing returns relative to the computational cost. BINNED refers to the case where the Hubble diagra… view at source ↗
Figure 15
Figure 15. Figure 15: Redshift trend of distance modulus residuals relative to the nominal case for the different shift amplitudes. The shift amplitudes are annotated in the figures. The first, second and third columns are Train, Fit and Train+Fit respectively [PITH_FULL_IMAGE:figures/full_fig_p020_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Mean standard error of the distance-modulus shift as a function of calibration shift amplitude, for Train, Fit, and Train+Fit systematics. Each amplitude corresponds to equal zero-point and central wavelength shifts; for example, an amplitude of 2 corresponds to 2 mmag and 2 ˚A, respectively [PITH_FULL_IMAGE:figures/full_fig_p021_16.png] view at source ↗
read the original abstract

In the coming years, the Vera Rubin Observatory's Legacy Survey of Space and Time (Rubin-LSST) and the Nancy Grace Roman Space Telescope's (Roman) High Latitude Time Domain Survey (HLTDS) are expected to discover more than a million Type Ia supernovae (SNe Ia), several orders of magnitude more than current samples and with a tighter control on systematic uncertainties. One of the largest systematic uncertainties in cosmological analyses with SNe Ia is the accuracy of the spectro-photometric model for SNe Ia time series data, which depends on the photometric calibration of the surveys. To quantify the impact of this uncertainty, we analyze simulated Rubin-LSST and HLTDS data, perturb the photometric zero-points and filter mean wavelengths, and propagate these systematics to spectral model recovery, estimated distances, and dark energy figure of merit (FoM) based on the $w_0 w_a$CDM model. Zero-point shifts of 5 mmag and filter mean wavelength shifts of 5 angstrom lead to a $\sim 50\%$ decrease in the FoM relative to a statistical-only case when calibration uncertainties are propagated only through light-curve fitting. The same calibration shifts applied only during model training produce a smaller $\sim 13\%$ degradation. Contrary to previous analyses, calibration uncertainties in light-curve fitting dominate over those from model training. Their effect during light-curve fitting varies smoothly with redshift and is nearly degenerate with cosmology, preventing mitigation through self-calibration. Finally, we show that the FoM dependence on the size of the calibration uncertainties (in the range of expected sizes) is roughly linear.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper uses forward simulations of Rubin-LSST and Roman HLTDS Type Ia supernova light curves to quantify calibration systematics in the SALT3 model. Zero-point shifts of 5 mmag and filter mean-wavelength shifts of 5 Å are applied either only during model training or only during light-curve fitting; the resulting distance biases are propagated to the w0waCDM figure of merit (FoM). The analysis finds a ~50% FoM degradation when shifts affect only fitting, a ~13% degradation when they affect only training, and a near-degeneracy between the fitting-stage bias and cosmological parameters that prevents self-calibration. The FoM dependence on calibration uncertainty size is reported as roughly linear.

Significance. If the simulated propagation faithfully reproduces real-data covariances, the result is significant for Stage-IV survey planning: it indicates that calibration resources should be prioritized for the fitting stage rather than training, quantifies the FoM penalty, and shows why self-calibration is ineffective. The forward-simulation framework itself is a strength, as it allows controlled isolation of training versus fitting contributions.

major comments (2)
  1. [§3] §3 (Simulation and propagation pipeline): the central 50%-versus-13% dominance claim rests on the fidelity of the simulated SN SEDs, training-sample composition, and noise model. No quantitative validation against real calibration residuals or assessment of omitted covariances between zero-point/wavelength errors and color/luminosity parameters is provided; if such covariances exist in the data, the reported relative importance of fitting over training could be an artifact.
  2. [§5] §5 (FoM results and degeneracy): the statement that the fitting-stage bias 'varies smoothly with redshift and is nearly degenerate with cosmology' is load-bearing for the self-calibration conclusion, yet the manuscript does not show the explicit redshift-dependent bias curves or the Fisher-matrix eigenvectors that would demonstrate the degeneracy strength.
minor comments (2)
  1. [Abstract / §4] The abstract and §4 state that the FoM dependence is 'roughly linear' over the explored range, but no figure or table quantifies the slope or reports the goodness-of-fit to linearity.
  2. [Tables] Table captions and axis labels should explicitly state whether the reported FoM values include or exclude the calibration-induced bias term.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and for recognizing the significance of our forward-simulation framework for isolating calibration systematics in SALT3. We address each major comment below. Where the manuscript was incomplete, we have revised it by adding discussion and a new figure; we also clarify the controlled nature of the simulation and note limitations honestly.

read point-by-point responses

  1. Referee: [§3] §3 (Simulation and propagation pipeline): the central 50%-versus-13% dominance claim rests on the fidelity of the simulated SN SEDs, training-sample composition, and noise model. No quantitative validation against real calibration residuals or assessment of omitted covariances between zero-point/wavelength errors and color/luminosity parameters is provided; if such covariances exist in the data, the reported relative importance of fitting over training could be an artifact.

    Authors: We agree that direct quantitative validation against observed calibration residuals from existing surveys would be valuable. Our simulations adopt the SALT3 SED model and noise properties calibrated to the published Rubin-LSST and Roman HLTDS specifications, with training-sample composition drawn from realistic redshift and magnitude distributions used in prior SALT3 analyses. We have added a new paragraph in §3 explicitly discussing the assumptions underlying the SED fidelity and noise model, together with a qualitative assessment of how covariances between zero-point/wavelength shifts and color/luminosity parameters could propagate. Because the forward-modeling approach isolates the training versus fitting stages by construction, any unmodeled covariance would affect both stages; we therefore retain the reported 50 % versus 13 % contrast as a lower bound on the fitting-stage dominance. We acknowledge that a full end-to-end validation against proprietary calibration data lies outside the present scope and have noted this limitation. revision: partial

  2. Referee: [§5] §5 (FoM results and degeneracy): the statement that the fitting-stage bias 'varies smoothly with redshift and is nearly degenerate with cosmology' is load-bearing for the self-calibration conclusion, yet the manuscript does not show the explicit redshift-dependent bias curves or the Fisher-matrix eigenvectors that would demonstrate the degeneracy strength.

    Authors: We accept the referee’s point that explicit visualization strengthens the claim. We have inserted a new figure (Figure 8) in §5 that displays the redshift-dependent distance-modulus bias curves arising from the fitting-stage zero-point and wavelength shifts, together with the leading eigenvectors of the Fisher matrix for the w0–wa plane. These eigenvectors confirm the near-degeneracy between the smooth redshift-dependent bias and the cosmological parameters, directly supporting the conclusion that self-calibration cannot remove the systematic. The revised text now references this figure when stating the degeneracy. revision: yes

Circularity Check

0 steps flagged

No circularity: forward simulation of external perturbations

full rationale

The paper's central results are obtained by generating simulated Rubin-LSST and Roman HLTDS light curves, imposing independent external zero-point (5 mmag) and wavelength (5 Å) shifts, then propagating those shifts separately through SALT3 model training versus light-curve fitting, and finally computing the w0waCDM FoM from the resulting distance estimates. These FoM values are direct numerical outputs of the pipeline applied to the perturbed mocks; they are not fitted parameters, not defined in terms of themselves, and not obtained by renaming or re-using the input perturbations. No load-bearing self-citations, ansatzes, or uniqueness theorems are invoked in the abstract or described chain. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The analysis relies on the established SALT3 model and standard supernova simulation techniques without introducing new free parameters, axioms beyond domain standards, or invented entities.

axioms (2)
  • domain assumption The SALT3 model provides an adequate description of Type Ia supernova spectral energy distributions for cosmological distance estimation.
    The entire training and fitting procedure is performed with SALT3.
  • domain assumption The simulated Rubin-LSST and Roman HLTDS datasets faithfully reproduce the statistical and systematic properties of future observations.
    Perturbations are applied to these simulations to propagate calibration effects.

pith-pipeline@v0.9.0 · 5680 in / 1481 out tokens · 50722 ms · 2026-05-10T01:14:50.100930+00:00 · methodology

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

67 extracted references · 67 canonical work pages · 2 internal anchors

  1. [1]

    Abdul Karimet al.[DESI], Phys

    Abdul Karim, M., Aguilar, J., Ahlen, S., et al. 2025, Physical Review D, 112, doi: 10.1103/tr6y-kpc6

    work page doi:10.1103/tr6y-kpc6 2025

  2. [2]

    G., Aguilar, J., Ahlen, S., et al

    Adame, A., Aguilar, J., Ahlen, S., et al. 2025, Journal of Cosmology and Astroparticle Physics, 2025, 021, doi: 10.1088/1475-7516/2025/02/021

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

  3. [3]

    2021, prd, 103, 083533, doi: 10.1103/PhysRevD.103.083533

    Alam, S., Aubert, M., Avila, S., et al. 2021, Phys. Rev. D, 103, 083533, doi: 10.1103/PhysRevD.103.083533

    work page doi:10.1103/physrevd.103.083533 2021

  4. [4]

    Report of the Dark Energy Task Force

    Albrecht, A., Bernstein, G., Cahn, R., et al. 2006, arXiv e-prints, astro, doi: 10.48550/arXiv.astro-ph/0609591

    work page Pith review doi:10.48550/arxiv.astro-ph/0609591 2006

  5. [5]

    , OPTeprint =

    Astier, P., Guy, J., Regnault, N., et al. 2006, A&A, 447, 31, doi: 10.1051/0004-6361:20054185

    work page doi:10.1051/0004-6361:20054185 2006

  6. [6]

    C., Vincenzi, M., Scolnic, D., et al

    Bailey, A. C., Vincenzi, M., Scolnic, D., et al. 2023, MNRAS, 524, 5432, doi: 10.1093/mnras/stad2179

    work page doi:10.1093/mnras/stad2179 2023

  7. [7]

    2014, , 568, A22, 10.1051/0004-6361/201423413

    Betoule, M., Kessler, R., Guy, J., et al. 2014, A&A, 568, A22, doi: 10.1051/0004-6361/201423413

    work page doi:10.1051/0004-6361/201423413 2014

  8. [8]

    2019 a , , 874, 150, 10.3847/1538-4357/ab08a0

    Brout, D., Scolnic, D., Kessler, R., et al. 2019, ApJ, 874, 150, doi: 10.3847/1538-4357/ab08a0

    work page doi:10.3847/1538-4357/ab08a0 2019

  9. [9]

    and Carr, Anthony and Zuntz, Joe and Kessler, Rick and Davis, Tamara M

    Brout, D., Scolnic, D., Popovic, B., et al. 2022a, The Astrophysical Journal, 938, 110, doi: 10.3847/1538-4357/ac8e04

    work page doi:10.3847/1538-4357/ac8e04

  10. [11]

    , keywords =

    Brout, D., Taylor, G., Scolnic, D., et al. 2022, ApJ, 938, 111, doi: 10.3847/1538-4357/ac8bcc

    work page doi:10.3847/1538-4357/ac8bcc 2022

  11. [12]

    C., Guo, Z., Scolnic, D., et al

    Chen, R. C., Guo, Z., Scolnic, D., et al. 2026, ApJ, 999, 81, doi: 10.3847/1538-4357/ae42be

    work page doi:10.3847/1538-4357/ae42be 2026

  12. [13]

    LSST Science Book, Version 2.0

    Collaboration, L. S., Abell, P. A., Allison, J., et al. 2009, LSST Science Book, Version 2.0. https://arxiv.org/abs/0912.0201

    work page Pith review arXiv 2009

  13. [14]

    2010, The Astrophysical Journal Supplement Series, 192, 1, 10.1088/0067-0049/192/1/1

    Conley, A., Guy, J., Sullivan, M., et al. 2011, ApJS, 192, 1, doi: 10.1088/0067-0049/192/1/1

    work page doi:10.1088/0067-0049/192/1/1 2011

  14. [15]

    O., Kenworthy, W

    Dai, M., Jones, D. O., Kenworthy, W. D., et al. 2023, ApJS, 267, 1, doi: 10.3847/1538-4365/acd051

    work page doi:10.3847/1538-4365/acd051 2023

  15. [16]

    2006, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol

    Pierfederici, F. 2006, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 6270, Observatory Operations: Strategies, Processes, and Systems, ed. D. R. Silva & R. E. Doxsey, 62701D, doi: 10.1117/12.671992

    work page doi:10.1117/12.671992 2006

  16. [17]

    A., et al

    Dilday, B., Kessler, R., Frieman, J. A., et al. 2008, ApJ, 682, 262, doi: 10.1086/587733

    work page doi:10.1086/587733 2008

  17. [18]

    , keywords =

    Guy, J., Astier, P., Nobili, S., Regnault, N., & Pain, R. 2005, A&A, 443, 781, doi: 10.1051/0004-6361:20053025

    work page doi:10.1051/0004-6361:20053025 2005

  18. [19]

    2007, , 466, 11, 10.1051/0004-6361:20066930

    Guy, J., Astier, P., Baumont, S., et al. 2007, A&A, 466, 11, doi: 10.1051/0004-6361:20066930

    work page doi:10.1051/0004-6361:20066930 2007

  19. [20]

    , keywords =

    Hamuy, M., Phillips, M. M., Suntzeff, N. B., et al. 1996, The Astronomical Journal, 112, 2391, doi: 10.1086/118190

    work page doi:10.1086/118190 1996

  20. [21]

    2020, The Journal of Open Source Software, 5, 2122, 10.21105/joss.02122

    Hinton, S., & Brout, D. 2020, The Journal of Open Source Software, 5, 2122, doi: 10.21105/joss.02122

    work page doi:10.21105/joss.02122 2020

  21. [22]

    and Campbell, Heather and Dilday, Ben and et al

    Hlozek, R., Kunz, M., Bassett, B., et al. 2012, The Astrophysical Journal, 752, 79, doi: 10.1088/0004-637X/752/2/79

    work page doi:10.1088/0004-637x/752/2/79 2012

  22. [23]

    arXiv e-prints , keywords =

    HLTDS, H. 2025, arXiv e-prints, arXiv:2505.10574, doi: 10.48550/arXiv.2505.10574

    work page doi:10.48550/arxiv.2505.10574 2025

  23. [24]

    J., et al

    Hounsell, R., Scolnic, D., Foley, R. J., et al. 2018, ApJ, 867, 23, doi: 10.3847/1538-4357/aac08b

    work page doi:10.3847/1538-4357/aac08b 2018

  24. [25]

    Ivezi ´c, S

    Ivezic, Z., Kahn, S. M., Tyson, J. A., et al. 2019, The Astrophysical Journal, 873, 111, doi: 10.3847/1538-4357/ab042c

    work page doi:10.3847/1538-4357/ab042c 2019

  25. [26]

    O., Kenworthy, W

    Jones, D. O., Kenworthy, W. D., Dai, M., et al. 2023, ApJ, 951, 22, doi: 10.3847/1538-4357/acd195

    work page doi:10.3847/1538-4357/acd195 2023

  26. [27]

    O., Scolnic , D

    Jones, D. O., Scolnic, D. M., Riess, A. G., et al. 2018, ApJ, 857, 51, doi: 10.3847/1538-4357/aab6b1

    work page doi:10.3847/1538-4357/aab6b1 2018

  27. [28]

    O., Scolnic , D

    Jones, D. O., Scolnic, D. M., Foley, R. J., et al. 2019, ApJ, 881, 19, doi: 10.3847/1538-4357/ab2bec

    work page doi:10.3847/1538-4357/ab2bec 2019

  28. [29]

    Kirshner, R. P. 2010, ApJ, 715, 743, doi: 10.1088/0004-637X/715/2/743

    work page doi:10.1088/0004-637x/715/2/743 2010

  29. [30]

    , keywords =

    Kenworthy, W. D., Jones, D. O., Dai, M., et al. 2021, ApJ, 923, 265, doi: 10.3847/1538-4357/ac30d8

    work page doi:10.3847/1538-4357/ac30d8 2021

  30. [31]

    D., Goobar, A., Jones, D

    Kenworthy, W. D., Goobar, A., Jones, D. O., et al. 2025, A&A, 697, A125, doi: 10.1051/0004-6361/202452578

    work page doi:10.1051/0004-6361/202452578 2025

  31. [32]

    2025, arXiv e-prints, arXiv:2506.04402

    Kessler, R., Hounsell, R., Joshi, B., et al. 2025, arXiv e-prints, arXiv:2506.04402. https://arxiv.org/abs/2506.04402

    work page arXiv 2025

  32. [33]

    , keywords =

    Kessler, R., & Scolnic, D. 2017, ApJ, 836, 56, doi: 10.3847/1538-4357/836/1/56

    work page doi:10.3847/1538-4357/836/1/56 2017

  33. [34]

    2023, ApJL, 952, L8, doi: 10.3847/2041-8213/ace34d

    Kessler, R., Vincenzi, M., & Armstrong, P. 2023, ApJL, 952, L8, doi: 10.3847/2041-8213/ace34d

    work page doi:10.3847/2041-8213/ace34d 2023

  34. [35]

    P., Cinabro, D., et al

    Kessler, R., Becker, A. C., Cinabro, D., et al. 2009a, The Astrophysical Journal Supplement Series, 185, 32, doi: 10.1088/0067-0049/185/1/32

    work page doi:10.1088/0067-0049/185/1/32

  35. [36]

    P., Cinabro , D., et al

    Kessler, R., Bernstein, J. P., Cinabro, D., et al. 2009b, Publications of the Astronomical Society of the Pacific, 121, 1028, doi: 10.1086/605984

    work page doi:10.1086/605984

  36. [37]

    , keywords =

    Kessler, R., Narayan, G., Avelino, A., et al. 2019a, PASP, 131, 094501, doi: 10.1088/1538-3873/ab26f1 —. 2019b, PASP, 131, 094501, doi: 10.1088/1538-3873/ab26f1

    work page doi:10.1088/1538-3873/ab26f1

  37. [38]

    A., & Hlozek, R

    Kunz, M., Bassett, B. A., & Hlozek, R. A. 2007, PhRvD, 75, 103508, doi: 10.1103/PhysRevD.75.103508

    work page doi:10.1103/physrevd.75.103508 2007

  38. [39]

    C., et al

    Lampeitl, H., Smith, M., Nichol, R. C., et al. 2010, ApJ, 722, 566, doi: 10.1088/0004-637X/722/1/566

    work page doi:10.1088/0004-637x/722/1/566 2010

  39. [40]

    2022, ApJS, 259, 58, doi: 10.3847/1538-4365/ac5033

    Lochner, M., Scolnic, D., Almoubayyed, H., et al. 2022, ApJS, 259, 58, doi: 10.3847/1538-4365/ac5033 17

    work page doi:10.3847/1538-4365/ac5033 2022

  40. [41]

    , keywords =

    Marriner, J., Bernstein, J. P., Kessler, R., et al. 2011, ApJ, 740, 72, doi: 10.1088/0004-637X/740/2/72

    work page doi:10.1088/0004-637x/740/2/72 2011

  41. [42]

    doi:10.48550/arXiv.2512.06319 , archivePrefix =

    Mitra, A., Kessler, R., Chen, R. C., et al. 2025, arXiv e-prints, arXiv:2512.06319. https://arxiv.org/abs/2512.06319

    work page arXiv 2025

  42. [43]

    2024, arXiv e-prints, arXiv:2501.05632, doi: 10.48550/arXiv.2501.05632

    Collaboration, The Roman HLIS Project Infrastructure Team, et al. 2024, arXiv e-prints, arXiv:2501.05632, doi: 10.48550/arXiv.2501.05632

    work page doi:10.48550/arxiv.2501.05632 2024

  43. [44]

    Measurements of Omega and Lambda from 42 High-Redshift Supernovae

    Perlmutter, S., Aldering, G., Goldhaber, G., et al. 1999, ApJ, 517, 565, doi: 10.1086/307221

    work page internal anchor Pith review doi:10.1086/307221 1999

  44. [45]

    Phillips, M. M. 1993, ApJL, 413, L105, doi: 10.1086/186970

    work page doi:10.1086/186970 1993

  45. [46]

    Pierel, J. D. R., Jones, D. O., Dai, M., et al. 2021, ApJ, 911, 96, doi: 10.3847/1538-4357/abe867

    work page doi:10.3847/1538-4357/abe867 2021

  46. [47]

    Pierel, J. D. R., Jones, D. O., Kenworthy, W. D., et al. 2022, ApJ, 939, 11, doi: 10.3847/1538-4357/ac93f9 Planck Collaboration, Aghanim, N., Akrami, Y., et al. 2020, A&A, 641, A1, doi: 10.1051/0004-6361/201833880

    work page doi:10.3847/1538-4357/ac93f9 2022

  47. [48]

    , keywords =

    Popovic, B., Brout, D., Kessler, R., & Scolnic, D. 2023, ApJ, 945, 84, doi: 10.3847/1538-4357/aca273

    work page doi:10.3847/1538-4357/aca273 2023

  48. [49]

    Popovic et al.,A Reassessment of the Pantheon+ and DES 5YR Calibration Uncertainties: Dovekie,2506.05471

    Popovic, B., Kenworthy, W. D., Ginolin, M., et al. 2025, arXiv e-prints, arXiv:2506.05471, doi: 10.48550/arXiv.2506.05471

    work page doi:10.48550/arxiv.2506.05471 2025

  49. [51]

    2021, The Astronomical Journal, 162, 67, doi: 10.3847/1538-3881/ac0824

    Qu, H., Sako, M., M¨ oller, A., & Doux, C. 2021, The Astronomical Journal, 162, 67, doi: 10.3847/1538-3881/ac0824

    work page doi:10.3847/1538-3881/ac0824 2021

  50. [52]

    Ridgway, S. T. 2016, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 9911, Modeling, Systems Engineering, and Project Management for Astronomy VI, ed. G. Z. Angeli & P. Dierickx, 991125, doi: 10.1117/12.2232680

    work page doi:10.1117/12.2232680 2016

  51. [53]

    Observational Evidence from Supernovae for an Accelerating Universe and a Cosmological Constant

    Riess, A. G., Filippenko, A. V., Challis, P., et al. 1998, AJ, 116, 1009, doi: 10.1086/300499 Roman Core Community Survey Definition Committees. 2025, Interim Report: The Community Definition of Roman’s Core Community Surveys, Tech. rep., NASA Goddard Space Flight Center. https: //asd.gsfc.nasa.gov/roman/comm forum/forum 17/ Core Community Survey Reports-...

    work page Pith review doi:10.1086/300499 1998

  52. [54]

    M., Baltay, C., Hounsell, R., et al

    Rose, B. M., Baltay, C., Hounsell, R., et al. 2021, arXiv e-prints, arXiv:2111.03081, doi: 10.48550/arXiv.2111.03081

    work page doi:10.48550/arxiv.2111.03081 2021

  53. [55]

    M., Vincenzi, M., Hounsell, R., et al

    Rose, B. M., Vincenzi, M., Hounsell, R., et al. 2025, arXiv e-prints, arXiv:2506.05161, doi: 10.48550/arXiv.2506.05161

    work page doi:10.48550/arxiv.2506.05161 2025

  54. [56]

    Rubinet al., Union Through UNITY: Cosmology with 2,000 SNe Using a Unified Bayesian Framework, Astrophys

    Rubin, D., Aldering, G., Betoule, M., et al. 2025, The Astrophysical Journal, 986, 231, doi: 10.3847/1538-4357/adc0a5

    work page doi:10.3847/1538-4357/adc0a5 2025

  55. [57]

    arXiv e-prints , keywords =

    Rubin, D., Aldering, G., Fruchter, A., et al. 2025, arXiv e-prints, arXiv:2506.04327, doi: 10.48550/arXiv.2506.04327 S´ anchez, B. O., Kessler, R., Scolnic, D., et al. 2022, ApJ, 934, 96, doi: 10.3847/1538-4357/ac7a37

    work page doi:10.48550/arxiv.2506.04327 2025

  56. [58]

    P., Suntzeff, N

    Schmidt, B. P., Suntzeff, N. B., Phillips, M. M., et al. 1998, ApJ, 507, 46, doi: 10.1086/306308

    work page doi:10.1086/306308 1998

  57. [59]

    , archivePrefix = "arXiv", eprint =

    Scolnic, D., Rest, A., Riess, A., et al. 2014, ApJ, 795, 45, doi: 10.1088/0004-637X/795/1/45

    work page doi:10.1088/0004-637x/795/1/45 2014

  58. [60]

    M., Jones , D

    Scolnic, D. M., Jones, D. O., Rest, A., et al. 2018, ApJ, 859, 101, doi: 10.3847/1538-4357/aab9bb

    work page doi:10.3847/1538-4357/aab9bb 2018

  59. [61]

    Wide-Field InfrarRed Survey Telescope-Astrophysics Focused Telescope Assets WFIRST-AFTA 2015 Report

    Spergel, D., Gehrels, N., Baltay, C., et al. 2015, arXiv e-prints, arXiv:1503.03757, doi: 10.48550/arXiv.1503.03757

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1503.03757 2015

  60. [62]

    2010, Monthly Notices of the Royal Astronomical Society, 408, 1181, doi: 10.1111/j.1365-2966.2010.17197.x

    Sullivan, M., Conley, A., Howell, D. A., et al. 2010, MNRAS, 406, 782, doi: 10.1111/j.1365-2966.2010.16731.x

    work page doi:10.1111/j.1365-2966.2010.16731.x 2010

  61. [63]

    2012, , 746, 85, 10.1088/0004-637X/746/1/85

    Suzuki, N., Rubin, D., Lidman, C., et al. 2012, The Astrophysical Journal, 746, 85, doi: 10.1088/0004-637X/746/1/85

    work page doi:10.1088/0004-637x/746/1/85 2012

  62. [64]

    Taylor, G., Lidman, C., Popovic, B., & Abbot, H. J. 2024, MNRAS, 528, 4643, doi: 10.1093/mnras/stae293

    work page doi:10.1093/mnras/stae293 2024

  63. [65]

    Taylor, D

    Taylor, G., Jones, D. O., Popovic, B., et al. 2023, MNRAS, 520, 5209, doi: 10.1093/mnras/stad320 The LSST Dark Energy Science Collaboration,

    work page doi:10.1093/mnras/stad320 2023

  64. [66]

    The lsst dark energy science collaboration (desc) science requirements document,

    Mandelbaum, R., Eifler, T., et al. 2018, arXiv e-prints, arXiv:1809.01669, doi: 10.48550/arXiv.1809.01669

    work page doi:10.48550/arxiv.1809.01669 2018

  65. [67]

    , keywords =

    Vincenzi, M., Brout, D., Armstrong, P., et al. 2024a, ApJ, 975, 86, doi: 10.3847/1538-4357/ad5e6c —. 2024b, ApJ, 975, 86, doi: 10.3847/1538-4357/ad5e6c

    work page doi:10.3847/1538-4357/ad5e6c

  66. [68]

    O., Pierel, J

    Wang, Q., Jones, D. O., Pierel, J. D. R., et al. 2025, arXiv e-prints, arXiv:2512.25064, doi: 10.48550/arXiv.2512.25064

    work page doi:10.48550/arxiv.2512.25064 2025

  67. [69]

    Wang, Y. 2008, PhRvD, 77, 123525, doi: 10.1103/PhysRevD.77.123525 18 APPENDIX A.VALIDATION OF TRAINED MODELS We compare both models’ performance at recovering SN parameters for some simulatedRomanlight curves. Both models show broadly similar distributions for most of the key SN parameters and their uncertainties as well as redshift and redshift errors. T...

    work page doi:10.1103/physrevd.77.123525 2008