Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

A neutron star's radius can be pinned to tens of meters from its mass and two oscillation frequencies alone, with no equation of state needed; simulation-based inference supplies the honest error bars.

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 12:08 UTC pith:QEJDOVS4

load-bearing objection Useful new method and a new neutron-star radius relation, but the error-bar calibration is tuned on the test set, so the reliability claims need independent validation. the 3 major comments →

arxiv 2601.03945 v2 pith:QEJDOVS4 submitted 2026-01-07 gr-qc astro-ph.HE

Combining simulation-based inference and universal relations for precise and accurate neutron star science

Christian J. Kr\"uger , Sebastian H. V\"olkel This is my paper
classification gr-qc astro-ph.HE
keywords neutron starsuniversal relationssimulation-based inferenceneural posterior estimationequation of stateasteroseismologyf-moderadius measurement
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 argues that the unknown nuclear equation of state (EOS) inside a neutron star can be treated as intrinsic noise rather than a missing ingredient, and that likelihood-free simulation-based inference (SBI) can exploit this view to predict bulk properties directly from simulated data. As a demonstration it reports a new explicit universal relation, R = a0 + a1 sqrt(M) + a2/f + a3 (M f)^2 + a4 (f/p1), that yields the radius from mass, fundamental mode frequency f, and first pressure-mode frequency p1 with 'well beyond percent' accuracy—deviations mostly under 80 meters. The same data, fed to a neural posterior estimator, outperform the explicit formula: radius estimates of a few tens of meters with systematic errors that pass coverage checks. The paper also shows that the standard error bars of universal relations are unreliable, and introduces a calibration step that turns them into honest systematic uncertainties. If correct, this gives future asteroseismic and gravitational-wave observations a direct, EOS-agnostic route from oscillation frequencies to radius—and thus to dense-matter constraints.

Core claim

The central claim is that a non-rotating neutron star's radius is fixed, to well beyond percent accuracy, by its mass and two oscillation frequencies—the fundamental mode f and the first pressure mode p1—through an explicit five-parameter formula (Eq. 1). A neural posterior estimator trained on 1,491 equations of state treats the spread of EOS predictions as intrinsic noise; given (M, f, p1) it returns a radius posterior whose mean misses the true radius by typically less than 80 m, often by only a few meters, with 68% widths as small as ~18 m. The same search recovers known relations among moment of inertia, tidal deformability, and the f-mode frequency. The paper also claims that the least

What carries the argument

The load-bearing object is the map (M, f, p1) -> R, realized in two forms. The explicit universal relation is R = a0 + a1 sqrt(M) + a2/f + a3 (M f)^2 + a4 (f/p1), with best-fit values a0 = -3.312, a1 = 4.864, a2 = 4.360e-2, a3 = -2.828e3, a4 = 3.973; it provides a hand-evaluable, EOS-insensitive approximation. The other form is a neural posterior estimator (simulation-based inference via normalizing flows) that, instead of committing to a functional form, learns the conditional distribution of R given (M, f, p1) from the same 1,491 simulated EOS realizations. What carries the argument is the treatment of the EOS variation as 'EOS noise': the width of the SBI posterior is literally an estimat

Load-bearing premise

The 1,491 equations of state, drawn from four parametrizations and filtered by mass, radius, and tidal-deformability constraints, are assumed to be a broad and unbiased sample of the true nuclear EOS space; if this sample is skewed, both the universal relation and the SBI error bars inherit that skew.

What would settle it

Hold out all equations of state from one of the four parametrization families, retrain both the SBI estimator and the universal relation on the remaining three, and test on the held-out family; if the median radius deviation exceeds ~100 m or the 68% HDI coverage drops markedly below 0.68, the claimed EOS insensitivity is an artifact of the training prior.

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

If this is right

  • Future asteroseismic or gravitational-wave observations that return M, f, and p1 for one star can produce a radius posterior with tens-of-meters error bars, with no nuclear EOS assumed in advance.
  • The explicit formula can be evaluated by hand or in simple code, making it a practical cross-check for the neural network and a fallback when the network is unavailable.
  • The automated SBI screening ranks which subsets of bulk quantities carry information, so analysts can decide where to invest effort before constructing analytic relations.
  • The calibration procedure gives existing and future universal relations honest error bars, preventing downstream EOS inferences from being systematically overconfident.
  • Because the calibration is generic, it can be applied to other universal relations beyond radius, including those for tidal deformability or moment of inertia.

Where Pith is reading between the lines

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

  • Editorial extension: the same SBI treatment should transfer to other EOS-sensitive quantities such as tidal deformability or moment of inertia, where the training prior is already shown to cover current observational constraints; the paper explicitly leaves these as future work.
  • Editorial extension: the success of a single calibration scale hints that the residual noise of the radius relation is close to Gaussian and roughly stationary across the mass–frequency plane; if that holds more broadly, simple analytic error formulas could replace full simulation-based inference in many applications.
  • Editorial extension: because the uncalibrated covariance and the SBI posterior can disagree strongly in the tails, a practical observer could treat the two as a systematics check—when they disagree, the training EOS sample or the functional ansatz may be missing structure.
  • Editorial extension: the reported few-tens-of-meters accuracy assumes noiseless inputs for M, f, and p1; real detector noise will widen the posteriors, so the headline accuracy is an upper bound on asteroseismic precision, not a guaranteed observational result.

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 / 5 minor

Summary. The paper proposes a workflow that combines simulation-based inference (SBI, implemented via neural posterior estimation) with conventional universal-relation construction for neutron star bulk properties. Using a set of 1491 equation-of-state (EOS) realizations from four parametrizations, the authors train an SBI network on 602 possible divisions of six bulk quantities. The method ranks parameter combinations and identifies a new explicit universal relation R = a0 + a1 sqrt(M) + a2/f + a3 (M f)^2 + a4 f/p1 (Eq. 1). The paper claims that this relation predicts the radius to sub-percent accuracy, that SBI can outperform it and provide radius estimates with reliable systematic error bars, and that a calibration procedure can make the universal relation's covariance matrix accurately reflect systematic EOS noise. These claims are supported by held-out test data: deviation histograms, HDI width distributions, and coverage plots.

Significance. If the claims are correct, the paper offers a practical way to infer neutron star radii from mass and two oscillation frequencies without knowing the high-density EOS, with O(10-100 m) accuracy and honest uncertainty estimates. The explicit universal relation is simple and potentially useful for gravitational-wave asteroseismology. The paper also demonstrates a useful automated search strategy for discovering universal relations. However, the load-bearing uncertainty-quantification claims rest on a validation protocol that uses the test set for model selection and calibration, which is a serious methodological gap. The EOS-sample breadth is also inherited from a prior study and could limit the 'universal' claim.

major comments (3)
  1. [Sec. IIIC, p. 4-5; Appendix B; Appendix C] The coverage analysis that underpins the central claim of 'reliable systematic errors' is not independent. In Sec. IIIC the authors state that they 'repeated the training and the tests described in Sec. B several times until we obtained a well-calibrated network.' This selects a network by repeated evaluation on the test set, so the reported coverage and the KS p-value in Appendix B are optimistic. Likewise, Appendix C calibrates the universal relation by varying a single effective error 'until the value that gives the closest match to the diagonal' on the test data; Fig. 4 is then computed on the same test set. This double use of the test set invalidates the coverage evidence and the quantitative measure of systematic uncertainty. A proper validation would use a separate calibration/validation set or a nested cross-validation, and would report the number of retraining attempts and the s
  2. [Sec. IIA, Ref. [38]] The EOS set consists of 1491 realizations from four parametrizations taken from a prior study [38]. Universality is claimed relative to this sample, but all training, validation, and test points are drawn from the same four families. If the prior is skewed or incomplete, the EOS-insensitivity of Eq. (1) and the SBI posterior widths inherit that skew. The paper would be substantially stronger with a leave-one-family-out test (train on three parametrizations, test on the fourth) or a comparison against an independent EOS ensemble (e.g., chiral effective field theory or different agnostic parametrizations). Without such a test, the 'universal' and 'EOS-noise' claims are conditioned on the assumed prior and may not generalize to the true EOS distribution.
  3. [Figs. 2, 3 and Secs. IIIA-IIIB] The claim that 'SBI can outperform the universal relation' is not quantitatively established. The comparison is presented only as histograms of deviations and of 68% HDI widths. No aggregate statistics (e.g., mean absolute deviation, root-mean-square error, fraction within a given tolerance) or their uncertainties are reported, and the histograms appear broadly overlapping. The abstract's statement that SBI 'provides radius estimates of only a few tens of meters' is not supported by Fig. 3, which shows widths up to 150 m and only a minority below a few tens of meters. Please report numeric summary statistics for both methods (and for the calibrated UR) with proper uncertainties, and clarify how the 'few tens of meters' claim is derived from the data.
minor comments (5)
  1. [Sec. IIC] The phrase 'well beyond percent accuracy' is ambiguous; if it means sub-percent accuracy, please state so explicitly and give a numerical value (e.g., median relative error).
  2. [Eq. (1)] The best-fit coefficients are given without uncertainties or covariance matrix. Since the paper discusses the covariance matrix, please include the numeric covariance matrix (or at least standard errors) in a table.
  3. [Figure 1] The single example shown is not identified as typical or worst-case. Please state how it was selected and how representative it is of the test set.
  4. [Appendix A] Equation (A1) is a normal approximation to the binomial confidence interval. For finite N_test = 1490 it is adequate, but the Wilson interval would be safer for p near 0 or 1. Please mention this choice.
  5. [Appendix C] The residual-based calibration (Fig. 7) is a useful sanity check, but the text says it is 'slightly worse' for central HDI probabilities. Please quantify this deviation (e.g., maximum coverage deviation) so the reader can judge the accuracy.

Circularity Check

2 steps flagged

Error-bar/coverage claims are partly circular: the UR calibration parameter is tuned to the test-set coverage curve and the SBI network is retrained until well calibrated on the same test data; the point prediction of Eq. (1) is not circular.

specific steps
  1. fitted input called prediction [Appendix C (Calibration of universal relations), controlling Fig. 4]
    "By varying the effective error (one single calibration parameter for all data points) and repeating the universal relation fit, we obtain different covariance matrices and different calibration curves. Finally, we use the value that gives the closest match to the diagonal."

    The one free calibration parameter is chosen by matching the coverage curve to the diagonal on the test set, and the same test set is then used to report the calibrated universal relation's coverage in Fig. 4 as evidence that its error bars are accurate. The near-diagonal coverage is therefore enforced by construction rather than independently predicted. This affects the claim that the calibrated universal relation provides reliable systematic errors, not the point accuracy of Eq. (1).

  2. other [Sec. IIIC (Accuracy of systematic error prediction) and Appendix B]
    "In our case, we repeated the training and the tests described in Sec. B several times until we obtained a well-calibrated network."

    The SBI network is selected by repeatedly training and testing until the test-set calibration looks good; the same test set is then used for the coverage curve in Fig. 4 and the KS p-value in Appendix B. Retraining on the basis of test-set calibration means the reported coverage is a fitted outcome, not an out-of-sample validation. Consequently the paper's central claim that the SBI systematic error estimates are reliable is not independently established, although the point predictions remain informative.

full rationale

The explicit universal relation R(M,f,p1), Eq. (1), is not circular: it is a five-parameter empirical fit to a training subset and its point accuracy is evaluated on held-out test data (Figs. 1-3). The SBI point predictions are likewise trained and evaluated on separate splits. No step reduces Eq. (1) to its target by definition, and no load-bearing uniqueness theorem is imported from the authors' earlier work. The circularity is confined to the uncertainty-quantification/calibration claims. Appendix C tunes the single effective EOS-noise parameter so that the universal relation's coverage curve on the test data matches the diagonal, then presents that same test-set coverage in Fig. 4 as evidence of accurate systematic errors. The residual-based calibration in Appendix C likewise estimates the Gaussian width from test residuals and then validates on the same test points. For SBI, the paper states that training/tests were repeated until a well-calibrated network was obtained, again selecting on the test data before reporting coverage and a KS p-value. These steps make the 'reliable systematic errors' assertion partly circular: the agreement with the diagonal is, to an unquantified degree, a fitted result rather than a prediction. The prior EOS sample is inherited from the authors' Ref. [38], but that is a data-generation choice, not a circular derivation of the target relation. Overall, the point-prediction content is independent; the error-bar/coverage validation is partially circular, warranting a score of 6 rather than 8 or 10.

Axiom & Free-Parameter Ledger

3 free parameters · 6 axioms · 0 invented entities

The central claims rest almost entirely on the simulated EOS sample and fitted forms. There are no first-principles derivations; the functional form of Eq. (1) is an empirical ansatz suggested by SBI, and the calibrated error bars depend on a scalar parameter tuned to the validation data. No new physical entities are introduced.

free parameters (3)
  • Universal relation coefficients a0..a4 = a0=-3.312, a1=4.864, a2=4.360e-2, a3=-2.828e3, a4=3.973
    Fitted by least squares to the training data; define Eq. (1), the central explicit relation.
  • Effective EOS-noise calibration parameter
    Appendix C: a single scaling parameter varied until the universal relation's coverage curve matches the diagonal on the test data. This makes the calibrated covariance matrix a fitted quantity, not an independent prediction.
  • Neural network weights (27,510 parameters)
    Reported in Sec. IIIB as the number of hidden-layer parameters; fit to the 72% training split. These are standard ML fitted parameters, not hand-chosen, but they are fit to data.
axioms (6)
  • domain assumption TOV equations, Hartle slow-rotation equations, Love-number equations, and the standard eigenvalue formulation for modes yield accurate bulk quantities.
    Used in Sec. IIA to generate all data; if any of these solvers is inaccurate, the universal relations and SBI posteriors inherit the error.
  • domain assumption The four EOS parametrizations [33-36] from Ref. [38] constitute a sufficiently broad and unbiased prior over viable neutron-star matter.
    All claims of EOS-insensitivity and EOS-noise calibration rest on this sample; the paper provides no independent check of sample coverage.
  • domain assumption The astrophysical selection cuts (causality, M_TOV>=1.97 Msun, R1.6>10.6 km, 11.5<=R1.4<=13.5 km, 120<=Lambda1.4<=800) define the viable parameter space.
    Used in Sec. IIA to filter EOS realizations; too narrow or too wide cuts would change both the universal relation and the noise estimates.
  • ad hoc to paper Residual EOS scatter in R at fixed (M,f,p1) can be approximated by a single Gaussian effective error ('EOS noise').
    Appendix C: the effective error is a single scaling parameter chosen by matching the coverage curve to the diagonal; no first-principles justification is given, and the authors note there is no guarantee it works in more complicated cases.
  • domain assumption NPE with normalizing flows accurately approximates the true posterior p(R|M,f,p1).
    Validated with sbi rank tests in Appendix B, but the network was selected by repeated retraining until well-calibrated, weakening the out-of-sample character of the validation.
  • domain assumption The p1-mode is a well-defined observable and is computed to 1e-5 accuracy for all sampled EOS.
    The novelty of the proposed relation depends on p1 carrying independent information; no external validation of the p1 solver is provided.

pith-pipeline@v1.3.0-alltime-deepseek · 13299 in / 13371 out tokens · 129932 ms · 2026-08-03T12:08:50.452863+00:00 · methodology

0 comments
read the original abstract

In this work, we propose a novel approach for identifying, constructing, and validating precise and accurate universal relations for neutron star bulk quantities. A central element is simulation-based inference (SBI), which we adopt to treat uncertainties due to the unknown nuclear equation of state (EOS) as intrinsic non-trivial noise. By assembling a large set of bulk properties of non-rotating neutron stars across multiple state-of-the-art EOS models, we are able to systematically explore universal relations in high-dimensional parameter spaces. Our framework further identifies the most promising parameter combinations, enabling a more focused and traditional construction of explicit universal relations. At the same time, SBI does not rely on explicit relations; instead, it directly provides predictive distributions together with a quantitative measure of systematic uncertainties, which are not captured by conventional approaches. As an example, we report a new universal relation that allows us to obtain the radius as a function of mass, fundamental mode, and one pressure mode. Our analysis shows that SBI can surpass the predictive power of this universal relation while also mitigating systematic errors. Finally, we demonstrate how universal relations can be further calibrated to mitigate systematic errors accurately.

Figures

Figures reproduced from arXiv: 2601.03945 by Christian J. Kr\"uger, Sebastian H. V\"olkel.

Figure 2
Figure 2. Figure 2: FIG. 2. Histograms of the deviations of the mean values of [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Calibration plot visualizing the reliability of the pos [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Empirical cumulative density function of the poste [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Distribution of the ranks. They grey area denotes [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Calibration plot for the universal relation where we [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] 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 1 Pith paper

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

  1. Spectroscopy of analogue black holes using simulation-based inference

    gr-qc 2026-04 unverdicted novelty 5.0

    Simulation-based inference reliably extracts physical parameters from noisy spectra of analogue black holes.

Reference graph

Works this paper leans on

63 extracted references · 47 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Hewish, S.J

    A. Hewish, S.J. Bell, J.D.H. Pilkington, P. F.Scott,and R. A. Collins, Observation of a rapidly pulsating radio source, Nature217, 709 (1968)

    1968

  2. [2]

    Gold, Rotating neutron stars as the origin of the pul- sating radio sources, Nature218, 731 (1968)

    T. Gold, Rotating neutron stars as the origin of the pul- sating radio sources, Nature218, 731 (1968)

    1968

  3. [3]

    J. M. Lattimer and M. Prakash, The physics of neutron stars, Science304, 536 (2004), arXiv:astro-ph/0405262

    Pith/arXiv arXiv 2004

  4. [4]

    J. M. Lattimer and D. N. Schramm, Black-hole-neutron- star collisions, Astrophys. J. Lett.192, L145 (1974)

    1974

  5. [5]

    B. P. Abbottet al.(LIGO Scientific, Virgo, Fermi GBM, INTEGRAL, IceCube, AstroSat Cadmium Zinc Telluride Imager Team, IPN, Insight-Hxmt, ANTARES, Swift, AGILE Team, 1M2H Team, Dark Energy Camera GW-EM, DES, DLT40, GRAWITA, Fermi-LAT, ATCA, ASKAP, Las Cumbres Observatory Group, OzGrav, DWF (Deeper Wider Faster Program), AST3, CAAS- TRO, VINROUGE, MASTER, J...

    Pith/arXiv arXiv 2017

  6. [6]

    M. R. Droutet al., Light Curves of the Neutron Star Merger GW170817/SSS17a: Implications for R- Process Nucleosynthesis, Science358, 1570 (2017), arXiv:1710.05443 [astro-ph.HE]

    Pith/arXiv arXiv 2017

  7. [7]

    R. A. Hulse and J. H. Taylor, Discovery of a pulsar in a binary system, Astrophys. J. Lett.195, L51 (1975)

    1975

  8. [8]

    P. C. C. Freire and N. Wex, Gravity experiments with radio pulsars, Living Rev. Rel.27, 5 (2024), arXiv:2407.16540 [gr-qc]

    Pith/arXiv arXiv 2024

  9. [9]

    B. P. Abbottet al.(LIGO Scientific, Virgo), GW170817: Observation of Gravitational Waves from a Binary Neu- tron Star Inspiral, Phys. Rev. Lett.119, 161101 (2017), arXiv:1710.05832 [gr-qc]

    Pith/arXiv arXiv 2017

  10. [10]

    B. P. Abbottet al.(LIGO Scientific, Virgo), GW190425: Observation of a Compact Binary Coalescence with To- tal Mass„3.4M d, Astrophys. J. Lett.892, L3 (2020), arXiv:2001.01761 [astro-ph.HE]

    Pith/arXiv arXiv 2020

  11. [11]

    Abbottet al.(LIGO Scientific, KAGRA, VIRGO), Observation of Gravitational Waves from Two Neutron Star–Black Hole Coalescences, Astrophys

    R. Abbottet al.(LIGO Scientific, KAGRA, VIRGO), Observation of Gravitational Waves from Two Neutron Star–Black Hole Coalescences, Astrophys. J. Lett.915, L5 (2021), arXiv:2106.15163 [astro-ph.HE]

    Pith/arXiv arXiv 2021

  12. [12]

    A. G. Abacet al.(LIGO Scientific, VIRGO, KAGRA), GWTC-4.0: Updating the Gravitational-Wave Transient Catalog with Observations from the First Part of the Fourth LIGO-Virgo-KAGRA Observing Run, arXiv e- prints (2025), arXiv:2508.18082 [gr-qc]

    Pith/arXiv arXiv 2025

  13. [13]

    Andersson and K

    N. Andersson and K. D. Kokkotas, Towards gravitational wave asteroseismology, Mon. Not. Roy. Astron. Soc.299, 1059 (1998), arXiv:gr-qc/9711088

    Pith/arXiv arXiv 1998

  14. [14]

    Benhar, E

    O. Benhar, E. Berti, and V. Ferrari, The Imprint of the equation of state on the axial w modes of oscillating neu- tron stars, Mon. Not. Roy. Astron. Soc.310, 797 (1999), arXiv:gr-qc/9901037

    Pith/arXiv arXiv 1999

  15. [15]

    Gaertig and K

    E. Gaertig and K. D. Kokkotas, Oscillations of rapidly ro- tating relativistic stars, Phys. Rev. D78, 064063 (2008), arXiv:0809.0629 [gr-qc]

    Pith/arXiv arXiv 2008

  16. [16]

    Breu and L

    C. Breu and L. Rezzolla, Maximum mass, moment of inertia and compactness of relativistic stars, Mon. Not. 7 Roy. Astron. Soc.459, 646 (2016), arXiv:1601.06083 [gr- qc]

    Pith/arXiv arXiv 2016

  17. [17]

    Musolino, C

    C. Musolino, C. Ecker, and L. Rezzolla, On the Max- imum Mass and Oblateness of Rotating Neutron Stars with Generic Equations of State, Astrophys. J.962, 61 (2024), arXiv:2307.03225 [gr-qc]

    Pith/arXiv arXiv 2024

  18. [18]

    Rezzolla, E

    L. Rezzolla, E. R. Most, and L. R. Weih, Using gravitational-wave observations and quasi-universal re- lations to constrain the maximum mass of neutron stars, Astrophys. J. Lett.852, L25 (2018), arXiv:1711.00314 [astro-ph.HE]

    Pith/arXiv arXiv 2018

  19. [19]

    Bauswein, T

    A. Bauswein, T. W. Baumgarte, and H. T. Janka, Prompt merger collapse and the maximum mass of neutron stars, Phys. Rev. Lett.111, 131101 (2013), arXiv:1307.5191 [astro-ph.SR]

    Pith/arXiv arXiv 2013

  20. [20]

    C. J. Krüger and F. Foucart, Estimates for Disk and Ejecta Masses Produced in Compact Binary Mergers, Phys. Rev. D101, 103002 (2020), arXiv:2002.07728 [astro-ph.HE]

    Pith/arXiv arXiv 2020

  21. [21]

    Vretinaris, N

    S. Vretinaris, N. Stergioulas, and A. Bauswein, Empir- ical relations for gravitational-wave asteroseismology of binary neutron star mergers, Phys. Rev. D101, 084039 (2020), arXiv:1910.10856 [gr-qc]

    Pith/arXiv arXiv 2020

  22. [22]

    Manoharan and K

    P. Manoharan and K. D. Kokkotas, Finding universal re- lations using statistical data analysis, Phys. Rev. D109, 103033 (2024), arXiv:2307.13063 [gr-qc]

    Pith/arXiv arXiv 2024

  23. [23]

    Papigkiotis and G

    G. Papigkiotis and G. Pappas, Universal relations for rapidly rotating neutron stars using supervised machine- learning techniques, Phys. Rev. D107, 103050 (2023), arXiv:2303.04273 [astro-ph.HE]

    Pith/arXiv arXiv 2023

  24. [24]

    Papigkiotis, G

    G. Papigkiotis, G. Vardakas, A. Likas, and N. Ster- gioulas, Universal description of a neutron star’s surface and its key global properties: A machine learning ap- proach for nonrotating and rapidly rotating stellar mod- els, Phys. Rev. D111, 083056 (2025), arXiv:2501.18544 [astro-ph.HE]

    Pith/arXiv arXiv 2025

  25. [25]

    Papigkiotis, G

    G. Papigkiotis, G. Vardakas, and N. Stergioulas, Assess- ing Universal Relations for Rapidly Rotating Neutron Stars: Insights from an Interpretable Deep Learning Per- spective, arXiv e-prints (2025), arXiv:2508.05850 [astro- ph.HE]

    arXiv 2025

  26. [26]

    R.Kashyap, A.Dhani,andB.Sathyaprakash,Systematic errors due to quasiuniversal relations in binary neutron stars and their correction for unbiased model selection, Phys. Rev. D106, 123001 (2022), arXiv:2209.02757 [gr- qc]

    Pith/arXiv arXiv 2022

  27. [27]

    Cranmer, J

    K. Cranmer, J. Brehmer, and G. Louppe, The fron- tier of simulation-based inference, Proceedings of the National Academy of Sciences117, 30055 (2020), https://www.pnas.org/doi/pdf/10.1073/pnas.1912789117

    doi:10.1073/pnas.1912789117 2020

  28. [28]

    M. Dax, S. R. Green, J. Gair, J. H. Macke, A. Buonanno, and B. Schölkopf, Real-Time Gravitational Wave Science with Neural Posterior Estimation, Phys. Rev. Lett.127, 241103 (2021), arXiv:2106.12594 [gr-qc]

    Pith/arXiv arXiv 2021

  29. [29]

    M. Dax, S. R. Green, J. Gair, M. Pürrer, J. Wildberger, J. H. Macke, A. Buonanno, and B. Schölkopf, Neural Im- portance Sampling for Rapid and Reliable Gravitational- Wave Inference, Phys. Rev. Lett.130, 171403 (2023), arXiv:2210.05686 [gr-qc]

    Pith/arXiv arXiv 2023

  30. [30]

    M. Dax, S. R. Green, J. Gair, N. Gupte, M. Pürrer, V. Raymond, J. Wildberger, J. H. Macke, A. Buonanno, and B. Schölkopf, Real-time inference for binary neu- tron star mergers using machine learning, Nature639, 49 (2025), arXiv:2407.09602 [gr-qc]

    Pith/arXiv arXiv 2025

  31. [31]

    Crisostomi, K

    M. Crisostomi, K. Dey, E. Barausse, and R. Trotta, Neu- ral posterior estimation with guaranteed exact coverage: The ringdown of GW150914, Phys. Rev. D108, 044029 (2023), arXiv:2305.18528 [gr-qc]

    Pith/arXiv arXiv 2023

  32. [32]

    Sivia and J

    D. Sivia and J. Skilling,Data Analysis: A Bayesian Tu- torial, Oxford science publications (OUP Oxford, 2006)

    2006

  33. [33]

    J. S. Read, B. D. Lackey, B. J. Owen, and J. L. Fried- man,Constraintsonaphenomenologicallyparameterized neutron-star equation of state, Phys. Rev. D79, 124032 (2009), arXiv:0812.2163 [astro-ph]

    Pith/arXiv arXiv 2009

  34. [34]

    S. K. Greif, G. Raaijmakers, K. Hebeler, A. Schwenk, and A.L.Watts,Equationofstatesensitivitieswheninferring neutronstaranddensematterproperties,Mon.Not.Roy. Astron. Soc.485, 5363 (2019), arXiv:1812.08188 [astro- ph.HE]

    Pith/arXiv arXiv 2019

  35. [35]

    M. F. O’Boyle, C. Markakis, N. Stergioulas, and J. S. Read, Parametrized equation of state for neutron star matter with continuous sound speed, Phys. Rev. D102, 083027 (2020), arXiv:2008.03342 [astro-ph.HE]

    Pith/arXiv arXiv 2020

  36. [36]

    Annala, T

    E. Annala, T. Gorda, A. Kurkela, J. Nättilä, and A. Vuorinen, Evidence for quark-matter cores in mas- sive neutron stars, Nature Phys.16, 907 (2020), arXiv:1903.09121 [astro-ph.HE]

    Pith/arXiv arXiv 2020

  37. [38]

    C. J. Krüger and M. Celato, Universal relations for fast rotating neutron stars without equation of state bias, arXiv e-prints (2025), arXiv:2509.11882 [gr-qc]

    arXiv 2025

  38. [39]

    Galassi, J

    M. Galassi, J. Davies, J. Theiler, B. Gough, G. Jung- man, M. Booth, and F. Rossi,GNU Scientific Library Reference Manual (3rd Ed.)(NetworkTheoryLtd.,2009) available athttp://www.gnu.org/software/gsl/

    2009

  39. [40]

    Antoniadiset al., A Massive Pulsar in a Com- pact Relativistic Binary, Science340, 6131 (2013), arXiv:1304.6875 [astro-ph.HE]

    J. Antoniadiset al., A Massive Pulsar in a Com- pact Relativistic Binary, Science340, 6131 (2013), arXiv:1304.6875 [astro-ph.HE]

    Pith/arXiv arXiv 2013

  40. [41]

    Bauswein, O

    A. Bauswein, O. Just, H.-T. Janka, and N. Stergioulas, Neutron-star radius constraints from GW170817 and fu- ture detections, Astrophys. J. Lett.850, L34 (2017), arXiv:1710.06843 [astro-ph.HE]

    Pith/arXiv arXiv 2017

  41. [42]

    Raaijmakers, S

    G. Raaijmakers, S. K. Greif, K. Hebeler, T. Hinderer, S. Nissanke, A. Schwenk, T. E. Riley, A. L. Watts, J. M. Lattimer, and W. C. G. Ho, Constraints on the Dense Matter Equation of State and Neutron Star Prop- erties from NICER’s Mass–Radius Estimate of PSR J0740+6620 and Multimessenger Observations, Astro- phys. J. Lett.918, L29 (2021), arXiv:2105.06981...

    Pith/arXiv arXiv 2021

  42. [43]

    Annala, T

    E. Annala, T. Gorda, A. Kurkela, and A. Vuori- nen, Gravitational-wave constraints on the neutron-star- matter Equation of State, Phys. Rev. Lett.120, 172703 (2018), arXiv:1711.02644 [astro-ph.HE]

    Pith/arXiv arXiv 2018

  43. [44]

    B. P. Abbottet al.(LIGO Scientific, Virgo), GW170817: Measurements of neutron star radii and equation of state, Phys. Rev. Lett.121, 161101 (2018), arXiv:1805.11581 [gr-qc]

    Pith/arXiv arXiv 2018

  44. [45]

    J. B. Hartle, Slowly rotating relativistic stars. 1. Equa- tions of structure, Astrophys. J.150, 1005 (1967)

    1967

  45. [46]

    Hinderer, Erratum: Tidal Love numbers of neutron stars, Astrophys

    T. Hinderer, Erratum: Tidal Love numbers of neutron stars, Astrophys. J.697, 964 (2009), arXiv:0711.2420 [astro-ph]

    Pith/arXiv arXiv 2009

  46. [47]

    Lindblom and S

    L. Lindblom and S. L. Detweiler, The quadrupole oscilla- tions of neutron stars, Astrophys. J. Suppl.53, 73 (1983). 8

    1983

  47. [48]

    S. L. Detweiler and L. Lindblom, On the nonradial pul- sations of general relativistic stellar models, Astrophys. J.292, 12 (1985)

    1985

  48. [49]

    Andersson, K

    N. Andersson, K. D. Kokkotas, and B. F. Schutz, A New numerical approach to the oscillation modes of relativis- tic stars, Mon. Not. Roy. Astron. Soc.274, 1039 (1995), arXiv:gr-qc/9503014

    Pith/arXiv arXiv 1995

  49. [50]

    Tejero-Cantero, J

    Á. Tejero-Cantero, J. Boelts, M. Deistler, J.-M. Lueck- mann, C. Durkan, P. J. Gonçalves, D. S. Greenberg, and J. H. Macke, sbi: A toolkit for simulation-based infer- ence, Journal of Open Source Software5, 2505 (2020)

    2020

  50. [51]

    Tejero-Cantero, J

    Á. Tejero-Cantero, J. Boelts, M. Deistler, J.-M. Lueck- mann, C. Durkan, P. J. Gonçalves, D. S. Greenberg, and J. H. Macke, sbi: A toolkit for simulation-based inference (2022)

    2022

  51. [52]

    Boelts, M

    J. Boelts, M. Deistler, M. Gloeckler, Álvaro Tejero- Cantero, J.-M. Lueckmann, G. Moss, P. Steinbach, T. Moreau, F. Muratore, J. Linhart, C. Durkan, J. Vet- ter, B. K. Miller, M. Herold, A. Ziaeemehr, M. Pals, T. Gruner, S. Bischoff, N. Krouglova, R. Gao, J. K. Lappalainen, B. Mucsányi, F. Pei, A. Schulz, Z. Ste- fanidi, P. Rodrigues, C. Schröder, F. A. Z...

    2025

  52. [53]

    Papamakarios and I

    G. Papamakarios and I. Murray, Fastϵ-free Inference of Simulation Models with Bayesian Conditional Density Estimation (2016), arXiv:1605.06376 [stat.ML]

    Pith/arXiv arXiv 2016

  53. [54]

    Lueckmann, P

    J.-M. Lueckmann, P. J. Goncalves, G. Bassetto, K. Öcal, M. Nonnenmacher, and J. H. Macke, Flexible statisti- cal inference for mechanistic models of neural dynamics (2017), arXiv:1711.01861 [stat.ML]

    Pith/arXiv arXiv 2017

  54. [55]

    D. S. Greenberg, M. Nonnenmacher, and J. H. Macke, Automaticposteriortransformationforlikelihood-freein- ference (2019), arXiv:1905.07488 [cs.LG]

    Pith/arXiv arXiv 2019

  55. [56]

    Deistler, P

    M. Deistler, P. J. Goncalves, and J. H. Macke, Truncated proposalsforscalableandhassle-freesimulation-basedin- ference (2022), arXiv:2210.04815 [stat.ML]

    Pith/arXiv arXiv 2022

  56. [57]

    H. K. Lau, P. T. Leung, and L. M. Lin, Inferring physi- cal parameters of compact stars from their f-mode grav- itational wave signals, Astrophys. J.714, 1234 (2010), arXiv:0911.0131 [gr-qc]

    Pith/arXiv arXiv 2010

  57. [58]

    Yagi and N

    K. Yagi and N. Yunes, I-Love-Q, Science341, 365 (2013), arXiv:1302.4499 [gr-qc]

    Pith/arXiv arXiv 2013

  58. [59]

    T. K. Chan, Y. H. Sham, P. T. Leung, and L. M. Lin, Multipolar universal relations between f-mode frequency and tidal deformability of compact stars, Phys. Rev. D 90, 124023 (2014), arXiv:1408.3789 [gr-qc]

    Pith/arXiv arXiv 2014

  59. [60]

    Konstantinou and S

    A. Konstantinou and S. M. Morsink, Universal Rela- tions for the Increase in the Mass and Radius of a Rotating Neutron Star, Astrophys. J.934, 2 (2022), arXiv:2206.12515 [astro-ph.HE]

    Pith/arXiv arXiv 2022

  60. [61]

    C. J. Krüger and S. H. Völkel, Rapidly rotating neutron stars: Universal relations and EOS inference, Phys. Rev. D108, 124056 (2023), arXiv:2309.05643 [gr-qc]

    Pith/arXiv arXiv 2023

  61. [62]

    S. H. Völkel and C. J. Krüger, Constraining the nuclear equation of state from rotating neutron stars, Phys. Rev. D105, 124071 (2022), arXiv:2203.05555 [gr-qc]

    Pith/arXiv arXiv 2022

  62. [63]

    S. R. Cook, A. Gelman, and D. B. Rubin, Validation of Software for Bayesian Models Using Posterior Quantiles, J. Comp. Graph. Stat.15, 675 (2006)

    2006

  63. [64]

    idealized

    S. Talts, M. Betancourt, D. Simpson, A. Vehtari, and A. Gelman, Validating bayesian inference al- gorithms with simulation-based calibration (2020), arXiv:1804.06788 [stat.ME]. Appendix A: Posterior coverage analysis for SBI and universal relation To assess the extent to which the SBI posteriors and universal relation capture the systematic uncertainty, w...

    Pith/arXiv arXiv 2020