Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Gravitational-wave analyses can use cheaper waveform models for most signals and still recover unbiased black hole mass and spin distributions, cutting analysis cost by about 20%.

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 07:49 UTC pith:NVZIRSO7

load-bearing objection Practical model-selection rule with a real cost saving, but the threshold is calibrated in-sample and the 'unchanged' claim needs a quantitative test before the spin results are fully credible. the 4 major comments →

arxiv 2601.18916 v2 pith:NVZIRSO7 submitted 2026-01-26 gr-qc astro-ph.HEastro-ph.IM

Reconsidering the consistent use of precessing, higher order multipole models for gravitational wave analyses

Charlie Hoy This is my paper
classification gr-qc astro-ph.HEastro-ph.IM
keywords gravitational wavesbinary black holeswaveform modelsspin precessionhigher-order multipoleshierarchical Bayesian inferencecomputational costselection function
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 consistently using the most expensive, most accurate gravitational-wave models is unnecessary for most events. It proposes a selection function based on matched-filter signal-to-noise ratios for precession and higher-order multipoles to decide when cheaper models suffice. Testing on a deliberately extreme population of black hole binaries, a threshold of 1.5 reproduces the mass and spin distributions obtained with the full model, while reducing the cost of Bayesian inference by ~20%. The savings are expected to be much larger for real, astrophysically motivated populations.

Core claim

This paper claims that a simple SNR threshold on the matched-filtered precession SNR ρp and higher-order-multipole SNR ρHM can safely dictate when to use cheaper waveform models. For a worst-case population of 90 rapidly spinning, asymmetric binary black holes, analysing events below the threshold with models that neglect precession and/or higher-order multipoles—and reweighting their spin posteriors—yields inferred mass and spin distributions consistent with consistently using the full PhenomXPHM model. The threshold ρthres = 1.5 produces comparable population estimates and reduces the total cost of hierarchical Bayesian inference by ~20%.

What carries the argument

The selection function built on simple-pe, a fast matched-filtering tool that extracts ρp and ρHM from strain data. simple-pe's outputs assign each event to the cheapest model that still captures its observable physics: PhenomXAS when both SNRs are low, PhenomXHM or PhenomXP when only one is high, and PhenomXPHM when both are high. A companion reweighting-and-conditioning step converts aligned-spin posterior samples from XAS/XHM into comparable precessing-spin distributions, enabling unbiased hierarchical population inference.

Load-bearing premise

The reweighting-and-conditioning step that converts aligned-spin posterior samples into precessing-spin-like distributions must faithfully reconstruct the spin posterior for events that pass the SNR threshold; App. B shows it visibly biases spin magnitudes and tilts at threshold 2.0.

What would settle it

Applying the recommended selection function to a set of real gravitational-wave events (e.g., from GWTC-4) and comparing the hierarchical population hyperparameters against those obtained with consistent PhenomXPHM analysis; if the inferred spin magnitude or tilt distributions differ beyond statistical fluctuations, the central claim fails. Alternatively, injecting a large set of signals with ρp just below 1.5 and checking whether the reweighted XAS posteriors reproduce XPHM posteriors.

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

If this is right

  • If the central claim holds, the standard practice of analysing every binary black hole with precessing, higher-order-multipole models can be relaxed without biasing mass and spin population measurements.
  • The ~20% cost reduction measured on a worst-case population implies larger savings (the paper estimates up to 78%) for real populations where most events have weak precession and multipole power.
  • The selection function may extend to other inference targets, such as the Hubble constant or gravitational lensing, where the same unbiasedness should apply.
  • The matched-filter approach could be incorporated into the sampling itself, avoiding a separate pre-selection step.
  • As waveform models become more accurate and more expensive, the relative computational benefit of this selection strategy will grow.
  • The reweighting-and-conditioning procedure's reliability is only demonstrated for one moderate-precession case; testing it on a suite of injections with varying ρp and ρHM values and comparing against full XPHM posteriors would sharpen confidence in the population-level result.
  • The threshold of 1.5 is tuned to current detector sensitivities; for next-generation detectors with louder signals, the fraction of events with measurable precession/multipoles grows, and the selection approach may need re-optimization.

Where Pith is reading between the lines

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

  • A natural extension would be to include the total network SNR in the threshold, since the paper notes that for loud misidentified events, the cheaper model could bias spins; this trade-off is not yet fully quantified.
  • The reweighting-and-conditioning procedure's reliability is only demonstrated for one moderate-precession case; testing it on a suite of injections with varying ρp and ρHM values and comparing against full XPHM posteriors would sharpen confidence in the population-level result.
  • The threshold of 1.5 is tuned to current detector sensitivities; for next-generation detectors with louder signals, the fraction of events with measurable precession/multipoles grows, and the selection approach may need re-optimization or fail to yield large savings.

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

4 major / 5 minor

Summary. The paper proposes a selection function for gravitational-wave population analyses: using the matched-filter SNRs rho_p and rho_HM (computed with simple-pe) to decide which waveform model to use for each event, reserving the most expensive precessing higher-order-multipole model (PhenomXPHM) for events where these effects are measurable. The selection rule is tested on a simulated 90-event 'worst-case' population of high-spin, asymmetric binaries. The authors report that a threshold rho_thres = 1.5 on rho_p and rho_HM yields inferred mass and spin distributions comparable to those obtained by consistently using XPHM, at a reduced computational cost of about 20% for this population, with larger projected savings for more astrophysical populations. The paper includes single-event and hierarchical Bayesian inference methodology, a reweighting-and-conditioning procedure to convert aligned-spin posteriors from XAS/XHM into approximate precessing-spin posteriors, and appendices examining alternative thresholds, a misidentified event, and a GWTC-like population.

Significance. If the central claim is robust, the paper offers a practical and timely way to reduce the growing computational cost of Bayesian GW population inference without biasing mass and spin measurements. The use of publicly available code and posterior samples is a significant strength, as is the choice of a deliberately pessimistic simulated population. The paper also engages with existing literature on model selection and waveform systematics. However, the validation is currently based on a single simulated catalog, the recommended threshold is selected after inspecting the same data, and the spin-reweighting step is validated on only one moderate-precession case while showing visible bias at a nearby threshold. These issues need quantitative and independent testing before the 'unchanged' population-inference claim can be accepted. The paper's practical value is clear, but at present the evidence is suggestive rather than definitive.

major comments (4)
  1. [Sec. 4, Fig. 5] The recommended threshold rho_thres = 1.5 is chosen after inspecting the results for six thresholds applied to the same 90-event population. The claim that the inferred mass and spin distributions are 'unchanged' rests on visual overlap of 90% credible intervals in a single realization. Since the same data are used to select and to demonstrate performance, this is post-hoc calibration. Please provide an independent validation (e.g., a second simulated population, or a quantitative metric such as the shift in hyperposterior means, a Bayes factor, or a coverage test) that distinguishes 'no detectable difference with 90 events' from 'no bias'.
  2. [Sec. 4, Fig. 4 and App. B] The reweighting-and-conditioning procedure used to convert XAS/XHM aligned-spin posteriors into full spin posteriors is validated only on one moderate-precession case (rho_p = 2.0, rho_HM = 0.6). Appendix B explicitly states that at rho_thres = 2.0 this procedure produces visibly biased spin magnitudes and tilts. Since the same procedure is applied to every event analysed with XAS or XHM at the recommended rho_thres = 1.5, the central claim of unbiased population spin inference depends on this approximation being accurate at the recommended threshold. The manuscript does not quantify the residual bias at rho_thres = 1.5 (e.g., by comparing the reweighted XAS/XHM posteriors to the XPHM posteriors on a population of events with rho_p or rho_HM just below 1.5). Please add such a test.
  3. [App. C] The paper concedes that for a misidentified event with network SNR > 20, XAS would likely recover biased spin estimates, and floats the idea of adding a total-SNR guard ('we could always use XPHM if the total signal power is greater than e.g. 20') but does not implement it. This is a load-bearing caveat because the selection function uses only rho_p and rho_HM, and the worst-case population deliberately contains loud, extreme events. Either implement the total-SNR guard and re-run the population analysis, or demonstrate quantitatively that such loud misidentified events are rare enough in the target population that the population-level spin inference is unaffected.
  4. [Sec. 3 and App. A2] The validation population contains only 90 events, and detectability is modelled by a simplified cut (network SNR > 12) rather than an injection campaign into realistic search pipelines. The 6% false-negative rate is also computed on the same simulated catalog used for the main inference. This is acceptable as a first test, but it means the paper currently demonstrates consistency for one noise realization of one population, not a general property. Please clarify how the quoted 20% cost reduction and the population-level consistency would be affected by a different noise realization, a different population draw, or a more realistic detection selection function.
minor comments (5)
  1. [Abstract and throughout] The phrase 'worse-case scenario' appears repeatedly; this should be 'worst-case scenario'.
  2. [Eq. (1)] The definition of chi_p contains a typo: the second term in the max should be A2 * chi_2 * sin(theta_2), not A2 * chi_1 * sin(theta_1).
  3. [Sec. 2.3] The text says 'Bayes factor’s between the different models' and elsewhere uses an apostrophe in plural forms (e.g., 'binary’s'); please correct the grammar.
  4. [Fig. 5 and Fig. B1] The green/orange/dashed line styles may be difficult to distinguish for color-blind readers; consider adding different line types or markers, and ensure the caption explicitly describes the credibility levels.
  5. [Sec. 4] The claim that 'it is unlikely that the uncertainties will shrink enough to expose differences' for O(1000)-event catalogs is speculative. The cited Monte-Carlo error concerns are real, but they do not remove the need for a quantitative null test at the current 90-event scale.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

No new physical entities, particles, or forces are introduced. The reweighting/conditioning procedure is an algorithmic step, not an entity. The central free parameters are the threshold choices, which are calibrated on the simulated data.

free parameters (3)
  • rho_thres = 1.5
    Recommended SNR threshold on matched-filter rho_p and rho_HM; chosen post hoc because rho_thres=2.0 produces visible spin biases in the same simulated population (Sec. 4, Figs. 5, B1).
  • detection threshold = SNR > 12
    Events are assumed detectable if total network SNR exceeds 12, replacing full search injection campaigns (Sec. 3, App. A2). This affects which events enter the population and the stated cost savings.
  • significant-effect definition = rho_p = rho_HM = 2.0
    A binary is defined as having significant evidence of precession/higher multipoles when true SNR is >= 2.0 (Sec. 3, Fig. 3). This cut defines false-negative rates.
axioms (5)
  • domain assumption The simulated worst-case population (90 events, high spins, isotropic tilts, asymmetric masses) is representative of the most challenging population for which the selection criterion must work.
    The paper uses this population to set the threshold and claims it is an upper limit on false negatives (Sec. 3).
  • domain assumption simple-pe matched-filter rho_p and rho_HM reliably measure the observability of precession and higher multipoles in the strain data.
    The selection function depends on these estimates being trustworthy before running expensive inference (Sec. 3, Fig. 3).
  • domain assumption Idealized Gaussian noise and the SNR>12 detection cut are sufficient to validate the method for real detector data.
    All injections use idealized Gaussian noise and no actual search; the paper states the systematic from the detection cut is neglected (App. A2).
  • ad hoc to paper The reweighting-and-conditioning procedure recovers the spin posterior that would have been obtained with a precessing-spin prior.
    This is a new procedure introduced in Sec. 4; its failure at higher thresholds is visible in App. B.
  • domain assumption Waveform models PhenomXAS/XP/XHM/XPHM form a nested set of sufficient approximations to GR for the purposes of inference.
    The whole method assumes that a model containing the relevant physics is unbiased, while simpler models are unbiased only when rho_p, rho_HM are small (Sec. 2.3).

pith-pipeline@v1.3.0-alltime-deepseek · 20943 in / 13543 out tokens · 142394 ms · 2026-08-03T07:49:52.037265+00:00 · methodology

0 comments
read the original abstract

The growing number of gravitational-wave (GW) observations allows for constraints to be placed on the underlying population of black holes; current estimates show that black hole spins are small, with binaries more likely to have comparable component masses. Since general relativistic effects, such as spin-induced orbital precession and higher order multipole moments, are more likely to be observed for asymmetric binary systems, a direct measurement remains unlikely. Nevertheless, we continue to consistently probe these effects by performing Bayesian inference with our most accurate and computationally expensive models. As the number of GW detections increases, it may soon become infeasible to consistently use these models for analyses. In this paper, we provide a selection criterion that determines when less accurate and computationally cheaper models can be used without giving biased estimates for the population properties of black holes in the Universe. We show that when using our selection criterion, comparable estimates can be obtained for the underlying mass and spin distribution of black holes for a simulated ``worst-case'' scenario population, while reducing the overall cost of performing Bayesian inference on our population by $\sim 20\%$. We anticipate a reduction of up to $78\%$ in the overall cost for an astrophysically motivated population, since there are fewer events with observable spin-precession and higher order multipole power.

Figures

Figures reproduced from arXiv: 2601.18916 by Charlie Hoy.

Figure 1
Figure 1. Figure 1: Plot showing the amplitude of the plus polarization, h+, for two GW signals when assuming different physics; XAS assumes no spin-precession and higher order multipoles, XP assumes no higher order multipoles, XHM assumes no spin-precession, and XPHM includes both spin-precession and higher order multipoles. The Top panel compares the GW signals for an equal mass ratio binary with small spin magnitudes, χ1 =… view at source ↗
Figure 2
Figure 2. Figure 2: Plot showing the inferred primary mass and mass ratio when performing Bayesian inference on the GW signals shown in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Plot comparing the matched filtered SNR in precession ρp and higher order multipoles ρHM to the true SNR for our worst-case scenario population. The shaded red region shows false negatives: a region where we identify the binary as having no evidence for precession and/or higher order multipoles from the filtered SNR despite there being observable features in the signal. The green region shows false positiv… view at source ↗
Figure 4
Figure 4. Figure 4: Plot showing the reweighting and conditioning procedure that is applied to the spin magnitude, χ, and spin tilt, cos θ, posterior distributions obtained with XAS and XHM. The posterior distributions were obtained by analysing a simulated GW signal injected into idealised Gaussian noise. The binary had component masses m1 = 34, M⊙, m2 = 34 M⊙, spin magnitudes χ1 = 0.7, χ2 = 0.8 and spin tilts cos θ1 = 0.4, … view at source ↗
Figure 5
Figure 5. Figure 5: Plot comparing the inferred mass and spin distribution for a simulated highly spinning population. The top and middle panels show the inferred distribution for the primary mass m1 and mass ratio q = m2/m1 respectively. The bottom left and bottom right panels show the inferred spin magnitude, χ, and spin orientation, cos θ, respectively. The black solid line shows the distribution inferred when solely using… view at source ↗
Figure 6
Figure 6. Figure 6: Plot showing the GW signals in our simulated, worst￾case scenario population that were analysed by each model. We show this as a function of ρp and ρHM. We use the recommended SNR threshold of ρthres = 1.5. App. B. As the threshold is further increased, more signif￾icant biases in the spin distribution is seen, with a prefer￾ence for lower spin magnitudes and more aligned binaries. For this reason, althoug… 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 2 Pith papers

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

  1. Mitigating Systematic Errors in Parameter Estimation of Binary Black Hole Mergers in O1-O3 LIGO-Virgo Data

    astro-ph.HE 2026-04 unverdicted novelty 4.0

    Parametric models incorporating waveform phase and amplitude uncertainties mitigate systematic errors in gravitational wave parameter estimation, producing consistent results across models and raw/deglitched data for ...

  2. Mitigating Systematic Errors in Parameter Estimation of Binary Black Hole Mergers in O1-O3 LIGO-Virgo Data

    astro-ph.HE 2026-04 unverdicted novelty 4.0

    Reanalysis of flagged LVK events with waveform uncertainty models produces consistent spin and precession inferences across raw/deglitched data and multiple waveform approximants.

Reference graph

Works this paper leans on

161 extracted references · 14 canonical work pages · cited by 1 Pith paper

  1. [1]

    Aasi J., et al., 2015, @doi [Class. Quant. Grav.] 10.1088/0264-9381/32/7/074001 , 32, 074001

    doi:10.1088/0264-9381/32/7/074001 2015

  2. [2]

    G., et al., 2025d, arXiv:2509.04348

    Abac A. G., et al., 2025d, arXiv:2509.04348

    Pith/arXiv arXiv

  3. [3]

    G., et al., 2025b, arXiv:2508.18081

    Abac A. G., et al., 2025b, arXiv:2508.18081

    Pith/arXiv arXiv

  4. [4]

    G., et al., 2025c, arXiv:2508.18083

    Abac A. G., et al., 2025c, arXiv:2508.18083

    Pith/arXiv arXiv

  5. [5]

    G., et al., 2025e, arXiv:2512.16347

    Abac A. G., et al., 2025e, arXiv:2512.16347

    arXiv

  6. [6]

    G., et al., 2025a, arXiv:2508.18082

    Abac A. G., et al., 2025a, arXiv:2508.18082

    Pith/arXiv arXiv

  7. [7]

    G., et al., 2025f, @doi [Astrophys

    Abac A. G., et al., 2025f, @doi [Astrophys. J. Lett.] 10.3847/2041-8213/ae0d54 , 993, L21

    doi:10.3847/2041-8213/ae0d54 2041

  8. [8]

    Abbott R., et al., 2020a, @doi [Phys. Rev. D] 10.1103/PhysRevD.102.043015 , 102, 043015

    doi:10.1103/physrevd.102.043015

  9. [9]

    Abbott R., et al., 2020b, @doi [Astrophys. J. Lett.] 10.3847/2041-8213/ab960f , 896, L44

    doi:10.3847/2041-8213/ab960f 2041

  10. [10]

    Abbott R., et al., 2021, @doi [Astrophys. J. Lett.] 10.3847/2041-8213/abe949 , 913, L7

    doi:10.3847/2041-8213/abe949 2021

  11. [11]

    Abbott R., et al., 2023a, @doi [Phys. Rev. X] 10.1103/PhysRevX.13.011048 , 13, 011048

    doi:10.1103/physrevx.13.011048

  12. [12]

    Abbott R., et al., 2023b, @doi [Phys. Rev. X] 10.1103/PhysRevX.13.041039 , 13, 041039

    doi:10.1103/physrevx.13.041039

  13. [13]

    Abbott R., et al., 2024, @doi [Phys. Rev. D] 10.1103/PhysRevD.109.022001 , 109, 022001

    doi:10.1103/physrevd.109.022001 2024

  14. [14]

    Acernese F., et al., 2015, @doi [Class. Quant. Grav.] 10.1088/0264-9381/32/2/024001 , 32, 024001

    doi:10.1088/0264-9381/32/2/024001 2015

  15. [15]

    Adams T., et al., 2016, @doi [Class. Quant. Grav.] 10.1088/0264-9381/33/17/175012 , 33, 175012

    doi:10.1088/0264-9381/33/17/175012 2016

  16. [16]

    J.] 10.3847/1538-4357/adda3a , 987, 47

    Agarwal A., et al., 2025, @doi [Astrophys. J.] 10.3847/1538-4357/adda3a , 987, 47

    doi:10.3847/1538-4357/adda3a 2025

  17. [17]

    Ajith P., et al., 2011, @doi [Phys. Rev. Lett.] 10.1103/PhysRevLett.106.241101 , 106, 241101

    doi:10.1103/physrevlett.106.241101 2011

  18. [18]

    Ak c ay S., Hoy C., Mac Uilliam J., 2025, @doi [Phys. Rev. D] 10.1103/rkgr-psrt , 112, 084032

    work page doi:10.1103/rkgr-psrt 2025

  19. [19]

    Akutsu T., et al., 2021, @doi [PTEP] 10.1093/ptep/ptaa125 , 2021, 05A101

    doi:10.1093/ptep/ptaa125 2021

  20. [20]

    A., Cutler C., Sussman G

    Apostolatos T. A., Cutler C., Sussman G. J., Thorne K. S., 1994, @doi [Phys. Rev. D] 10.1103/PhysRevD.49.6274 , 49, 6274

    doi:10.1103/physrevd.49.6274 1994

  21. [21]

    Ashton G., et al., 2019, @doi [Astrophys. J. Suppl.] 10.3847/1538-4365/ab06fc , 241, 27

    doi:10.3847/1538-4365/ab06fc 2019

  22. [22]

    Babak S., Taracchini A., Buonanno A., 2017, @doi [Phys. Rev. D] 10.1103/PhysRevD.95.024010 , 95, 024010

    doi:10.1103/physrevd.95.024010 2017

  23. [23]

    Babak S., Petiteau A., Hewitson M., 2021, arXiv:2108.01167

    Pith/arXiv arXiv 2021

  24. [24]

    Baird E., Fairhurst S., Hannam M., Murphy P., 2013, @doi [Phys. Rev. D] 10.1103/PhysRevD.87.024035 , 87, 024035

    doi:10.1103/physrevd.87.024035 2013

  25. [25]

    M., Capano C

    Biwer C. M., Capano C. D., De S., Cabero M., Brown D. A., Nitz A. H., Raymond V., 2019, @doi [Publ. Astron. Soc. Pac.] 10.1088/1538-3873/aaef0b , 131, 024503

    doi:10.1088/1538-3873/aaef0b 2019

  26. [26]

    Boh\'e A., et al., 2017, @doi [Phys. Rev. D] 10.1103/PhysRevD.95.044028 , 95, 044028

    doi:10.1103/physrevd.95.044028 2017

  27. [27]

    A., 2004, Other thesis ( @eprint arXiv 0705.1514 )

    Brown D. A., 2004, Other thesis ( @eprint arXiv 0705.1514 )

    Pith/arXiv arXiv 2004

  28. [28]

    M., P \"u rrer M., 2016, @doi [Phys

    Calder \'o n Bustillo J., Husa S., Sintes A. M., P \"u rrer M., 2016, @doi [Phys. Rev. D] 10.1103/PhysRevD.93.084019 , 93, 084019

    doi:10.1103/physrevd.93.084019 2016

  29. [29]

    E., Gair J

    Canizares P., Field S. E., Gair J. R., Tiglio M., 2013, @doi [Phys. Rev. D] 10.1103/PhysRevD.87.124005 , 87, 124005

    work page doi:10.1103/physrevd.87.124005 2013

  30. [30]

    E., Gair J., Raymond V., Smith R., Tiglio M., 2015, @doi [Phys

    Canizares P., Field S. E., Gair J., Raymond V., Smith R., Tiglio M., 2015, @doi [Phys. Rev. Lett.] 10.1103/PhysRevLett.114.071104 , 114, 071104

    doi:10.1103/physrevlett.114.071104 2015

  31. [31]

    J.] 10.1088/0004-637X/748/2/136 , 748, 136

    Cannon K., et al., 2012, @doi [Astrophys. J.] 10.1088/0004-637X/748/2/136 , 748, 136

    doi:10.1088/0004-637x/748/2/136 2012

  32. [32]

    Chatziioannou K., Cornish N., Klein A., Yunes N., 2015, @doi [Astrophys. J. Lett.] 10.1088/2041-8205/798/1/L17 , 798, L17

    doi:10.1088/2041-8205/798/1/l17 2015

  33. [33]

    Chu Q., et al., 2022, @doi [Phys. Rev. D] 10.1103/PhysRevD.105.024023 , 105, 024023

    doi:10.1103/physrevd.105.024023 2022

  34. [34]

    Chua A. J. K., Vallisneri M., 2020, @doi [Phys. Rev. Lett.] 10.1103/PhysRevLett.124.041102 , 124, 041102

    work page doi:10.1103/physrevlett.124.041102 2020

  35. [35]

    Colleoni M., Vidal F. A. R., Garc \' a-Quir \'o s C., Ak c ay S., Bera S., 2025, @doi [Phys. Rev. D] 10.1103/PhysRevD.111.104019 , 111, 104019

    doi:10.1103/physrevd.111.104019 2025

  36. [36]

    Colpi M., et al., 2024, arXiv:2402.07571

    Pith/arXiv arXiv 2024

  37. [37]

    J., 2010, arXiv:1007.4820

    Cornish N. J., 2010, arXiv:1007.4820

    Pith/arXiv arXiv 2010

  38. [38]

    J., 2021, @doi [Phys

    Cornish N. J., 2021, @doi [Phys. Rev. D] 10.1103/PhysRevD.104.104054 , 104, 104054

    doi:10.1103/physrevd.104.104054 2021

  39. [39]

    Cotesta R., Buonanno A., Boh\'e A., Taracchini A., Hinder I., Ossokine S., 2018, @doi [Phys. Rev. D] 10.1103/PhysRevD.98.084028 , 98, 084028

    doi:10.1103/physrevd.98.084028 2018

  40. [40]

    Cotesta R., Marsat S., P\"urrer M., 2020, @doi [Phys. Rev. D] 10.1103/PhysRevD.101.124040 , 101, 124040

    doi:10.1103/physrevd.101.124040 2020

  41. [41]

    R., Gair J., Macke J

    Dax M., Green S. R., Gair J., Macke J. H., Buonanno A., Sch \"o lkopf B., 2021, @doi [Phys. Rev. Lett.] 10.1103/PhysRevLett.127.241103 , 127, 241103

    doi:10.1103/physrevlett.127.241103 2021

  42. [42]

    R., Louppe G., 2020, arXiv:2010.12931

    Delaunoy A., Wehenkel A., Hinderer T., Nissanke S., Weniger C., Williamson A. R., Louppe G., 2020, arXiv:2010.12931

    Pith/arXiv arXiv 2020

  43. [43]

    H., Buonanno A., Estelles H., Gair J., Pfeiffer H

    Dhani A., V \"o lkel S. H., Buonanno A., Estelles H., Gair J., Pfeiffer H. P., Pompili L., Toubiana A., 2025, @doi [Phys. Rev. X] 10.1103/5pks-qz6b , 15, 031036

    work page doi:10.1103/5pks-qz6b 2025

  44. [44]

    K., Dudi R., Tichy W., 2019, @doi [Phys

    Dietrich T., Samajdar A., Khan S., Johnson-McDaniel N. K., Dudi R., Tichy W., 2019, @doi [Phys. Rev. D] 10.1103/PhysRevD.100.044003 , 100, 044003

    doi:10.1103/physrevd.100.044003 2019

  45. [45]

    Essick R., 2023, @doi [Phys. Rev. D] 10.1103/PhysRevD.108.043011 , 108, 043011

    doi:10.1103/physrevd.108.043011 2023

  46. [46]

    Essick R., Farr W., 2022, arXiv:2204.00461

    Pith/arXiv arXiv 2022

  47. [47]

    J.] 10.3847/1538-4357/ad1604 , 962, 169

    Essick R., Fishbach M., 2024, @doi [Astrophys. J.] 10.3847/1538-4357/ad1604 , 962, 169

    doi:10.3847/1538-4357/ad1604 2024

  48. [48]

    Estell\'es H., Ramos-Buades A., Husa S., Garc\' a-Quir\'os C., Colleoni M., Haegel L., Jaume R., 2021, @doi [Phys. Rev. D] 10.1103/PhysRevD.103.124060 , 103, 124060

    doi:10.1103/physrevd.103.124060 2021

  49. [49]

    Estell\'es H., Husa S., Colleoni M., Keitel D., Mateu-Lucena M., Garc\' a-Quir\'os C., Ramos-Buades A., Borchers A., 2022a, @doi [Phys. Rev. D] 10.1103/PhysRevD.105.084039 , 105, 084039

    doi:10.1103/physrevd.105.084039

  50. [50]

    Estell\'es H., Colleoni M., Garc\' a-Quir\'os C., Husa S., Keitel D., Mateu-Lucena M., Planas M. d. L., Ramos-Buades A., 2022b, @doi [Phys. Rev. D] 10.1103/PhysRevD.105.084040 , 105, 084040

    doi:10.1103/physrevd.105.084040

  51. [51]

    Estell \'e s H., Buonanno A., Enficiaud R., Foo C., Pompili L., 2025, arXiv:2506.19911

    arXiv 2025

  52. [52]

    Fairhurst S., Green R., Hoy C., Hannam M., Muir A., 2020a, @doi [Phys. Rev. D] 10.1103/PhysRevD.102.024055 , 102, 024055

    doi:10.1103/physrevd.102.024055

  53. [53]

    Fairhurst S., Green R., Hannam M., Hoy C., 2020b, @doi [Phys. Rev. D] 10.1103/PhysRevD.102.041302 , 102, 041302

    doi:10.1103/physrevd.102.041302

  54. [54]

    A., 2023, @doi [Phys

    Fairhurst S., Hoy C., Green R., Mills C., Usman S. A., 2023, @doi [Phys. Rev. D] 10.1103/PhysRevD.108.082006 , 108, 082006

    doi:10.1103/physrevd.108.082006 2023

  55. [55]

    Fairhurst S., Mills C., Colpi M., Schneider R., Sesana A., Trinca A., Valiante R., 2024, @doi [Mon. Not. Roy. Astron. Soc.] 10.1093/mnras/stae443 , 529, 2116

    work page doi:10.1093/mnras/stae443 2024

  56. [56]

    S., Tonolini F., Murray-Smith R., 2022, @doi [Nature Phys.] 10.1038/s41567-021-01425-7 , 18, 112

    Gabbard H., Messenger C., Heng I. S., Tonolini F., Murray-Smith R., 2022, @doi [Nature Phys.] 10.1038/s41567-021-01425-7 , 18, 112

    doi:10.1038/s41567-021-01425-7 2022

  57. [57]

    Gamboa A., et al., 2025, @doi [Phys. Rev. D] 10.1103/jxrc-z298 , 112, 044038

    doi:10.1103/jxrc-z298 2025

  58. [58]

    Garc\' a-Quir\'os C., Colleoni M., Husa S., Estell\'es H., Pratten G., Ramos-Buades A., Mateu-Lucena M., Jaume R., 2020, @doi [Phys. Rev. D] 10.1103/PhysRevD.102.064002 , 102, 064002

    doi:10.1103/physrevd.102.064002 2020

  59. [59]

    Gerosa D., Bellotti M., 2024, @doi [Class. Quant. Grav.] 10.1088/1361-6382/ad4509 , 41, 125002

    doi:10.1088/1361-6382/ad4509 2024

  60. [60]

    M., 2021, @doi [Phys

    Gerosa D., Mould M., Gangardt D., Schmidt P., Pratten G., Thomas L. M., 2021, @doi [Phys. Rev. D] 10.1103/PhysRevD.103.064067 , 103, 064067

    work page doi:10.1103/physrevd.103.064067 2021

  61. [61]

    N., MacFarlane A

    Goldberg J. N., MacFarlane A. J., Newman E. T., Rohrlich F., Sudarshan E. C. G., 1967, @doi [J. Math. Phys.] 10.1063/1.1705135 , 8, 2155

    doi:10.1063/1.1705135 1967

  62. [62]

    R., Gair J., 2021, @doi [Mach

    Green S. R., Gair J., 2021, @doi [Mach. Learn. Sci. Tech.] 10.1088/2632-2153/abfaed , 2, 03LT01

    doi:10.1088/2632-2153/abfaed 2021

  63. [63]

    R., Simpson C., Gair J., 2020, @doi [Phys

    Green S. R., Simpson C., Gair J., 2020, @doi [Phys. Rev. D] 10.1103/PhysRevD.102.104057 , 102, 104057

    doi:10.1103/physrevd.102.104057 2020

  64. [64]

    K., Du Z., Wen L., Gu Y., 2018, @doi [Comput

    Guo X., Chu Q., Chung S. K., Du Z., Wen L., Gu Y., 2018, @doi [Comput. Phys. Commun.] 10.1016/j.cpc.2018.05.002 , 231, 62

    doi:10.1016/j.cpc.2018.05.002 2018

  65. [65]

    Hamilton E., et al., 2021, @doi [Phys. Rev. D] 10.1103/PhysRevD.104.124027 , 104, 124027

    doi:10.1103/physrevd.104.124027 2021

  66. [66]

    Hamilton E., et al., 2025, arXiv:2507.02604

    Pith/arXiv arXiv 2025

  67. [67]

    Hanna C., et al., 2020, @doi [Phys. Rev. D] 10.1103/PhysRevD.101.022003 , 101, 022003

    doi:10.1103/physrevd.101.022003 2020

  68. [68]

    Hannam M., Schmidt P., Boh\'e A., Haegel L., Husa S., Ohme F., Pratten G., P\"urrer M., 2014, @doi [Phys. Rev. Lett.] 10.1103/PhysRevLett.113.151101 , 113, 151101

    doi:10.1103/physrevlett.113.151101 2014

  69. [69]

    Hannam M., et al., 2022, @doi [Nature] 10.1038/s41586-022-05212-z , 610, 652

    doi:10.1038/s41586-022-05212-z 2022

  70. [70]

    R., et al., 2020, @doi [Nature] 10.1038/s41586-020-2649-2 , 585, 357

    Harris C. R., et al., 2020, @doi [Nature] 10.1038/s41586-020-2649-2 , 585, 357

    doi:10.1038/s41586-020-2649-2 2020

  71. [71]

    Harry I., Hoy C., 2025, arXiv: 2503.09773

    arXiv 2025

  72. [72]

    Heinzel J., Vitale S., 2025, arXiv:2509.07221

    Pith/arXiv arXiv 2025

  73. [73]

    Hild S., et al., 2011, @doi [Class. Quant. Grav.] 10.1088/0264-9381/28/9/094013 , 28, 094013

    doi:10.1088/0264-9381/28/9/094013 2011

  74. [74]

    Hoy C., 2022, @doi [Phys. Rev. D] 10.1103/PhysRevD.106.083003 , 106, 083003

    doi:10.1103/physrevd.106.083003 2022

  75. [75]

    Hoy C., Leyde K., 2025, arXiv:2511.07551

    Pith/arXiv arXiv 2025

  76. [76]

    Hoy C., Raymond V., 2021, @doi [SoftwareX] 10.1016/j.softx.2021.100765 , 15, 100765

    arXiv 2021

  77. [77]

    Hoy C., Mills C., Fairhurst S., 2022, @doi [Phys. Rev. D] 10.1103/PhysRevD.106.023019 , 106, 023019

    work page doi:10.1103/physrevd.106.023019 2022

  78. [78]

    R., Nuttall L

    Hoy C., Weaving C. R., Nuttall L. K., Harry I., 2024, @doi [Class. Quant. Grav.] 10.1088/1361-6382/ad8f26 , 41, 245012

    work page doi:10.1088/1361-6382/ad8f26 2024

  79. [79]

    E., 2025a, @doi [Nature Astron.] 10.1038/s41550-025-02579-7 , 9, 1256

    Hoy C., Akcay S., Mac Uilliam J., Thompson J. E., 2025a, @doi [Nature Astron.] 10.1038/s41550-025-02579-7 , 9, 1256

    doi:10.1038/s41550-025-02579-7

  80. [80]

    Hoy C., Fairhurst S., Mandel I., 2025b, @doi [Phys. Rev. D] 10.1103/PhysRevD.111.023037 , 111, 023037

    doi:10.1103/physrevd.111.023037

Showing first 80 references.