Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

Machine-learning consistency tests leave one strong-lensing candidate in GWTC-3: the previously known pair GW170104-GW170814.

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 →

T0 review · deepseek-v4-flash

2026-08-04 22:51 UTC pith:UBBTOVWD

load-bearing objection Solid, well-validated methodology for scalable lensed-GW searches; the GWTC-3 conclusion that only one pair survives is plausible but the LR-based exclusion of one candidate is not robust. the 3 major comments →

arxiv 2509.06901 v1 pith:UBBTOVWD submitted 2025-09-08 gr-qc astro-ph.CO

Machine Learning Assisted Parameter-Space Searches for Lensed Gravitational Waves

classification gr-qc astro-ph.CO
keywords gravitational lensinggravitational wavesstrong lensingnormalizing flowsposterior consistencyGWTC-3parameter estimationlensed event search
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 develops a set of machine-learning statistical tests to find pairs of gravitational-wave events that could be repeated images of the same merger caused by strong gravitational lensing. It uses normalizing flows to approximate the full 15-dimensional posterior of each event, then computes three consistency statistics in a compressed six-parameter detector basis: a non-Gaussian parameter-shift probability, a likelihood-ratio at maximum posterior, and the information content (KL divergence) of the pair. The method is validated on simulated catalogs with known lensing status and then applied to the GWTC-3 catalog of 86 binary black holes, where it finds exactly one pair, GW170104-GW170814, that passes all tests. If the calibration transfers to real data, the approach offers a computationally feasible, calibration-free way to rank lensing candidates as catalogs grow, without the costly joint parameter estimation used in earlier searches.

Core claim

The central claim is that strong-lensing consistency between two gravitational-wave events can be tested without joint parameter estimation or large simulations of lensed and unlensed events. Instead, normalizing-flow approximations of each event's posterior are used to evaluate three non-Gaussian statistics directly: the probability that the parameter difference exceeds zero shift, a likelihood ratio comparing zero shift to the maximum-posterior shift, and the Kullback-Leibler divergence between prior and posterior as a measure of information content. Applied to GWTC-3, the method identifies GW170104-GW170814 as the only significant lensing candidate, a pair previously flagged by evidence-r

What carries the argument

The detector parameter basis: six parameters consisting of per-detector phases, inter-detector time delays, and a cycle count that replaces chirp mass. Normalizing flows, trained on each event's posterior and prior, make it computationally feasible to evaluate the KL divergence, the parameter-shift probability, and the likelihood ratio; the likelihood-ratio statistic is assigned a chi-squared distribution with the effective number of constrained parameters, extending Wilks' theorem to the maximum-posterior point.

Load-bearing premise

The calibration established on high-signal-to-noise, nearly Gaussian simulated events transfers to the real GWTC-3 events, which are lower signal-to-noise and significantly more non-Gaussian, and every normalizing flow is accurate enough that the reported significances—including rejections such as GW190828-GW200129—are not corrupted by flow error.

What would settle it

Take the GWTC-3 events that drive key conclusions, retrain their normalizing flows with independent seeds, and compare the flow-based parameter-shift probability and likelihood ratio at zero shift against direct sample-based calculations for those same pairs; if the flows fail the KS test on those events or the shift probabilities move by more than about one sigma, the reported candidate ranking is not robust.

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

Share X Bluesky LinkedIn Reddit HN

If this is right

  • Pairwise strong-lensing searches can be run at scale: a fast Gaussian preselection trims the O(n^2) pair set, and full non-Gaussian statistics are then evaluated with flow evaluations rather than joint parameter estimation.
  • The non-Gaussian parameter shift in the detector basis is better calibrated and more informative than the overlap basis, and it avoids false rejections caused by mass-spin degeneracy and multimodal localization.
  • The information-content plane (parameter-shift significance vs. KL divergence) provides a practical ranking tool for candidate selection that separates promising, ambiguous, and unlikely lensing pairs.
  • The only surviving candidate in GWTC-3, GW170104-GW170814, matches a previously identified candidate, providing independent validation of the method against more costly techniques.
  • The method is directly extendable to other event classes and to sub-threshold searches, with the caveat that the decision boundary in the information-content plane still needs calibration on larger simulated catalogs.

Where Pith is reading between the lines

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

  • A natural testable extension is to calibrate the n_sigma-D_KL decision boundary on injection catalogs that reproduce GWTC-3's lower signal-to-noise ratios and strongly multimodal posteriors, since the current boundary is deduced from a small simulated sample.
  • The reported significance of individual pairs, especially rejections like GW190828-GW200129, depends on the accuracy of flows for those specific events; a flow that fails the KS test could shift a pair's n_sigma by enough to change the candidate list.
  • The likelihood-ratio statistic's chi-squared approximation is conservative in the sense that its true variance exceeds that of the approximating distribution, so reported rejection significances may be overestimated when parameters are only partially constrained.
  • The same machinery, with the detector basis and information plane, could be applied to future catalogs with thousands of events, where the dominant bottleneck shifts from computational cost to the reliability of flow calibration across heterogeneous event morphologies.

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 develops a machine-learning workflow for identifying strongly lensed gravitational-wave candidates from event parameter posteriors. It extends the detector basis of Ezquiaga et al. with normalizing-flow density estimates, defines a KL-based information-content statistic, a non-Gaussian parameter-shift statistic, and a likelihood-ratio statistic whose null distribution is approximated as chi^2(Neff). Validation is carried out on two simulated catalogs: noise-varying realizations of a high-SNR event and 24 injections containing eight lensed pairs, followed by application to the 86 BBH events of GWTC-3. The central catalog claim is that only GW170104-GW170814 passes all criteria, while GW190512-GW190925 and GW190828-GW200129 are discarded mainly through the likelihood-ratio test. The normalizing-flow code is released as part of the tensiometer package.

Significance. If the calibration holds, this is a valuable methodological step: it replaces kernel-density and evidence-ratio bottlenecks with tractable flow-based statistics, demonstrates that the detector basis is more informative than the overlap basis, publicly releases the code, and independently recovers a previously discussed lensing candidate. The validation is honest about false negatives and about the limitations of the chi^2 approximation. However, the strongest catalog conclusion depends on a likelihood-ratio significance that the authors themselves show can overestimate rejection confidence, and whose null calibration is the weakest of the three estimators they consider. The manuscript is therefore a promising contribution whose headline GWTC-3 claim needs either empirical recalibration or careful softening.

major comments (3)
  1. [Sec. VIII and App. A.2, Eq. (19)] The rejection of GW190828-GW200129 at 4.5 sigma is load-bearing for the claim that GW170104-GW170814 is the only significant candidate. That rejection uses the likelihood-ratio statistic QL with significance assigned via the chi^2(Neff) approximation. Appendix A.2 proves that Var(QL) >= 2 Neff and states that the approximation 'can in principle overestimate the significance of rejection', being exact only when parameters are fully data- or prior-dominated. The text attributes the 4.5 sigma rejection to multimodal localization parameters, precisely the regime where the GLM assumptions underlying the approximation fail. Because the NV null calibration for QL is also the weakest of the three estimators (KS significance 5%, Fig. 5b), the 4.5 sigma number cannot be taken at face value. Please provide an empirical null calibration of QL on realistic low-SNR and multimodal pairs, or explicitly
  2. [Sec. VII A, Fig. 5(b)] The likelihood-ratio p-values on the noise-varying catalog have a KS significance of only 5%. This is above the 1% threshold customarily used to flag problems, but it is marginal, and the NV events are high-SNR and nearly Gaussian. The GWTC-3 pairs are lower SNR and markedly more non-Gaussian, as the paper itself documents in Appendix B. No null calibration is provided in that regime. The paper should either calibrate QL on simulated unlensed pairs with realistic non-Gaussianity or attach a systematic uncertainty to the reported n_sigma values for the likelihood-ratio test, particularly for the pairs that determine the final candidate list.
  3. [Sec. III and Sec. VIII] Only 90% of trained flows pass the KS test at the 5% level, compared with the 95% expected for perfect modeling; the text calls this 'very close to the ideal case'. For the specific events that drive the catalog conclusion (GW190828, GW200129, GW190512, GW190925), an inadequately trained flow could change both the parameter-shift and likelihood-ratio values. Flow variance is shown as error bars in the injection studies but is not propagated into the reported GWTC-3 significance values or the final selection decision. Please state which GWTC-3 flows, if any, fail the KS test and quantify how the final candidate list changes under flow-model uncertainty.
minor comments (5)
  1. [Sec. II, Eq. (1)] Typo in the text: 'the same gravitational wave eventh(t)' should read 'event h(t)'.
  2. [Sec. III] The architecture description '2 log2 Nparams spline flows' should define Nparams explicitly and indicate whether this is the number of parameters after pre-processing.
  3. [Fig. 5] The caption describes panel (a) as the parameter-shift estimator in two bases, but the figure also compares methods in panel (b). Please expand the caption so the two panels are independently described.
  4. [App. A.2] The sentence 'Wilks' theorem ensures that ... Delta theta_f = Delta theta_MAP' is misleading: for a fixed Delta theta_f this is an assumption about the null hypothesis, not a consequence of Wilks' theorem. Please rephrase.
  5. [Throughout] The detector-network name is rendered as 'L VK' with an extra space in several places; likely a typographical artifact.

Circularity Check

0 steps flagged

No significant circularity; the central claim is externally validated and not forced by fitted inputs.

full rationale

The derivation chain is self-contained. Parameter posteriors are taken from standard GW parameter estimation; normalizing flows are validated by KS tests; the KL, parameter-shift, and likelihood-ratio statistics are defined from those posteriors/priors without fitting any parameter to the lensing conclusion. The method is calibrated on two independent simulated catalogs (NV and GW injections) with known lensing status, and the GWTC-3 application uses a preselection rule fixed before application rather than tuned to recover GW170104-GW170814. The detector basis and tensiometer machinery are cited from the authors' prior work, but the paper independently demonstrates the basis' information content and the flows' accuracy; these citations are not used as unexamined constraints. The main caveats — QL's chi^2(Neff) approximation can overestimate rejection confidence (App. A2), and the LR's empirical KS significance on NV is 5% (Fig. 5b) — are calibration limitations that could weaken the rejection of pairs such as GW190828-GW200129, but they are not circular reductions: the significance is not constructed to equal the conclusion. Hence no step reduces by definition to its inputs.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

No new physical entities are introduced. The method rests on standard strong-lensing theory, the fidelity of PE posteriors, and normalizing flow accuracy. The hand-tuned thresholds and the chi^2 approximation are the main added assumptions.

free parameters (3)
  • Preselection threshold for Gaussian shift = n_sigma <= 4
    Chosen to avoid false negatives in the phazap preselection; the value is justified heuristically, not derived.
  • Decision boundary in the n_sigma-D_KL plane = Diagonal line, not quantified
    The classification of promising lensing candidates relies on a hand-drawn boundary; the paper states further injection studies are needed to quantify it.
  • KS-test acceptance threshold for flows = 5%
    Standard significance threshold, but flows failing it are still used in the analysis, so the effective quality control is only statistical.
axioms (5)
  • domain assumption Strong lensing produces repeated chirps differing only by magnification, time delay and Morse phase (Eq. 1).
    Standard result of gravitational lensing in geometric optics, cited from Schneider et al. and Blandford & Narayan.
  • domain assumption The waveform model and the bilby parameter estimation posteriors faithfully characterize each event.
    The entire method operates on PE posteriors; if any posterior is biased, the lensing consistency tests inherit the bias. Invoked throughout Secs. II and V.
  • standard math Normalizing flows can accurately represent GW posteriors, including periodic and multimodal features.
    Theoretical existence from Chen & Gopinath; empirical validation via KS tests in Sec. III. This is the load-bearing tooling assumption.
  • standard math The likelihood ratio test statistic Q_L is approximately chi^2(Neff), extending Wilks' theorem to the MAP point.
    Derived in App. A.2 under the Gaussian Linear Model. The derivation itself shows the variance is >= 2Neff, so the approximation can overestimate significance.
  • domain assumption Projecting one event's posterior into the detector basis of the other preserves the lensing consistency signal.
    Uses the detector basis of Ezquiaga, Hu and Lo; the paper takes the smaller consistency probability over the two orderings to be conservative.

pith-pipeline@v1.3.0-alltime-deepseek · 32372 in / 14993 out tokens · 159530 ms · 2026-08-04T22:51:16.018339+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Machine Learning Assisted Parameter-Space Searches for Lensed Gravitational Waves." pith.science (2026). https://pith.science/paper/UBBTOVWD

@misc{pith2026250906901,
  author       = {Pith},
  title        = {Pith review of: Machine Learning Assisted Parameter-Space Searches for Lensed Gravitational Waves},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UBBTOVWD}},
  note         = {Machine review of arXiv:2509.06901}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

When a gravitational wave encounters a massive object along the line of sight, repeated copies of the original signal may be produced due to gravitational lensing. In this paper, we develop a series of new machine-learning based statistical methods to identify promising strong lensing candidates in gravitational wave catalogs. We employ state-of-the-art normalizing flow generative models to perform statistical calculations on the posterior distributions of gravitational wave events that would otherwise be computationally unfeasible. Our lensing identification strategy, developed on two simulated gravitational wave catalogs that test noise realization and event signal variations, selects event pairs with low parameter differences in the optimal detector basis that also have a high information content and favorable likelihood for coincident parameters. We then apply our method to the GWTC-3 catalog and find a single pair still consistent with the lensing hypothesis. This pair has been previously identified through more costly evidence ratio techniques, but rejected on astrophysical grounds, which further validates our technique.

Figures

Figures reproduced from arXiv: 2509.06901 by Giulia Campailla, Jose Mar\'ia Ezquiaga, Marco Raveri, Wayne Hu.

Figure 1
Figure 1. Figure 1: FIG. 1. Joint marginalized posterior distribution for se [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: (b) for GWTC-3. For the former, all pairs sat￾isfy the SNR cut, and the selection is driven solely by the phazap significance of the parameter difference. No lensed pair is excluded by the significance threshold of 4 sigma. In total the full analysis is performed on 81 pairs out of the 276 in the catalog. For the GWTC-3 pairs, as we can see in [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Relation between KL divergence and posterior/prior volume ratio computed in the detector parameter basis for [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Comparison between KL divergences computed on the 9 overlap parameters and the 6 detector basis parameters. [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Empirical cumulative distribution functions (CDFs) of computed [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Comparison between lensing metrics for noise simulations: Gaussian and non-Gaussian parameter shifts significance, [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Comparison plot of non-Gaussian parameter shift, panel (a), and likelihood ratio significance, panel (b), computed [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Localization posterior distribution for [PITH_FULL_IMAGE:figures/full_fig_p015_8.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Parameter difference posterior distributions for [PITH_FULL_IMAGE:figures/full_fig_p016_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11. Comparison between lensing metrics for simulated injections: Gaussian and non-Gaussian parameter shifts signifi [PITH_FULL_IMAGE:figures/full_fig_p017_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12. Localization (ra, dec) posterior distributions for in [PITH_FULL_IMAGE:figures/full_fig_p018_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13. Lensing identification plane, panel (a), and likelihood ratio verification, panel (b), for simulated events pairs. The [PITH_FULL_IMAGE:figures/full_fig_p019_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14. Lensing identification plane, panel (a), and likelihood ratio verification, panel (b), for the GWTC-3 catalog. Low [PITH_FULL_IMAGE:figures/full_fig_p020_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: FIG. 15. Localization (ra, dec) posterior distributions for [PITH_FULL_IMAGE:figures/full_fig_p020_15.png] view at source ↗
Figure 17
Figure 17. Figure 17: FIG. 17. Comparison between lensing metrics for the GWTC-3 catalog: Gaussian and non-Gaussian parameter shifts [PITH_FULL_IMAGE:figures/full_fig_p025_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: FIG. 18. Posterior distributions for parameter differences [PITH_FULL_IMAGE:figures/full_fig_p026_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: FIG. 19. Posterior distribution for the shift in ∆ [PITH_FULL_IMAGE:figures/full_fig_p026_19.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 2 Pith papers

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

  1. Search for strong lensing of gravitational waves in the binary black hole events from O1-O4a

    gr-qc 2026-07 accept novelty 6.0

    Posterior Overlap 2.0 finds no lensed BBH pairs in O1–O4a (p_L < 0.6% for all pairs) and sets a 90% upper bound of 1.4% on the strong-lensing fraction.

  2. Identifying lensed gravitational waves with physics-informed posterior learning

    gr-qc 2026-07 conditional novelty 6.0

    Fusing a simulation-trained common-source mass posterior with waveform features raises lensed-event detection efficiency from 20.8% to 35.2% at 1% false-positive rate and lowers the SNR for 50% efficiency from 45.3 to 33.5.

Reference graph

Works this paper leans on

66 extracted references · 18 canonical work pages · cited by 2 Pith papers · 5 internal anchors

  1. [1]

    Lensing significance in different parameter bases We begin by comparing the significance of the parame- ter shift and likelihood ratio test statistics in the detector basis and overlap parameter bases, as shown in Fig. 7. Comparing the significance of detector basis and over- lap basis lensing results, we find that in both bases the majority of lensed pai...

  2. [2]

    In particular, based on the conclusions drawn in the previous section, we focus on results obtained in the detector basis only, and we show these results in Fig

    Lensing significance with different estimators In this section, we compare different estimators within the same parameter basis. In particular, based on the conclusions drawn in the previous section, we focus on results obtained in the detector basis only, and we show these results in Fig. 11. For typical pairsn σ is highly correlated between the estimato...

  3. [3]

    Dai and T

    L. Dai and T. Venumadhav, (2017), arXiv:1702.04724 [gr-qc]

  4. [4]

    (6), and we derive the for- mula for the KL divergence averaged over data realiza- tions, Eq

    Proofs of Sec IV A In the following, we demonstrate the validity of the data processing inequality (DPI) for the Kullback- Leibler (KL) divergence, Eq. (6), and we derive the for- mula for the KL divergence averaged over data realiza- tions, Eq. (11), both presented in Sec. IV A. Consider a non-invertible transformationf:θ→ θ′ =f(θ). We express the distri...

  5. [5]

    IV C, along with the method used to approximate its distribution

    Proofs of Sec IV C In this appendix, we provide additional details con- cerning the likelihood ratio test statistic introduced in Sec. IV C, along with the method used to approximate its distribution. Consider the likelihood ratio test statistic defined in Eq. (18). In the GLM formalism it can be rewritten as: QL ≡(∆θ f −∆θ ML)T C−1(∆θf −∆θ ML) (A16) −(∆θ...

  6. [6]

    Schneider, J

    P. Schneider, J. Ehlers, and E. E. Falco,Gravitational Lenses, Astronomy and Astrophysics Library (Springer, 1992)

  7. [7]

    Blandford and R

    R. Blandford and R. Narayan, Astrophys. J.310, 568 (1986)

  8. [8]

    Takahashi and T

    R. Takahashi and T. Nakamura, Astrophys. J.595, 1039 (2003), arXiv:astro-ph/0305055

  9. [9]

    J. M. Ezquiaga, D. E. Holz, W. Hu, M. Lagos, and R. M. Wald, Phys. Rev. D103, 064047 (2021), arXiv:2008.12814 [gr-qc]. 26 0.0 0.5 1.0 P/P max 0.0 0.5 1.0 P/P max −5 0 5 10 15 log L 0.0 0.5 1.0 P/P max 0 2 4 ∆τHL −5 0 5 10 15 log L −10 0 10 ∆τHV −10 0 10 ∆τHV GW190512 180714 vs GW190925 232845 Gaussian approximation FIG. 18. Posterior distributions for par...

  10. [10]

    R. K. L. Lo, L. Vujeva, J. M. Ezquiaga, and J. C. L. Chan, Phys. Rev. Lett.134, 151401 (2025), arXiv:2407.17547 [gr-qc]

  11. [11]

    J. M. Ezquiaga, R. K. L. Lo, and L. Vujeva, (2025), arXiv:2503.22648 [gr-qc]

  12. [12]

    T. T. Nakamura and S. Deguchi, Prog. Theor. Phys. Suppl.133, 137 (1999)

  13. [13]

    C ¸ alı¸ skan, J

    M. C ¸ alı¸ skan, J. M. Ezquiaga, O. A. Hannuksela, and D. E. Holz, Phys. Rev. D107, 063023 (2023), arXiv:2201.04619 [astro-ph.CO]

  14. [14]

    Vujeva, J

    L. Vujeva, J. M. Ezquiaga, R. K. L. Lo, and J. C. L. Chan, (2025), arXiv:2501.02096 [astro-ph.CO]

  15. [15]

    Aasiet al.(LIGO Scientific), Class

    J. Aasiet al.(LIGO Scientific), Class. Quant. Grav.32, 074001 (2015), arXiv:1411.4547 [gr-qc]

  16. [16]

    Acerneseet al.(VIRGO), Class

    F. Acerneseet al.(VIRGO), Class. Quant. Grav.32, 024001 (2015), arXiv:1408.3978 [gr-qc]

  17. [17]

    Akutsuet al.(KAGRA), PTEP2021, 05A101 (2021), arXiv:2005.05574 [physics.ins-det]

    T. Akutsuet al.(KAGRA), PTEP2021, 05A101 (2021), arXiv:2005.05574 [physics.ins-det]

  18. [18]

    K. K. Y. Ng, K. W. K. Wong, T. Broadhurst, and T. G. F. Li, Phys. Rev. D97, 023012 (2018), arXiv:1703.06319 [astro-ph.CO]

  19. [20]

    Abbottet al.(LIGO Scientific, VIRGO), Astrophys

    R. Abbottet al.(LIGO Scientific, VIRGO), Astrophys. J.923, 14 (2021), arXiv:2105.06384 [gr-qc]

  20. [21]

    Janquartet al., Mon

    J. Janquartet al., Mon. Not. Roy. Astron. Soc.526, 3832 (2023), arXiv:2306.03827 [gr-qc]

  21. [23]

    (Table I). The significance of the rejection is higher in the detector basis because the overlap basis does not include phase parameters, so it is re- jecting the lensing consistency just based on source localization parameters. This is simply another ex- ample of same issue seen inGW14-GW21but with different relative significances for the phases and loca...

  22. [24]

    S.-S. Li, S. Mao, Y. Zhao, and Y. Lu, Mon. Not. Roy. Astron. Soc.476, 2220 (2018), arXiv:1802.05089 [astro- ph.CO]

  23. [25]

    F. Xu, J. M. Ezquiaga, and D. E. Holz, Astrophys. J. 929, 9 (2022), arXiv:2105.14390 [astro-ph.CO]

  24. [26]

    Goyal, H

    S. Goyal, H. D., S. J. Kapadia, and P. Ajith, Phys. Rev. D104, 124057 (2021), arXiv:2106.12466 [gr-qc]

  25. [27]

    Magare, A

    S. Magare, A. More, and S. Choudhary, Mon. Not. Roy. Astron. Soc.535, 990 (2024), arXiv:2403.02994 [astro- ph.HE]

  26. [28]

    Chakraborty and S

    A. Chakraborty and S. Mukherjee, Astrophys. J.984, 107 (2025), arXiv:2410.06995 [gr-qc]

  27. [29]

    Haris, A

    K. Haris, A. K. Mehta, S. Kumar, T. Venumadhav, and P. Ajith, (2018), arXiv:1807.07062 [gr-qc]

  28. [30]

    Barsode, S

    A. Barsode, S. Goyal, and P. Ajith, Astrophys. J.980, 258 (2025), arXiv:2412.01278 [gr-qc]

  29. [31]

    R. K. L. Lo and I. Magana Hernandez, Phys. Rev. D 107, 123015 (2023), arXiv:2104.09339 [gr-qc]

  30. [32]

    Janquart, O

    J. Janquart, O. A. Hannuksela, H. K., and C. Van Den Broeck, Mon. Not. Roy. Astron. Soc.506, 5430 (2021), arXiv:2105.04536 [gr-qc]

  31. [33]

    A. K. Y. Li, R. K. L. Lo, S. Sachdev, J. C. L. Chan, E. T. Lin, T. G. F. Li, and A. J. Weinstein (LIGO Scientific, Virgo), Phys. Rev. D107, 123014 (2023), arXiv:1904.06020 [gr-qc]

  32. [34]

    A. K. Y. Li, J. C. L. Chan, H. Fong, A. H. Y. Chong, A. J. Weinstein, and J. M. Ezquiaga, (2023), arXiv:2311.06416 [gr-qc]

  33. [35]

    Goyal, S

    S. Goyal, S. J. Kapadia, J.-R. Cudell, A. K. Y. Li, and J. C. L. Chan, Phys. Rev. D109, 023028 (2024), arXiv:2306.04397 [gr-qc]

  34. [36]

    lenscat: a Public and Community-Contributed Catalog of Known Strong Gravitational Lenses

    L. Vujeva, R. K. L. Lo, J. M. Ezquiaga, and J. C. L. 27 Chan, Phil. Trans. Roy. Soc. Lond. A383, 20240168 (2025), arXiv:2406.04398 [astro-ph.GA]

  35. [37]

    Farah, J

    A. Farah, J. M. Ezquiaga, M. Fishbach, and D. Holz, (2025), arXiv:2507.07964 [astro-ph.HE]

  36. [38]

    C. J. Stone, The Annals of Statistics8, 1348 (1980)

  37. [39]

    D. W. Scott and M. P. Wand, Biometrika78, 197 (1991)

  38. [40]

    Raveri and C

    M. Raveri and C. Doux, Phys. Rev. D104, 043504 (2021), arXiv:2105.03324 [astro-ph.CO]

  39. [41]

    J. M. Ezquiaga, W. Hu, and R. K. L. Lo, Phys. Rev. D 108, 103520 (2023), arXiv:2308.06616 [astro-ph.CO]

  40. [42]

    Christensen and R

    N. Christensen and R. Meyer, Rev. Mod. Phys.94, 025001 (2022), arXiv:2204.04449 [gr-qc]

  41. [43]

    Abbottet al.(LIGO Scientific, KAGRA, VIRGO), Astrophys

    R. Abbottet al.(LIGO Scientific, KAGRA, VIRGO), Astrophys. J.970, 191 (2024), arXiv:2304.08393 [gr-qc]

  42. [44]

    O. A. Hannuksela, K. Haris, K. K. Y. Ng, S. Kumar, A. K. Mehta, D. Keitel, T. G. F. Li, and P. Ajith, Astrophys. J. Lett.874, L2 (2019), arXiv:1901.02674 [gr- qc]

  43. [45]

    Waveform systematics in identifying strongly gravitationally lensed gravitational waves: Posterior overlap method

    A. Garr´ on and D. Keitel, Class. Quant. Grav.41, 015005 (2024), arXiv:2306.12908 [gr-qc]

  44. [46]

    Papamakarios and I

    G. Papamakarios and I. Murray, (2016), arXiv:1605.06376 [stat.ML]

  45. [47]

    D. P. Kingma, T. Salimans, R. Jozefowicz, X. Chen, I. Sutskever, and M. Welling, (2016), arXiv:1606.04934 [cs.LG]

  46. [48]

    D. J. Rezende and S. Mohamed, (2015), arXiv:1505.05770 [stat.ML]

  47. [49]

    Papamakarios, T

    G. Papamakarios, T. Pavlakou, and I. Murray, (2017), arXiv:1705.07057 [stat.ML]

  48. [50]

    Chen and R

    S. Chen and R. Gopinath, inAdvances in Neural Infor- mation Processing Systems, Vol. 13, edited by T. Leen, T. Dietterich, and V. Tresp (MIT Press, 2000)

  49. [51]

    Raveri and W

    M. Raveri and W. Hu, Phys. Rev. D99, 043506 (2019), arXiv:1806.04649 [astro-ph.CO]

  50. [52]

    Durkan, A

    C. Durkan, A. Bekasov, I. Murray, and G. Papamakar- ios, arXiv e-prints (2019), 10.48550/arXiv.1906.04032, arXiv:1906.04032 [stat.ML]

  51. [53]

    Normalizing Flows on Tori and Spheres

    D. Jimenez Rezende, G. Papamakarios, S. Racani` ere, M. S. Albergo, G. Kanwar, P. E. Shanahan, and K. Cranmer, (2020), arXiv:2002.02428 [stat.ML]

  52. [54]

    Raveri, C

    M. Raveri, C. Doux, and S. Pandey, (2024), arXiv:2409.09101 [astro-ph.IM]

  53. [55]

    Coccaro, M

    A. Coccaro, M. Letizia, H. Reyes-Gonzalez, and R. Torre, Symmetry16, 942 (2024), arXiv:2302.12024 [stat.ML]

  54. [56]

    N. J. Beaudry and R. Renner, arXiv e-prints , arXiv:1107.0740 (2011), arXiv:1107.0740 [quant-ph]

  55. [57]

    Albrechtet al., (2006), arXiv:astro-ph/0609591

    A. Albrechtet al., (2006), arXiv:astro-ph/0609591

  56. [58]

    Decision-theoretic justifications for Bayesian hypothesis testing using credible sets

    M. Thulin, arXiv e-prints , arXiv:1210.1066 (2012), arXiv:1210.1066 [math.ST]

  57. [59]

    S. S. Wilks, Annals Math. Statist.9, 60 (1938)

  58. [60]

    Ashtonet al., Astrophys

    G. Ashtonet al., Astrophys. J. Suppl.241, 27 (2019), arXiv:1811.02042 [astro-ph.IM]

  59. [61]

    L. S. Collaboration, https://dcc.ligo.org/LIGO- T2000012-v1/public

  60. [62]

    Abbottet al.(KAGRA, VIRGO, LIGO Scientific), Phys

    R. Abbottet al.(KAGRA, VIRGO, LIGO Scientific), Phys. Rev. X13, 041039 (2023), arXiv:2111.03606 [gr- qc]

  61. [63]

    L. Dai, B. Zackay, T. Venumadhav, J. Roulet, and M. Zaldarriaga, (2020), arXiv:2007.12709 [astro-ph.HE]

  62. [64]

    X. Liu, I. Magana Hernandez, and J. Creighton, Astro- phys. J.908, 97 (2021), arXiv:2009.06539 [astro-ph.HE]. [60]tensiometer, https://github.com/mraveri/tensiometer

  63. [65]

    G. P. Smithet al., Phil. Trans. Roy. Soc. Lond. A383, 20240134 (2025), arXiv:2503.19973 [astro-ph.HE]

  64. [66]

    A. M. Mathai and S. B. Provost, Statistics: A Series of Textbooks and Monographs (1992)

  65. [67]

    H. Liu, Y. Tang, and H. H. Zhang, Computational Statistics & Data Analysis54, 858 (2010)

  66. [68]

    P. B. Patnaik, Biometrika37, 78 (1950)