pith. machine review for the scientific record. sign in

arxiv: 2604.04957 · v2 · submitted 2026-04-03 · 🌌 astro-ph.IM · gr-qc· physics.data-an

Recognition: 2 theorem links

· Lean Theorem

FluxMC: Rapid and High-Fidelity Inference for Space-Based Gravitational-Wave Observations

Bo Liang , Chang Liu , Hanlin Song , Tianyu Zhao , Minghui Du , He Wang , Haohao Gu , Sensen He
show 6 more authors
Yuxiang Xu Wei-Liang Qian Li-e Qiang Peng Xu Ziren Luo Mingming Sun
Authors on Pith no claims yet

Pith reviewed 2026-05-13 18:25 UTC · model grok-4.3

classification 🌌 astro-ph.IM gr-qcphysics.data-an
keywords gravitational wavesBayesian inferenceMCMCflow matchingmassive black hole binariesparameter estimationspace-based detectorsmachine learning
0
0 comments X

The pith

FluxMC integrates flow matching with parallel tempering MCMC to converge high-fidelity gravitational-wave inferences for massive black hole binaries in hours rather than days.

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

The paper introduces FluxMC as a hybrid sampler that uses flow matching to supply a global approximation of the posterior and then hands control to parallel tempering Markov Chain Monte Carlo for local refinement and asymptotic guarantees. Applied to space-based gravitational-wave signals from massive black hole binaries, the method permits the use of detailed IMRPhenomHM waveforms without the prohibitive cost that traps conventional samplers. It reaches stable convergence in under five hours on these signals while standard parallel tempering MCMC shows no convergence after hundreds of hours and yields Jensen-Shannon divergences of order 0.1. The same framework also removes local-optima biases when faster but approximate waveform models are employed.

Core claim

FluxMC (Flow-guided Unbiased eXploration Monte Carlo) combines a flow-matching generative model with parallel tempering MCMC so that the flow supplies global transport while the tempered chains enforce rigorous convergence. On massive black hole binary injections analyzed with the high-fidelity IMRPhenomHM waveform, FluxMC reaches robust posterior convergence in less than five hours; conventional parallel tempering MCMC fails to converge after hundreds of hours and produces distributional errors measured by Jensen-Shannon divergence of O(10^{-1}). For computationally cheaper IMRPhenomD waveforms the method additionally eliminates systematic biases arising from entrapment in local modes.

What carries the argument

Flow matching, which learns a global approximation to the target posterior and uses it to guide the temperature ladder and proposal steps inside parallel tempering MCMC.

If this is right

  • High-fidelity IMRPhenomHM waveforms become practical for routine massive black hole binary analyses.
  • Distributional errors in recovered parameters drop by two to three orders of magnitude relative to untempered or unguided MCMC.
  • Systematic biases from local-optima entrapment are removed even when using reduced-fidelity but faster waveform models.
  • Parameter estimation speed no longer forces a compromise between waveform accuracy and analysis feasibility.

Where Pith is reading between the lines

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

  • The same flow-guided tempering structure could be applied to other high-dimensional, multimodal inference problems in astrophysics such as binary pulsar timing or exoplanet atmosphere retrieval.
  • Once LISA data arrive, the method could support near-real-time or follow-up analyses of detected events without requiring waveform approximations.
  • Alternative generative models could replace flow matching while preserving the MCMC convergence guarantees, provided they supply a sufficiently global posterior approximation.
  • Testing FluxMC on signals with more parameters or on injected signals with known higher-order modes would directly probe the scalability of the guidance step.

Load-bearing premise

The flow-matching approximation to the posterior is accurate enough that it does not inject new systematic biases into the final MCMC samples.

What would settle it

Generate a long reference posterior for the same massive black hole binary injection using a very expensive traditional sampler, then compare the one-dimensional marginals recovered by FluxMC against that reference to within sampling uncertainty.

read the original abstract

Bayesian inference in the physical sciences faces a fundamental challenge: the imperative for high-fidelity physical modeling often clashes with the intrinsic limitations of stochastic sampling algorithms. Complex, high-dimensional parameter spaces expose the universal vulnerability of conventional methods, e.g., Markov Chain Monte Carlo (MCMC), which struggle with the prohibitive costs of likelihood evaluations and the risk of entrapment in local optima. To resolve this impasse, we introduce FluxMC (Flow-guided Unbiased eXploration Monte Carlo), a machine learning-enhanced framework designed to shift the inference paradigm from blind local search to globally guided transport. It integrates Flow Matching with Parallel Tempering MCMC, effectively combining the global foresight of generative AI with the rigorous asymptotic convergence and local robustness of temperature-based sampling. We showcase the efficacy of this framework through the lens of space-based gravitational-wave (GW) astronomy -- a field representing the frontier of challenging parameter inversion. In the analysis of massive black hole binaries using high-fidelity waveforms (IMRPhenomHM), FluxMC achieves robust convergence in under five hours, whereas traditional Parallel Tempering MCMC fails to converge even after hundreds of hours, yielding high Jensen-Shannon divergences (JSD) of $O(10^{-1})$. Our method reduces the distributional error by two to three orders of magnitude. Furthermore, for computationally efficient models (IMRPhenomD), it eliminates systematic biases caused by local-optima entrapment. Ultimately, FluxMC removes the necessity to compromise between model accuracy and analysis speed, establishing a new computational foundation where scientific discovery is limited only by observational data quality, not by algorithmic capacity.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 2 minor

Summary. The manuscript introduces FluxMC, a hybrid inference framework that combines flow matching with parallel tempering MCMC for Bayesian parameter estimation of massive black hole binaries in space-based gravitational-wave data. It claims that the method achieves robust convergence in under five hours using high-fidelity IMRPhenomHM waveforms, whereas standard PTMCMC fails to converge after hundreds of hours, while reducing Jensen-Shannon divergence by two to three orders of magnitude and removing local-optima biases for simpler IMRPhenomD models.

Significance. If the central claims hold after detailed validation, this would constitute a meaningful methodological contribution to gravitational-wave astronomy. The approach preserves the asymptotic correctness guarantees of tempered MCMC while using a learned global transport map to accelerate mixing in high-dimensional spaces, potentially allowing routine analyses with expensive waveform models that are currently intractable.

major comments (3)
  1. [Abstract] Abstract: the performance claims (convergence in <5 h, 2–3 order JSD reduction) are presented without any description of the flow-matching training procedure, the precise manner in which the learned density enters the PTMCMC transition kernel, or quantitative checks that approximation error lies below the posterior width, all of which are required to substantiate the unbiasedness assertion.
  2. [Methods] Methods section on FluxMC integration: the manuscript does not specify whether the flow is held fixed after training or updated jointly, how the flow density modifies the Metropolis-Hastings ratio, or the exact support-overlap measure used to guarantee that tempered chains correct residual discrepancies, leaving the skeptic’s concern about possible stationary-distribution bias unaddressed.
  3. [Results] Results: no posterior-predictive diagnostics, comparison against known analytic posteriors, or reported error bars on the JSD values are supplied, so the claim that distributional error is reduced by two to three orders of magnitude without introducing systematic offset cannot be evaluated from the presented evidence.
minor comments (2)
  1. [Abstract] The acronym expansion “Flow-guided Unbiased eXploration Monte Carlo” appears only once; subsequent text should use the acronym consistently or restate the expansion at first use in the main body.
  2. [Figures] Figure captions should state the number of independent chains, the exact definition of the reported JSD (including binning or kernel choice), and whether the flow model was trained on the same noise realization used for the reported runs.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We are grateful to the referee for their detailed and insightful comments, which have helped us improve the clarity and rigor of our presentation. Below we respond point-by-point to the major comments. We have revised the manuscript accordingly to address the concerns.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the performance claims (convergence in <5 h, 2–3 order JSD reduction) are presented without any description of the flow-matching training procedure, the precise manner in which the learned density enters the PTMCMC transition kernel, or quantitative checks that approximation error lies below the posterior width, all of which are required to substantiate the unbiasedness assertion.

    Authors: We acknowledge that the abstract, being concise, omitted key methodological details. In the revised manuscript, we have expanded the abstract to include a brief description of the flow-matching training procedure (pre-training a neural network to match the flow from prior to posterior samples) and the integration into the PTMCMC kernel as a learned proposal distribution. For the unbiasedness, the parallel tempering ensures the stationary distribution is exact, independent of the flow quality, provided the flow is a valid density; we have added a note on the approximation error being validated to be sub-dominant to the posterior uncertainty through cross-validation on simpler models. revision: yes

  2. Referee: [Methods] Methods section on FluxMC integration: the manuscript does not specify whether the flow is held fixed after training or updated jointly, how the flow density modifies the Metropolis-Hastings ratio, or the exact support-overlap measure used to guarantee that tempered chains correct residual discrepancies, leaving the skeptic’s concern about possible stationary-distribution bias unaddressed.

    Authors: The flow is trained offline and held fixed during the PTMCMC sampling to maintain efficiency and avoid online training overhead. The flow density modifies the Metropolis-Hastings ratio by serving as the proposal density q(θ), so the acceptance probability becomes min(1, [p(θ') q(θ)] / [p(θ) q(θ')]), where p is the target posterior. The support-overlap is guaranteed by the tempering schedule, which ensures ergodicity across the temperature ladder, allowing the chains to correct any discrepancies from the flow approximation. We have added explicit pseudocode and a mathematical derivation in the Methods section demonstrating that the stationary distribution remains unbiased. revision: yes

  3. Referee: [Results] Results: no posterior-predictive diagnostics, comparison against known analytic posteriors, or reported error bars on the JSD values are supplied, so the claim that distributional error is reduced by two to three orders of magnitude without introducing systematic offset cannot be evaluated from the presented evidence.

    Authors: We agree that additional diagnostics would strengthen the results. In the revised manuscript, we have included posterior-predictive checks and comparisons to analytic posteriors for the IMRPhenomD models, confirming no systematic biases. For the JSD values, we now report error bars estimated from 10 independent runs. For the high-fidelity IMRPhenomHM waveforms, where analytic posteriors are unavailable, we rely on standard convergence diagnostics such as Gelman-Rubin statistics and autocorrelation times, which support the reported performance gains. We believe these additions allow proper evaluation of the claims. revision: partial

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper presents FluxMC as a hybrid framework that integrates Flow Matching for global transport guidance with Parallel Tempering MCMC to retain asymptotic correctness. Performance claims rest on direct empirical comparisons (convergence time under five hours and JSD reduction of 2-3 orders for IMRPhenomHM MBHB signals versus standard PTMCMC), which are falsifiable against external benchmarks and do not reduce to self-definition, fitted inputs renamed as predictions, or load-bearing self-citations. No uniqueness theorems, ansatzes, or renamings of known results are invoked via prior author work; the central assertion that the flow supplies a useful proposal while MCMC corrects residuals is independent of the reported metrics.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review reveals no explicit free parameters, axioms, or invented entities beyond standard assumptions of MCMC convergence and flow matching training.

pith-pipeline@v0.9.0 · 5639 in / 1071 out tokens · 53737 ms · 2026-05-13T18:25:13.938916+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

62 extracted references · 62 canonical work pages · 4 internal anchors

  1. [1]

    Science 19 Advances2(1), 1501177 (2016) https://doi.org/10.1126/sciadv.1501177 https://www.science.org/doi/pdf/10.1126/sciadv.1501177

    Bonomi, M., Camilloni, C., Cavalli, A., Vendruscolo, M.: Metainfer- ence: A bayesian inference method for heterogeneous systems. Science 19 Advances2(1), 1501177 (2016) https://doi.org/10.1126/sciadv.1501177 https://www.science.org/doi/pdf/10.1126/sciadv.1501177

    work page doi:10.1126/sciadv.1501177 2016

  2. [2]

    Quantum Science and Technology4(2), 024004 (2019) https: //doi.org/10.1088/2058-9565/aaf59e

    Arrazola, J.M., Bromley, T.R., Izaac, J., Myers, C.R., Br´ adler, K., Killoran, N.: Machine learning method for state preparation and gate synthesis on photonic quantum computers. Quantum Science and Technology4(2), 024004 (2019) https: //doi.org/10.1088/2058-9565/aaf59e

    work page doi:10.1088/2058-9565/aaf59e 2019

  3. [3]

    Nature619, 282–287 (2022)

    Layden, D., Mazzola, G., Mishmash, R.V., Motta, M., Wocjan, P., Kim, J.-S., Sheldon, S.: Quantum-enhanced markov chain monte carlo. Nature619, 282–287 (2022)

    work page 2022

  4. [4]

    Lewis, A., Bridle, S.: Cosmological parameters from CMB and other data: A Monte Carlo approach. Phys. Rev. D66, 103511 (2002) https://doi.org/10.1103/ PhysRevD.66.103511 arXiv:astro-ph/0205436

    work page Pith review arXiv 2002

  5. [5]

    (eds.) Promise of Bayesian Inference for Astrophysics, pp

    Loredo, T.J.: In: Feigelson, E.D., Babu, G.J. (eds.) Promise of Bayesian Inference for Astrophysics, pp. 275–297. Springer, New York, NY (1992). https://doi.org/ 10.1007/978-1-4613-9290-3 31 . https://doi.org/10.1007/978-1-4613-9290-3 31

    work page doi:10.1007/978-1-4613-9290-3 1992

  6. [6]

    2019, PASA, 36, e010, doi: 10.1017/pasa.2019.2

    Thrane, E., Talbot, C.: An introduction to bayesian inference in gravitational- wave astronomy: Parameter estimation, model selection, and hierarchical models. Publications of the Astronomical Society of Australia36, 010 (2019) https://doi. org/10.1017/pasa.2019.2

    work page doi:10.1017/pasa.2019.2 2019

  7. [7]

    A practical Bayesian method for gravitational-wave ringdown analysis with multiple modes

    Dong, Y., Wang, Z., Wang, H.-T., Zhao, J., Shao, L.: A practical Bayesian method for gravitational-wave ringdown analysis with multiple modes (2025) https://doi. org/10.1038/s41550-025-02766-6 arXiv:2502.01093 [gr-qc]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1038/s41550-025-02766-6 2025

  8. [8]

    The Journal of Chemical Physics21(6), 1087–1092 (1953) https://doi.org/10.1063/1.1699114

    Metropolis, N., Rosenbluth, A.W., Rosenbluth, M.N., Teller, A.H., Teller, E.: Equation of state calculations by fast computing machines. The Journal of Chemical Physics21(6), 1087–1092 (1953) https://doi.org/10.1063/1.1699114

    work page doi:10.1063/1.1699114 1953

  9. [9]

    General Bayesian updating and the loss-likelihood bootstrap.Biometrika, 106(2):465–478, June 2019

    Hastings, W.K.: Monte carlo sampling methods using markov chains and their applications. Biometrika57(1), 97–109 (1970) https://doi.org/10.1093/biomet/ 57.1.97 https://academic.oup.com/biomet/article-pdf/57/1/97/23940249/57-1- 97.pdf

    work page doi:10.1093/biomet/ 1970

  10. [10]

    Journal of the American Statistical Association106(496), 1317–1330 (2011) https://doi.org/10.1198/jasa.2011.ap10346

    Zhou, Q.: Multi-domain sampling with applications to structural inference of bayesian networks. Journal of the American Statistical Association106(496), 1317–1330 (2011) https://doi.org/10.1198/jasa.2011.ap10346

    work page doi:10.1198/jasa.2011.ap10346 2011

  11. [11]

    Electron

    Woodard, D., Schmidler, S., Huber, M.: Sufficient conditions for torpid mixing of parallel and simulated tempering. Electron. J. Probab.14, 29–780804 (2009) https://doi.org/10.1214/EJP.v14-638

    work page doi:10.1214/ejp.v14-638 2009

  12. [12]

    Arai, S., Kadowaki, T.: Quantum annealing enhanced Markov-Chain Monte 20 Carlo. Sci. Rep.15(1), 21427 (2025) https://doi.org/10.1038/s41598-025-07293-y arXiv:2502.08060 [quant-ph]

    work page doi:10.1038/s41598-025-07293-y 2025

  13. [13]

    P., Abbott, R., et al

    Collaboration, T.L.S., Aasi, J., Abbott, B.P., Abbott, R., Abbott, T., Abernathy, M.R., Ackley, K., Adams, C., Adams, T., Addesso, P., Adhikari, R.X., Adya, V., Affeldt, C., Aggarwal: Advanced ligo. Classical and Quantum Gravity32(7), 074001 (2015) https://doi.org/10.1088/0264-9381/32/7/074001

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

  14. [14]

    2015, Classical and Quantum Gravity, 32, 024001, doi: 10.1088/0264-9381/32/2/024001

    Acernese, F., Agathos, M., Agatsuma, K., Aisa, D., Allemandou, N., Allocca, A., Amarni, J., Astone, P., Balestri, G., Ballardin, G., Barone, F., Baronick, J.-P., Barsuglia, M., Basti, A., Basti, F., Bauer, T.S., Bavigadda, V., Bejger, M., Beker, M.G., Belczynski, C., Bersanetti, D., Bertolini, A., Bitossi, M., Bizouard, M.A., Bloemen, S., Blom, M., Boer, ...

    work page doi:10.1088/0264-9381/32/2/024001 2014

  15. [15]

    Akutsu, T., Ando, M., Arai, K., Arai, Y., Araki, S., Araya, A., Arit- omi, N., Aso, Y., Bae, S., Bae, Y., Baiotti, L., Bajpai, R., Barton, M.A., Cannon, K., Capocasa, E., Chan, M., Chen, C., Chen, K., Chen, Y., Chu, H., Chu, Y.-K., Eguchi, S., Enomoto, Y., Flaminio, R., Fujii, Y., Fukunaga: Overview of kagra: Detector design and construction history. Prog...

    work page doi:10.1093/ptep/ptaa125 2020

  16. [16]

    Amaro-Seoane, P., Audley, H., Babak, S., Baker, J., Barausse, E., Bender, P., Berti, E., Binetruy, P., Born, M., Bortoluzzi, D., et al.: Laser interferometer space antenna (2017)

    work page 2017

  17. [17]

    Baker, J., Bellovary, J., Bender, P.L., Berti, E., Caldwell, R., Camp, J., Conklin, J.W., Cornish, N., Cutler, C., DeRosa, R., et al.: The Laser Interferometer Space Antenna: unveiling the millihertz gravitational wave sky (2019)

    work page 2019

  18. [18]

    National Science Review4(5), 685–686 (2017) https: //doi.org/10.1093/nsr/nwx116

    Hu, W.-R., Wu, Y.-L.: The Taiji Program in Space for gravitational wave physics and the nature of gravity. National Science Review4(5), 685–686 (2017) https: //doi.org/10.1093/nsr/nwx116

    work page doi:10.1093/nsr/nwx116 2017

  19. [20]

    Research0(ja) https://doi.org/10.34133/research

    Liu, H., Wang, J., Tao, W., Qi, K., Wang, S., Gao, R., Li, P., Dong, P., Sha, W., 21 Luo, Z., Hu, W.: Recent development of the laser interferometer for taiji space gravitational wave detection. Research0(ja) https://doi.org/10.34133/research. 1252 https://spj.science.org/doi/pdf/10.34133/research.1252

    work page doi:10.34133/research

  20. [21]

    Classical and Quantum Gravity33(3), 035010 (2016) https://doi.org/ 10.1088/0264-9381/33/3/035010

    Luo, J., Chen, L.-S., Duan, H.-Z., Gong, Y.-G., Hu, S., Ji, J., Liu, Q., Mei, J., Milyukov, V., Sazhin, M.,et al.: Tianqin: a space-borne gravitational wave detector. Classical and Quantum Gravity33(3), 035010 (2016) https://doi.org/ 10.1088/0264-9381/33/3/035010

    work page doi:10.1088/0264-9381/33/3/035010 2016

  21. [22]

    Physical Review D93(2), 024003 (2016)

    Klein, A., Barausse, E., Sesana, A., Petiteau, A., Berti, E., Babak, S., Gair, J., Aoudia, S., Hinder, I., Ohme, F.,et al.: Science with the space-based interfer- ometer elisa: Supermassive black hole binaries. Physical Review D93(2), 024003 (2016)

    work page 2016

  22. [23]

    Science China Physics, Mechanics & Astronomy 69(3), 230412 (2026)

    Shen, P., Han, W.-B., Zhong, W.-X.: Revealing the origin of supermassive black holes with taiji-tianqin network. Science China Physics, Mechanics & Astronomy 69(3), 230412 (2026)

    work page 2026

  23. [24]

    Kyutoku, K., Seto, N.: Gravitational-wave cosmography with LISA and the Hubble tension. Phys. Rev. D95(8), 083525 (2017) https://doi.org/10.1103/ PhysRevD.95.083525 arXiv:1609.07142 [astro-ph.CO]

    work page arXiv 2017

  24. [25]

    Colpi, M., et al.: LISA Definition Study Report (2024) arXiv:2402.07571 [astro- ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  25. [26]

    2023, Living Reviews in Relativity, 26, 2, doi: 10.1007/s41114-022-00041-y

    Seoane, P.A.,et al.: Astrophysics with the Laser Interferometer Space Antenna. Living Rev. Rel.26(1), 2 (2023) https://doi.org/10.1007/s41114-022-00041-y arXiv:2203.06016 [gr-qc]

    work page doi:10.1007/s41114-022-00041-y 2023

  26. [27]

    Determining the Hubble Constant from Gravitational Wave Observations

    Schutz, B.F.: Determining the Hubble constant from gravitational wave observa- tions. Nature323(6086), 310–311 (1986) https://doi.org/10.1038/323310a0

    work page doi:10.1038/323310a0 1986

  27. [28]

    Physical Review D103(8), 083011 (2021)

    Marsat, S., Baker, J.G., Canton, T.D.: Exploring the bayesian parameter esti- mation of binary black holes with lisa. Physical Review D103(8), 083011 (2021)

    work page 2021

  28. [29]

    Kalaghatgi, C., Hannam, M., Raymond, V.: Parameter estimation with a spinning multimode waveform model. Phys. Rev. D101(10), 103004 (2020) https://doi. org/10.1103/PhysRevD.101.103004 arXiv:1909.10010 [gr-qc]

    work page doi:10.1103/physrevd.101.103004 2020

  29. [30]

    London, L., Khan, S., Fauchon-Jones, E., Garc´ ıa, C., Hannam, M., Husa, S., Jim´ enez-Forteza, X., Kalaghatgi, C., Ohme, F., Pannarale, F.: First higher- multipole model of gravitational waves from spinning and coalescing black- hole binaries. Phys. Rev. Lett.120, 161102 (2018) https://doi.org/10.1103/ PhysRevLett.120.161102

    work page 2018

  30. [31]

    Pompili, L.,et al.: Laying the foundation of the effective-one-body waveform 22 models SEOBNRv5: Improved accuracy and efficiency for spinning nonprecessing binary black holes. Phys. Rev. D108(12), 124035 (2023) https://doi.org/10.1103/ PhysRevD.108.124035 arXiv:2303.18039 [gr-qc]

    work page arXiv 2023

  31. [32]

    https: //doi.org/10.5281/zenodo.1037579

    Ellis, J., Haasteren, R.: jellis18/PTMCMCSampler: Official Release (2017). https: //doi.org/10.5281/zenodo.1037579 . https://doi.org/10.5281/zenodo.1037579

    work page doi:10.5281/zenodo.1037579 2017

  32. [33]

    https: //arxiv.org/abs/2211.06397

    Wong, K.W.K., Gabri´ e, M., Foreman-Mackey, D.: flowMC: Normalizing-flow enhanced sampling package for probabilistic inference in Jax (2022). https: //arxiv.org/abs/2211.06397

    work page arXiv 2022

  33. [34]

    Fast gravitational wave parameter estimation without com- promises,

    Wong, K.W.K., Isi, M., Edwards, T.D.P.: Fast gravitational wave parameter estimation without compromises (2023). https://arxiv.org/abs/2302.05333

    work page arXiv 2023

  34. [35]

    https://arxiv.org/abs/2404.11397

    Wouters, T., Pang, P.T.H., Dietrich, T., Broeck, C.V.D.: Robust parameter esti- mation within minutes on gravitational wave signals from binary neutron star inspirals (2025). https://arxiv.org/abs/2404.11397

    work page arXiv 2025

  35. [36]

    Proceedings of the National Academy of Sciences119(10), 2109420119 (2022) https://doi.org/10.1073/pnas.2109420119 https://www.pnas.org/doi/pdf/10.1073/pnas.2109420119

    Gabri´ e, M., Rotskoff, G.M., Vanden-Eijnden, E.: Adaptive monte carlo aug- mented with normalizing flows. Proceedings of the National Academy of Sciences119(10), 2109420119 (2022) https://doi.org/10.1073/pnas.2109420119 https://www.pnas.org/doi/pdf/10.1073/pnas.2109420119

    work page doi:10.1073/pnas.2109420119 2022

  36. [37]

    In: The Eleventh International Conference on Learning Representations (2023)

    Liu, X., Gong, C., liu: Flow straight and fast: Learning to generate and transfer data with rectified flow. In: The Eleventh International Conference on Learning Representations (2023). https://openreview.net/forum?id=XVjTT1nw5z

    work page 2023

  37. [38]

    In: The Eleventh International Conference on Learning Representations (2023)

    Lipman, Y., Chen, R.T.Q., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. In: The Eleventh International Conference on Learning Representations (2023). https://openreview.net/forum?id=PqvMRDCJT9t

    work page 2023

  38. [39]

    Fourier-domainmodulationsandde- lays of gravitational-wave signals,

    Marsat, S., Baker, J.G.: Fourier-domain modulations and delays of gravitational- wave signals. arXiv preprint arXiv:1806.10734 (2018)

    work page arXiv 2018

  39. [40]

    Improving and generalizing flow-based generative models with minibatch optimal transport

    Tong, A., Fatras, K., Malkin, N., Huguet, G., Zhang, Y., Rector-Brooks, J., Wolf, G., Bengio, Y.: Improving and generalizing flow-based generative models with minibatch optimal transport. arXiv preprint arXiv:2302.00482 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  40. [41]

    Khan, S., Husa, S., Hannam, M., Ohme, F., P¨ urrer, M., Forteza, X.J., Boh´ e, A.: Frequency-domain gravitational waves from nonprecessing black-hole binaries. ii. a phenomenological model for the advanced detector era. Physical Review D 93(4), 044007 (2016)

    work page 2016

  41. [42]

    Physical Review D89(6), 062001 (2014) 23

    Karnesis, N., Nofrarias, M., Sopuerta, C.F., Gibert, F., Armano, M., Audley, H., Congedo, G., Diepholz, I., Ferraioli, L., Hewitson, M.,et al.: Bayesian model selection for lisa pathfinder. Physical Review D89(6), 062001 (2014) 23

    work page 2014

  42. [43]

    Classical and Quantum Gravity 25(18), 184026 (2008)

    Babak, S., Baker, J.G., Benacquista, M.J., Cornish, N.J., Crowder, J., Larson, S.L., Plagnol, E., Porter, E.K., Vallisneri, M., Vecchio, A.,et al.: The mock lisa data challenges: from challenge 1b to challenge 3. Classical and Quantum Gravity 25(18), 184026 (2008)

    work page 2008

  43. [44]

    Gong, Y., Cao, Z., Zhao, J., Shao, L.: Including higher harmonics in gravitational- wave parameter estimation and cosmological implications for LISA. Phys. Rev. D108(6), 064046 (2023) https://doi.org/10.1103/PhysRevD.108.064046 arXiv:2308.13690 [astro-ph.HE]

    work page doi:10.1103/physrevd.108.064046 2023

  44. [45]

    Physical Review Letters127(24) (2021) https://doi.org/10.1103/physrevlett.127.241103

    Dax, M., Green, S.R., Gair, J., Macke, J.H., Buonanno, A., Sch¨ olkopf, B.: Real- time gravitational wave science with neural posterior estimation. Physical Review Letters127(24) (2021) https://doi.org/10.1103/physrevlett.127.241103

    work page doi:10.1103/physrevlett.127.241103 2021

  45. [46]

    Farr, W., Graff, P., Raymond, V., Blackburn, K., Christensen, N., Vitale, S., Coughlin, M., Aylott, B., Farr, B., Veitch, J.: Parameter estimation for compact binaries with ground-based gravitational-wave observations using the lalinference (2015)

    work page 2015

  46. [47]

    Marsat, S., Baker, J.G., Dal Canton, T.: Exploring the Bayesian parameter esti- mation of binary black holes with LISA. Phys. Rev. D103(8), 083011 (2021) https://doi.org/10.1103/PhysRevD.103.083011 arXiv:2003.00357 [gr-qc]

    work page doi:10.1103/physrevd.103.083011 2021

  47. [48]

    Weaving, C.R., Nuttall, L.K., Harry, I.W., Wu, S., Nitz, A.: Adapting the PyCBC pipeline to find and infer the properties of gravitational waves from massive black hole binaries in LISA. Class. Quant. Grav.41(2), 025006 (2024) https://doi.org/ 10.1088/1361-6382/ad134d arXiv:2306.16429 [astro-ph.IM]

    work page doi:10.1088/1361-6382/ad134d 2024

  48. [49]

    Plunkett, C., Hourihane, S., Chatziioannou, K.: Concurrent estimation of noise and compact-binary signal parameters in gravitational-wave data. Phys. Rev. D 106, 104021 (2022) https://doi.org/10.1103/PhysRevD.106.104021

    work page doi:10.1103/physrevd.106.104021 2022

  49. [50]

    Vallisneri, M.: Synthetic lisa: Simulating time delay interferometry in a model lisa. Phys. Rev. D71, 022001 (2005) https://doi.org/10.1103/PhysRevD.71.022001

    work page doi:10.1103/physrevd.71.022001 2005

  50. [51]

    Tinto, M., Estabrook, F.B., Armstrong, J.W.: Time delay interferometry with moving spacecraft arrays. Phys. Rev. D69, 082001 (2004) https://doi.org/10. 1103/PhysRevD.69.082001

    work page 2004

  51. [52]

    Estabrook, F.B., Tinto, M., Armstrong, J.W.: Time-delay analysis of lisa grav- itational wave data: Elimination of spacecraft motion effects. Phys. Rev. D62, 042002 (2000) https://doi.org/10.1103/PhysRevD.62.042002

    work page doi:10.1103/physrevd.62.042002 2000

  52. [53]

    Astrophys

    Yuan, Y., Du, M., Lin, X.-y., Zhou, H., Xu, P., Fan, X.: Bayesian Analysis of Wave-optics Gravitationally Lensed Massive Black Hole Binaries with a Space- based Gravitational-wave Detector. Astrophys. J.997(1), 11 (2026) https://doi. org/10.3847/1538-4357/ae29ad arXiv:2509.01888 [astro-ph.HE] 24

    work page doi:10.3847/1538-4357/ae29ad 2026

  53. [54]

    Husa, S., Khan, S., Hannam, M., P¨ urrer, M., Ohme, F., Jim´ enez Forteza, X., Boh´ e, A.: Frequency-domain gravitational waves from nonprecessing black- hole binaries. I. New numerical waveforms and anatomy of the signal. Phys. Rev. D93(4), 044006 (2016) https://doi.org/10.1103/PhysRevD.93.044006 arXiv:1508.07250 [gr-qc]

    work page doi:10.1103/physrevd.93.044006 2016

  54. [55]

    Khan, S., Husa, S., Hannam, M., Ohme, F., P¨ urrer, M., Forteza, X.J., Boh´ e, A.: Frequency-domain gravitational waves from nonprecessing black-hole binaries. ii. a phenomenological model for the advanced detector era. Phys. Rev. D93, 044007 (2016) https://doi.org/10.1103/PhysRevD.93.044007

    work page doi:10.1103/physrevd.93.044007 2016

  55. [56]

    Luo, Z., Wang, Y., Wu, Y., Hu, W., Jin, G.: The taiji program: A concise overview. Progress of Theoretical and Experimental Physics2021(5), 05–108 (2020) https://doi.org/10.1093/ptep/ptaa083 https://academic.oup.com/ptep/article- pdf/2021/5/05A108/37953044/ptaa083.pdf

    work page doi:10.1093/ptep/ptaa083 2020

  56. [57]

    Du, M.,et al.: Towards realistic detection pipelines of Taiji: New challenges in data analysis and high-fidelity simulations of space-based gravitational wave antenna. Sci. China Phys. Mech. Astron.69(4), 249501 (2026) https://doi.org/10.1007/ s11433-025-2870-8 arXiv:2505.16500 [gr-qc]

    work page arXiv 2026

  57. [58]

    https://arxiv.org/abs/1605

    Papamakarios, G., Murray, I.: Fastϵ-free Inference of Simulation Models with Bayesian Conditional Density Estimation (2018). https://arxiv.org/abs/1605. 06376

    work page 2018

  58. [59]

    https://arxiv.org/abs/1806.07366

    Chen, R.T.Q., Rubanova, Y., Bettencourt, J., Duvenaud, D.: Neural Ordinary Differential Equations (2019). https://arxiv.org/abs/1806.07366

    work page arXiv 2019

  59. [60]

    ArXivabs/2305.17161 (2023)

    Dax, M., Wildberger, J., Buchholz, S., Green, S.R., Macke, J.H., Scholkopf, B.: Flow matching for scalable simulation-based inference. ArXivabs/2305.17161 (2023)

    work page arXiv 2023

  60. [61]

    Flow Matching for Generative Modeling

    Lipman, Y., Chen, R.T., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. arXiv preprint arXiv:2210.02747 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  61. [62]

    Machine Learning: Science and Technology5(4), 045040 (2024) https://doi.org/10.1088/2632-2153/ad8da9

    Liang, B., Du, M., Wang, H., Xu, Y., Liu, C., Wei, X., Xu, P., Qiang, L.-e., Luo, Z.: Rapid parameter estimation for merging massive black hole binaries using continuous normalizing flows. Machine Learning: Science and Technology5(4), 045040 (2024) https://doi.org/10.1088/2632-2153/ad8da9

    work page doi:10.1088/2632-2153/ad8da9 2024

  62. [63]

    Liang, B., Liu, C., Zhao, T., Du, M., Liang, M., Shi, R., Guo, H., Xu, Y., Qiang, L.-e., Xu, P., Qian, W.-L., Luo, Z.: Accelerating Stochastic Gravitational Wave Backgrounds Parameter Estimation in Pulsar Timing Arrays with Flow Matching (2024). https://arxiv.org/abs/2412.19169 25 Appendix This appendix presents the complete posterior distribution plots (...

    work page arXiv 2024