{"id":"83f16b1d-eeeb-4b01-894b-99c5e57ac69d","arxiv_id":"2608.05456","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Transdimensional Bayesian inference with tBilby reconstructs core-collapse supernova gravitational-wave signals in simulated LIGO noise with overlaps up to 85%, and captures the dominant proto-neutron-star mode even at low SNR.","lead":"Researchers tested whether a flexible Bayesian algorithm called tBilby can piece together the messy gravitational-wave signals expected from collapsing supernovae. In simulated LIGO data they recovered up to 85% of the signal, and even low-quality reconstructions preserved the key frequency that tracks the newborn neutron star's size.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The PNS-size inference claim is asserted from spectrograms, not demonstrated: no quantitative radius estimate is produced from low-overlap reconstructions, so the central astrophysical payoff is unverified.","rationale":"Reader's weakest assumption concerns waveform realism. I partially agree, but the more immediate problem is that the paper never performs the inference it claims to enable. The central claim in the abstract and conclusions is explicitly about statements on PNS size; the evidence offered is visual inspection of spectrograms. A quantitative test is straightforward because the injections have known radius histories and universal relations already exist. The frequency cap and simple waveforms also matter, and the manuscript honestly flags them in Section V; they would additionally weaken extrapolation to real signals, but they do not excuse the absence of a direct test of the radius claim. If the quantitative test fails, the paper would still be a useful reconstruction study but the astrophysical payoff would need to be softened; if it passes for the simple models, the remaining obstacle is demonstrating it on more complex, full-band waveforms. A conditional acceptance is therefore appropriate, pending this test.","tokens_in":10554,"tokens_out":5324,"duration_ms":52743,"concrete_test":"Take the SNR=20 and SNR=35 posterior reconstruction samples for s18, y20, and m39 (both wavelet families). For each posterior sample, identify the dominant-mode chirp segment, fit its instantaneous frequency as a function of time, and map f(t) to PNS radius via the universal relation of Torres-Forné et al. (2019) or Powell & Müller (2022). Compare the resulting radius-vs-time posterior with the true radius history of the injected simulation. If the inferred radius is biased by more than ~20% or its credible interval excludes the true value for the low-overlap cases, the central claim that low-overlap reconstructions enable PNS-size statements is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's headline claim has two parts: tBilby can recover up to 85% of simple CCSN signals, and low-overlap reconstructions still capture enough of the dominant PNS mode to make statements about the evolving PNS radius. The first part is supported by the overlap numbers in Figure 3. The second part is not tested. Section IV presents spectrograms at SNR 60, 35, and 20 and states qualitatively that the dominant mode is visible, but the authors never extract a frequency track, apply the universal relations they cite (Refs. [35–37]), or compare a reconstructed PNS radius with the known radius of the injected simulation. At SNR 20, Figure 8 shows that most injections are reconstructed with one or two wavelets covering only the highest-amplitude segment, which the text associates with shock revival rather than later mode evolution. It is therefore an open question whether a low-overlap reconstruction contains enough frequency evolution to constrain the PNS radius within useful uncertainty. The waveform-complexity and 1024 Hz frequency-cap limitations identified in Section V are real, but they are secondary: even for the simple band-limited models used here, the paper does not quantitatively demonstrate the astrophysical inference it claims.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper applies the transdimensional Bayesian inference code tBilby to reconstruct four simulated core-collapse supernova gravitational-wave signals (s18, y20, m39, z85) injected into synthetic Advanced LIGO noise at network SNRs from 20 to 60, using two wavelet dictionaries (sine-Gaussians and chirplets). The authors report noise-weighted overlaps between injected and reconstructed waveforms of up to about 0.85, counts of the number of wavelets used, log Bayes factors for the two dictionaries, and spectrograms of example reconstructions. They conclude that tBilby can capture up to 85% of the CCSN signal features and that, even at low overlap, the dominant proto-neutron-star mode is sufficiently recovered to make statements about the evolving PNS radius.","tokens_in":10777,"tokens_out":7552,"duration_ms":72185,"significance":"The injection-recovery methodology is standard and the overlap computation is well defined; the use of simulated signals with known ground truth avoids circular reasoning, and the comparison between sine-Gaussian and chirplet dictionaries is a useful addition to the burst-reconstruction literature. If the quantitative PNS-radius inference were actually demonstrated, the paper would establish an important bridge between unmodelled burst reconstruction and astrophysical parameter estimation. However, as it stands the novel astrophysical payoff is asserted rather than measured, so the significance of the paper currently rests mainly on the reconstruction quality results.","major_comments":[{"comment":"The paper's central astrophysical claim — that low-overlap reconstructions still capture enough of the dominant mode to make statements about the evolving PNS size — is not tested quantitatively. Section IV (Figures 6–8) supports this by visual inspection of spectrograms only; the authors never extract an instantaneous-frequency track from the reconstructions, never apply the universal relations of Refs. [35–37], and never compare a reconstructed PNS radius with the known radius of the injected waveform. Because the overlap is a single global scalar, equal overlap values do not guarantee equal fidelity of the mode's frequency evolution. I recommend adding a quantitative analysis: estimate the dominant-mode frequency evolution from each reconstruction, apply a universal relation, and report bias and uncertainty in PNS radius as a function of SNR and overlap, using the injected waveform's true radius as ground truth.","section":"Abstract; Section IV; Section V"},{"comment":"At SNR 20, Figure 8 shows that most injections are reconstructed with one or two wavelets, and the text states that these capture only the highest-amplitude part of the signal, often around shock revival rather than the later PNS-mode evolution. This directly conflicts with the low-overlap PNS-size claim: if the reconstruction does not contain the frequency evolution of the mode, it cannot by itself support a radius estimate. The manuscript should state explicitly, with quantitative evidence, the lowest SNR or overlap at which a useful PNS-radius measurement is possible, and should separate shock-revival timing information from PNS-mode frequency evolution.","section":"Section IV, Figure 8"}],"minor_comments":[{"comment":"The printed Fourier-domain wavelet expressions appear to contain typos; for example, the second term in Eq. (2) has \\(\\exp[-Q^2 f/f_0]\\), which is not the standard negative-frequency component of a sine-Gaussian. Please check the equations against the implemented dictionary and correct them or provide the code.","section":"Section III, Eqs. (2)-(3)"},{"comment":"The caption states the overlap is 'calculated using Equation 5', but the overlap is defined in Eq. (4) and Eq. (5) defines the inner product; please correct the cross-reference.","section":"Figure 3 caption"},{"comment":"The text says the number of wavelets is the maximum likelihood value, whereas the figure caption says 'maximum posterior values'; please use consistent terminology and state which point estimate is shown.","section":"Section IV and Figure 4 caption"},{"comment":"There are a few proofreading errors, including 'the SASI mode mode' and 'paramaters'; please correct these.","section":"Section I and Section III"},{"comment":"Please define 'network SNR' explicitly (e.g., quadrature sum of single-detector SNRs) and state whether each plotted point is a single noise realization or an average, since Figures 3–5 show no uncertainty estimates.","section":"Section III"}],"recommendation":"major_revision","confidential_remarks":"The methodological core is sound and the paper is within the journal's scope, but the advertised astrophysical payoff (PNS-radius inference from low-overlap reconstructions) is currently asserted rather than demonstrated. I recommend major revision to add the quantitative inference test. If the editors view the paper strictly as a burst-reconstruction methods study, a minor revision with toned-down astrophysical claims could suffice, but the abstract as written makes the quantitative test necessary."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Plainly: the paper extends tBilby to CCSN waveform reconstruction, and that part mostly works. The injection-recovery setup is standard, the overlap numbers are clear, and the chirplet/sine-Gaussian comparison is new and useful. The problem is the headline claim about inferring PNS size from low-overlap reconstructions: it is supported only by visual inspection of spectrograms, not by any quantitative extraction. The stress-test note is right.\n\nCredit where due. This is the first tBilby application to CCSNe, and the paper does it carefully for four simulated waveforms. It reports overlaps up to ~0.85 for the rotating m39 model, lower for the others, and finds no significant overlap difference between sine Gaussians and chirplets while noting chirplets need fewer wavelets—useful for speed. The limitations are stated openly: 1024 Hz maximum frequency, simple one-mode waveforms, and the explicit acknowledgment that real CCSN signals will be more complex. Those are honest, not hidden.\n\nThe soft spot is the astrophysical inference. The paper says low-overlap reconstructions still capture enough of the dominant mode to 'make statements about the size of the evolving proto-neutron star,' but it never does the extraction. No frequency track is fitted, no universal relation from Refs. [35–37] is applied, and no reconstructed radius is compared with the known injected radius. At SNR 20, most injections are reconstructed with one or two wavelets covering only the highest-amplitude segment—associated by the authors with shock revival, not later mode evolution. That is not enough to constrain PNS radius, and the claim should either be demonstrated quantitatively or substantially tempered. Also, the comparison with BayesWave is not apples-to-apples: the paper cites earlier results from different waveforms and noise realizations, rather than running both algorithms on identical injections. That matters because the 'a little lower than BayesWave' sentence is easy to misread as a direct benchmark.\n\nMinor points: credible intervals are not reported, and only Gaussian design-sensitivity noise is used. Those are acceptable for a first demonstration but should be flagged for the reader.\n\nOverall, the method demonstration is sound and the paper is worth a serious referee. I would send it to review expecting major revisions: either remove the PNS-size claim or add a quantitative radius-recovery test. The code base is established, the waveforms are clearly described, and the limitations are honestly acknowledged—this is competent work, just oversold in the conclusions.","headline":"A solid, honest methods paper for tBilby on CCSN signals whose central astrophysical payoff—PNS radius from low-overlap reconstructions—is asserted from spectrograms, not demonstrated.","tokens_in":11329,"tokens_out":1770,"would_cite":false,"duration_ms":20376,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Transdimensional Bayesian inference reconstructs core-collapse supernova gravitational-wave signals with up to 85% overlap, and the dominant proto-neutron star mode survives even at low overlap.","keywords":["core-collapse supernovae","gravitational-wave burst reconstruction","transdimensional Bayesian inference","tBilby","sine-Gaussian wavelets","chirplets","proto-neutron star","universal relations"],"falsifier":"Run the same tBilby pipeline on a CCSN waveform that includes prompt convection, SASI emission below 200 Hz, and a dominant mode that climbs past 1500 Hz, injected at a signal-to-noise ratio of 20 in the same noise; if the reconstructed spectrogram's dominant-mode track deviates from the injected track by more than the width required by the PNS universal relations, then the claim that low-overlap reconstructions still permit PNS size statements is falsified.","tokens_in":10349,"feed_emoji":"💥","tokens_out":8106,"duration_ms":64538,"temperature":0.7,"pith_summary":"Core-collapse supernovae are expected to emit gravitational waves from the newborn proto-neutron star, but the signals are stochastic, contain multiple features, and span a wide frequency band, so reconstruction cannot rely on template waveforms. This paper shows that transdimensional Bayesian inference, in which the number of wavelets used to describe the signal is itself a sampled parameter, recovers up to 85% of the injected signal's overlap for the loudest simulated models. The key finding is that even when the overlap between injected and reconstructed waveforms is low, the rising-frequency oscillation of the dominant proto-neutron star mode remains visible in the spectrogram. That means a real detection could still yield a measurement of the proto-neutron star's shrinking size through universal relations, even in noisy data at signal-to-noise ratios as low as 20.","feed_headline":"Supernova signal recovered with up to 85% overlap","feed_subtitle":"At SNR 20 the dominant oscillation survives, so the newborn star's size can still be estimated.","key_machinery":"The central object is the transdimensional Bayesian model tBilby, where the number of basis functions $N$ is a free parameter sampled alongside each wavelet's amplitude, central frequency, quality factor, time, and phase. The basis functions are sine-Gaussian wavelets and chirplets; a chirplet adds a frequency-derivative parameter $\\dot{f}_0$, so the wavelet's frequency can sweep with time, and it reduces to a sine Gaussian when $\\dot{f}_0 = 0$. A nested-sampling algorithm draws $N$ and all wavelet parameters jointly, treating parameters beyond the drawn $N$ as ghosts that are marginalized away. This lets the posterior automatically choose how many wavelets the data require, which is what makes the reconstruction morphology-independent.","core_discovery":"The paper claims that the transdimensional Bayesian framework tBilby, previously used for binary black holes, can reconstruct simulated core-collapse supernova gravitational-wave signals added to Gaussian noise at a two-detector design sensitivity. Overlaps between the injected and median-reconstructed waveforms reach about 0.85 for the rapidly rotating model m39 at the highest signal-to-noise ratios, with the other models between about 0.35 and 0.75. Crucially, even at low overlap values the reconstructed spectrogram preserves the dominant proto-neutron star f/g-mode, the feature that sweeps upward in frequency as the proto-neutron star contracts. The paper argues that this preservation, not the raw overlap, is what enables astrophysical inference about the size of the proto-neutron star, and that the method works down to a network signal-to-noise ratio of 20.","pith_inferences":["A natural extension, implicit in the paper, is to feed the reconstructed dominant-mode track directly into the PNS universal relations inside the same Bayesian framework, producing a posterior for the radius rather than a post-processing step.","Because the method's utility lives in spectrogram features rather than overlap, future reconstruction studies could adopt frequency-track recovery as an explicit metric.","The transdimensional approach should be tested on waveforms with multiple simultaneous modes and on the higher-frequency emission that real CCSNe are expected to show; the 1024 Hz cutoff is a computational choice, not a physical one."],"forward_implications":["tBilby becomes a viable morphology-independent reconstruction tool for core-collapse supernova bursts.","A single detected core-collapse supernova in the Milky Way could yield a measurement of the proto-neutron star's radius evolution even at marginal signal-to-noise ratios.","Chirplets match sine-Gaussian reconstruction quality while needing roughly half the wavelets, so faster analyses are possible for the same fidelity.","Overlap alone is not the right figure of merit for CCSN reconstruction: spectrogram-level capture of the dominant mode is what determines astrophysical usability."],"supporting_citations":[{"why":"Introduces the tBilby transdimensional Bayesian inference framework that the paper applies to supernova signals.","marker":"[40]"},{"why":"Provides the s18 non-rotating 18 solar-mass CCSN waveform used as one of the four injections.","marker":"[16]"},{"why":"Provides the y20 and m39 CCSN waveforms, including the rapidly rotating m39 model that reaches the highest overlap.","marker":"[43]"},{"why":"Provides the z85 short, high-mass waveform that rapidly forms a black hole.","marker":"[44]"},{"why":"Earlier BayesWave reconstruction of CCSN signals in Advanced LIGO noise, the baseline with which the overlap values here are compared.","marker":"[31]"},{"why":"Previous comparison of sine-Gaussian and chirplet wavelets for burst reconstruction, which the paper's wavelet comparison extends to CCSN signals.","marker":"[48]"},{"why":"Establishes the SNR range where current CCSN search algorithms can find signals, motivating the SNR 20-60 injections.","marker":"[12]"},{"why":"Provides the universal relation between CCSN gravitational-wave frequency and proto-neutron star size used for astrophysical inference.","marker":"[35]"},{"why":"Shows how the gravitational-wave frequency and universal relations can be used to infer proto-neutron star parameters, the application the reconstructions target.","marker":"[37]"}],"fun_headline_variants":["Transdimensional Bayesian inference reconstructs supernova signals up to 85%","Even low overlap supernova signals reveal neutron star size","New Bayesian method recovers supernova signals with 85% overlap","Supernova gravitational waves reconstructed by transdimensional Bayes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire demonstration rests on four simulated waveforms that the paper itself calls 'the most simple first step,' because each mainly contains a single visible proto-neutron-star mode; real supernova signals are expected to add stochastic convection, SASI emission, extra modes, and frequencies above the 1024 Hz cutoff used here.","fun_headline_variants_meta":{"raw":{"variants":["Transdimensional Bayesian inference reconstructs supernova signals up to 85%","Even low overlap supernova signals reveal neutron star size","New Bayesian method recovers supernova signals with 85% overlap","Supernova gravitational waves reconstructed by transdimensional Bayes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001793,"raw_usage":{"total_tokens":7059,"prompt_tokens":936,"completion_tokens":6123,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":6054}},"tokens_in":552,"tokens_out":6123,"duration_ms":40459,"temperature":1.0,"reasoning_tokens":6054,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T12:43:06.454141+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same tBilby pipeline on a CCSN waveform that includes prompt convection, SASI emission below 200 Hz, and a dominant mode that climbs past 1500 Hz, injected at a signal-to-noise ratio of 20 in the same noise; if the reconstructed spectrogram's dominant-mode track deviates from the injected track by more than the width required by the PNS universal relations, then the claim that low-overlap reconstructions still permit PNS size statements is falsified.","supporting_citations":[{"cited_title":"Transdimensional inference for gravitational-wave astronomy with Bilby","cited_arxiv_id":"2404.04460","evidence_quote":"Introduces the tBilby transdimensional Bayesian inference framework that the paper applies to supernova signals."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the universal relation between CCSN gravitational-wave frequency and proto-neutron star size used for astrophysical inference."}],"review_version":1}