{"id":"301ee8c9-6f6e-4e16-aa92-adb18e5681e9","arxiv_id":"2412.11749","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A cWB overlap mismatch metric, Δmedian, separates simulated 40 solar mass black hole mergers with e20 ≳ 0.17 from circular ones even when parameter estimation uses only quasicircular waveforms.","lead":"Black hole mergers on wobbly, non-circular orbits can be flagged without running slow special-purpose waveform models, according to this simulation study. The authors show that mismatches between a minimal waveform reconstruction and a circular-orbit model grow with orbital eccentricity, offering a cheap triage test for gravitational wave events.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 4% threshold is calibrated on only three circular injections; without a measured null-hypothesis distribution of Δmedian the e20>0.17 separation could be a cWB reconstruction-fidelity artifact rather than eccentricity.","rationale":"The reader's weakest assumption—that the null distribution is a valid reference and cWB reconstruction fidelity does not vary with source parameters—is indeed the load-bearing point. I sharpen it: the paper does not measure the null-hypothesis distribution of Δmedian at all beyond three circular points, so the 4% threshold has no quantified false-alarm rate. This is not an internal inconsistency; the pipeline logic is coherent and the displayed trends are plausible. However, the central claim as stated ('4% upper bound classifies e20>0.17') is an in-sample calibration, not a demonstrated classifier. The proposed test—circular injections across the parameter space with multiple noise realizations—would settle whether the separation is specific to eccentricity or an artifact of cWB reconstruction systematics. Because the concern is concrete and addressable, and because the reader already reached CONDITIONAL, no verdict change is needed; the conditional status is the right call.","tokens_in":16834,"tokens_out":8544,"duration_ms":85003,"concrete_test":"Run the identical pipeline on at least 20 circular (zero-eccentricity) IMRPhenomXAS injections per mass ratio q=1,2,3, with the same total mass, SNR range, and detector configuration as the eccentric hybrids, each in an independent Gaussian noise realization. Compute Δmedian for every injection and compare the distribution against the 4% threshold. Also include a set of circular injections whose parameters are the median PE-recovered parameters of the eccentric hybrids (i.e., biased higher chirp masses) to directly test the footnote-3 reconstruction-fidelity confound. If any circular injection exceeds Δmedian=4%, the threshold is not specific to eccentricity; if none do, the Fig. 5 trend is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's operational claim is that Δmedian < 4% identifies non-eccentric signals and Δmedian > 4% flags e20 ≳ 0.17 (Sec. III B 2, Fig. 5). This threshold is meaningful only if the sampling distribution of Δmedian under the null hypothesis—circular signal, correct PE model—is known. The paper provides only three zero-eccentricity injections, one per mass ratio, with Δmedian = 0.01, 0.02 and 0.001. No repeated noise realizations, no circular injections at other masses or SNRs, and no estimate of the false-alarm rate are given. Footnote 3 explicitly concedes that cWB reconstruction errors may be sensitive to signal parameters, and Fig. 4(f) shows the null distribution itself shifting at high eccentricity due to loss of low-energy pixels. Because the same PE posterior samples enter both the null and on-source distributions, reconstruction errors do not cancel by construction; their parameter dependence must be calibrated. Without that calibration, the monotonic Δmedian trend in Fig. 5 could be produced by cWB reconstructing higher-mass circular templates differently from true hybrid signals, rather than by eccentricity per se. The 4% line is also chosen in-sample from the same 21 injections used to demonstrate the trend, so the e20 > 0.17 separation is not a validated classifier.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a low-latency pipeline to infer the presence of orbital eccentricity in gravitational-wave signals without using eccentric waveform models for parameter estimation. The authors inject 21 nonspinning eccentric hybrid BBH signals (total mass 40 M_sun, mass ratios q = 1, 2, 3) into Gaussian noise, perform Bayesian parameter estimation with the circular model IMRPhenomXAS, reconstruct the injections with the unmodelled pipeline cWB, and compare on-source overlap distributions with null overlap distributions built from cWB reconstructions of PE posterior samples. They define the figure of merit Δmedian as the difference between the medians of the null and on-source overlap distributions, find that Δmedian grows with eccentricity, and propose a 4% upper bound on Δmedian as a classifier for whether a source should be flagged as eccentric (e20 ≳ 0.17). They also report that ignoring eccentricity biases the recovered chirp mass outside the 90% credible interval for larger e20 values.","tokens_in":17105,"tokens_out":4086,"duration_ms":46067,"significance":"If validated, the proposed method would be a practically useful triage tool: it promises to identify eccentric binaries using only a circular-waveform PE run plus a fast unmodelled reconstruction, without the computational cost of eccentric PE. The paper is clearly written, uses physically motivated hybrid waveforms from Ref. [61], and presents a simple, reproducible overlap metric. The monotonic increase of Δmedian with e20 visible in Fig. 5 is a plausible and falsifiable signature. However, the central classification claim rests on a threshold chosen post hoc from the same 21 injections, with only three zero-eccentricity controls, a single noise realization per injection, and no false-alarm study. The paper also explicitly concedes (footnote 3) that cWB reconstruction errors may depend on signal parameters, which is exactly the confounding channel that would need to be controlled before attributing the Δmedian trend to eccentricity. The work is therefore a promising proof-of-principle rather than an established classifier.","major_comments":[{"comment":"The 4% threshold for Δmedian is selected after inspecting the same 21 injections used to demonstrate the trend, and the paper provides no null-hypothesis distribution of Δmedian for circular signals. Only three zero-eccentricity injections are shown (one per mass ratio), with Δmedian values of 0.01, 0.02, and 0.001, and there are no repeated noise realizations at any parameter point. Consequently, the false-alarm rate of the proposed classifier is undefined, and the e20 > 0.17 separation could in principle be a property of the particular noise realization rather than of eccentricity. The authors should compute the sampling distribution of Δmedian under the null hypothesis (circular injections) over many noise realizations and a grid of masses and SNRs, set the threshold from that distribution, and report the expected false-alarm probability.","section":"Sec. III B 2, Fig. 5"},{"comment":"The method's central assumption is that a deviation between the on-source and null overlap distributions is caused by additional physics absent from the PE model, but footnote 3 concedes that cWB reconstruction errors may be sensitive to variations in signal parameters. Because eccentric injections produce biased PE posteriors (e.g., biased chirp mass, Sec. III A), the on-source distribution may deviate from the null distribution simply because cWB reconstructs the effectively higher-mass circular-like templates differently, rather than because the signal is eccentric. This confounding is load-bearing and untested. A concrete control experiment would be to inject circular signals at the biased PE medians obtained from eccentric injections and measure Δmedian; if the same Δmedian values appear, the metric is not eccentricity-specific.","section":"Sec. II C, footnote 3"},{"comment":"The null injections are described as being placed \"near the event times\" in the GW data, which suggests that the same Gaussian noise realization is used for the on-source reconstruction, the PE posterior, and the null injections. If so, the null overlap distribution may be artificially narrowed or inflated because the same noise appears both in the reconstructed event and in the injected posterior samples. The authors should clarify whether the null injections reuse the same noise realization, and if so, repeat the null analysis in independent noise realizations or in off-source times to assess the impact on the null distribution and on Δmedian.","section":"Sec. II C"},{"comment":"The claims that the method \"can be applied to identify any physical effect of measurable strength\" (Sec. IV) and that it enables \"low-latency inferences of binary properties\" are broader than the evidence presented. The study covers only nonspinning, (2,2)-only hybrid signals with total mass 40 M_sun, mass ratios q = 1, 2, 3, and SNRs in the range 29–45. The generalization to other masses, spins, higher-order modes, and other subdominant effects is not supported by any test. The conclusions should be restricted to the tested parameter region, and the more general applicability should be explicitly labeled as future work.","section":"Sec. IV and abstract"}],"minor_comments":[{"comment":"The abstract says the method works \"in real time,\" but the pipeline includes a full Bayesian PE run; the total latency is not quantified anywhere. Please clarify what 'real time' means here.","section":"Abstract"},{"comment":"The error σ on Δmedian is computed from the 90% credible intervals of the two overlap distributions, but this does not account for noise-realization variance or sampling noise in the medians. At minimum, the text should state that these are distribution-width errors, not total uncertainties on the classifier.","section":"Eq. (4)"},{"comment":"The table title says 'Priors for parameters used in precessing spin recoveries,' but all injections and recoveries are nonspinning. The title appears to be a copy-paste error and should be corrected.","section":"Table I"},{"comment":"The paper uses 'off-source injection,' 'null-sources injections,' and 'null distribution' in adjacent sentences; please unify the terminology to avoid confusion.","section":"Sec. II C"},{"comment":"The caption says the colored horizontal lines denote injected chirp masses, but for some mass ratios these lines may be obscured by the violin plots; consider adding markers or a legend to make the injected values visible.","section":"Fig. 2 caption"}],"recommendation":"major_revision","confidential_remarks":"The reader's report and the stress-test note are largely aligned with my own reading. The core problem is not the physical plausibility of the trend but the absence of a statistical calibration for the 4% classifier: the threshold is post hoc, there is no false-alarm study, and the parameter-dependence of cWB reconstruction errors conceded in footnote 3 is exactly the confound that must be controlled. A well-executed revision with circular control injections at biased PE parameters, repeated noise realizations, and a threshold derived from a measured null distribution would make the central claim defensible. The paper is within the journal's scope, but in its current form the classification claim is not established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me give you the short version. This is a competent proof-of-principle paper, and the core idea is worth taking seriously. The genuinely new piece is the Δmedian figure of merit — the difference between medians of the on-source and null overlap distributions — and the claim that it grows monotonically with injected eccentricity across three mass ratios. Prior work already showed quasicircular PE biases eccentric signals and that cWB consistency tests can flag events like GW190521; this paper systematizes that into a proposed triage threshold. That is a real, incremental contribution.\n\nWhat it does well: the injection setup is sensible, using NR-PN eccentric hybrids from Chattaraj et al., and the PE bias story in Fig. 2 is clean. The cWB reconstruction plots (Fig. 3) are illustrative, and the trend in Fig. 5 is visible even with error bars. The authors are also upfront about the controlled setting: Gaussian noise, nonspinning, (2,2)-only hybrids, and footnote 3 explicitly concedes that cWB reconstruction errors may be parameter-dependent.\n\nWhere it gets soft. The 4% classification threshold is chosen in-sample from the same 21 injections that are used to demonstrate the trend. There are only three zero-eccentricity injections, one per mass ratio, and no repeated noise realizations, no null injections at other masses or SNRs, and no false-alarm estimate. The stress-test concern is legitimate: the monotonic Δmedian trend could in principle be produced by cWB reconstructing higher-mass circular templates differently rather than by eccentricity itself. Since the same posterior samples enter both the null and on-source distributions, reconstruction errors do not cancel by construction; their parameter dependence needs calibration. I don't think this kills the paper — the stability of the null distribution for moderate eccentricity and the monotonic trend make the eccentricity interpretation plausible — but the operational claim that 4% separates e20 > 0.17 is not yet validated. Also, the abstract says e20 ≳ 0.1 while the text and threshold use ~0.15–0.17; that inconsistency should be fixed.\n\nBottom line: This deserves a serious referee. It's a methods paper, not a discovery claim, and the central idea could be operationally useful for low-latency eccentricity triage in O4 and beyond. A referee should push for out-of-sample noise trials, repeated injections, and a measured null-hypothesis distribution of Δmedian. If those hold up, I'd cite it. As it stands, I'd treat the threshold as provisional.","headline":"A useful proof-of-principle for eccentricity triage via cWB overlap distributions, but the 4% threshold is calibrated in-sample and needs repeated-noise validation before operational use.","tokens_in":17632,"tokens_out":2710,"would_cite":false,"duration_ms":24187,"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":"Minimally modelled waveform reconstructions can flag binary eccentricity without eccentric templates.","keywords":["gravitational waves","eccentric binaries","waveform reconstruction","Coherent WaveBurst","parameter estimation biases","binary black hole mergers","low-latency inference","unmodelled searches"],"falsifier":"Run the same cWB overlap analysis on circular injections tuned to the biased masses, mass ratios, and SNRs recovered from the eccentric injections; if those circular injections also push $\\Delta_{\\rm median}$ above 4%, the metric is tracking reconstruction fidelity rather than eccentricity.","tokens_in":16657,"feed_emoji":"🌊","tokens_out":6610,"duration_ms":61335,"temperature":0.7,"pith_summary":"The paper tries to show that unmodelled waveform reconstruction can reveal a physical effect that the waveform model used for parameter estimation ignores, without needing a dedicated model for that effect. Concretely, it injects 21 eccentric 40-solar-mass binary black hole signals (mass ratios 1, 2, 3) into Gaussian noise and shows that cWB reconstructions of the event diverge from reconstructions of quasicircular posterior samples as eccentricity increases. The divergence, measured as $\\Delta_{\\rm median}$ between on-source and null overlap distributions, grows from about 0.01 at zero eccentricity to 0.14–0.20 near $e_{20} \\simeq 0.45$, and a 4% threshold classifies all injections with $e_{20} > 0.17$ as eccentric. The same injections show chirp-mass posteriors biased outside the 90% credible interval once $e_{20} \\gtrsim 0.15$. If the claim holds, this gives a low-latency way to triage gravitational-wave candidates for eccentricity follow-up using pipelines that already exist.","feed_headline":"Waveform mismatch flags eccentric mergers without eccentric templates","feed_subtitle":"In simulated 40-solar-mass mergers, a 4% overlap-distribution gap flags eccentricity that circular templates miss.","key_machinery":"The load-bearing object is the cWB point-estimate reconstruction, produced by inverse wavelet transform with minimal assumptions about signal morphology. The overlap $O(h_1,h_2) = \\langle h_1|h_2\\rangle / \\sqrt{\\langle h_1|h_1\\rangle \\langle h_2|h_2\\rangle}$ is the normalized noise-weighted inner product of whitened waveforms. The null distribution uses overlaps between each PE sample injection and its own cWB reconstruction, quantifying reconstruction error; the on-source distribution uses overlaps between the cWB reconstruction of the injected event and the cWB reconstructions of the PE samples, so it is sensitive to physics missing from the circular recovery model. Their difference in medians, $\\Delta_{\\rm median}$, with a 4% upper-bound criterion, is the figure of merit that classifies a signal as eccentric.","core_discovery":"The central claim is that orbital eccentricity, a subdominant effect, can be isolated even when eccentric waveforms are not readily accessible. The authors establish this by combining three ingredients: parameter estimation with a circular waveform (IMRPhenomXAS), minimally modelled cWB reconstructions of the event and of PE posterior samples, and a comparison of overlap distributions. They report that the discrepancy metric $\\Delta_{\\rm median}$ increases with $e_{20}$ for every mass ratio tested, and that the waveform-consistency test can therefore be repurposed as an eccentricity indicator. On this basis they conclude that the method can infer any subdominant physical effect of measurable strength, such as precession or higher-order modes, and can prioritize candidates for expensive follow-up analyses.","pith_inferences":["The 0.17 eccentricity threshold and the 4% cutoff are calibrated for one total mass, one distance, and one detector network, so they should not be treated as fixed; a broader injection campaign would show how far the criterion generalizes.","A direct test the paper leaves implicit is running the same pipeline on circular injections matched to the biased recovered masses and SNRs to measure how often the 4% classifier raises a false eccentricity alarm.","Because the null distribution inherits the circular-model posterior's biases, the method's sensitivity and false-alarm rate are coupled to the PE model's bias; an eccentric-aware prior or a noise-subtracted null set would decouple them.","If the approach survives real-noise validation, every search pipeline that already produces cWB reconstructions gains an eccentricity screen at essentially no extra computational cost, which could prioritize the small fraction of candidates needing eccentric follow-up."],"forward_implications":["A 4% cutoff on $\\Delta_{\\rm median}$ separates eccentric injections with $e_{20} > 0.17$ from noneccentric ones for 40-solar-mass binaries at matched-filter SNRs of roughly 29–45.","Ignoring eccentricity at $e_{20} \\gtrsim 0.15$ pushes chirp-mass recovery outside the 90% credible interval for all mass ratios studied.","cWB reconstruction plus circular-model parameter estimation can serve as a low-latency eccentricity screen, avoiding expensive eccentric parameter estimation for every candidate.","The same on-source-versus-null overlap logic extends, in principle, to any subdominant effect of measurable strength, including spin precession, higher-order modes, and environmental dephasing.","The demonstrated behaviour is tied to Gaussian noise; real noise will require re-evaluating the threshold and its classification power."],"supporting_citations":[{"why":"Supplies the 21 nonspinning eccentric inspiral-merger-ringdown hybrid waveforms used as injections, with mass ratios q=1,2,3 and eccentricities $e_{20}$.","marker":"[61]"},{"why":"Defines Coherent WaveBurst, the minimally modelled search and reconstruction pipeline whose waveform outputs the overlap test compares.","marker":"[92]"},{"why":"Documents the cWB reconstruction framework used to produce the point-estimate waveforms in the consistency analysis.","marker":"[99]"},{"why":"Introduces the on-source versus null overlap consistency test on GW190521 that this paper adapts into the $\\Delta_{\\rm median}$ figure of merit.","marker":"[10]"},{"why":"Defines the overlap equation and the waveform-consistency methodology used to compute null and on-source distributions.","marker":"[107]"},{"why":"Provides the quasicircular IMRPhenomXAS model used for parameter estimation, whose biased posteriors generate the PE-sample injections.","marker":"[123]"},{"why":"Provides the gw-eccentricity definition used to assign $e_{20}$ values to the hybrid injections.","marker":"[122]"},{"why":"Supplies the matching procedure used to construct the post-Newtonian and numerical-relativity hybrids by requiring more than 99% agreement in the matching region.","marker":"[113]"}],"fun_headline_variants":["Eccentric mergers betray themselves via waveform mismatch","No eccentric templates? Reconstruction mismatch still catches them","Waveform overlap gap flags orbital eccentricity promptly","Unmodelled reconstructions reveal eccentricity without templates","Low-latency eccentricity flag from waveform reconstruction"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the mismatch between on-source and null overlap distributions is caused by eccentricity, not by cWB reconstruction accuracy varying with the signal parameters that circular parameter estimation recovers.","fun_headline_variants_meta":{"raw":{"variants":["Eccentric mergers betray themselves via waveform mismatch","No eccentric templates? Reconstruction mismatch still catches them","Waveform overlap gap flags orbital eccentricity promptly","Unmodelled reconstructions reveal eccentricity without templates","Low-latency eccentricity flag from waveform reconstruction"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000302,"raw_usage":{"total_tokens":1687,"prompt_tokens":839,"completion_tokens":848,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":455,"completion_tokens_details":{"reasoning_tokens":775}},"tokens_in":455,"tokens_out":848,"duration_ms":10269,"temperature":1.0,"reasoning_tokens":775,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:36:51.621469+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same cWB overlap analysis on circular injections tuned to the biased masses, mass ratios, and SNRs recovered from the eccentric injections; if those circular injections also push $\\Delta_{\\rm median}$ above 4%, the metric is tracking reconstruction fidelity rather than eccentricity.","supporting_citations":[],"review_version":1}