{"id":"b7b2cac6-f410-4556-a4c4-1dceada4a15b","arxiv_id":"2608.07662","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Multi-pipeline follow-up recovers the G347.3-0.5 outlier in O3 data but finds no standard persistent continuous-wave signal in O4a or O4b, disfavouring the astrophysical interpretation under the assumed model.","lead":"A coordinated multi-pipeline follow-up confirms a candidate continuous gravitational-wave signal in O3 LIGO data but finds no trace of it in the newer O4a and O4b data. The study is a practical test of how to vet future detection candidates before claiming the first continuous gravitational-wave discovery.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The O4 null result is strong only within the standard CW model; the paper's own Section 5.2 caveat that large glitches or post-glitch recovery could hide an astrophysical source marks the true boundary of the claim.","rationale":"The reader's weakest_assumption identifies exactly the same load-bearing concern: the conclusion that no standard CW signal is present in O4 presupposes that the searched parameter spaces and signal model cover any real source behavior. The paper itself flags this limitation in Section 5.2, and Section 6 repeats the restriction to the adopted assumptions. This makes the concern real but already acknowledged, so the appropriate verdict remains CONDITIONAL rather than a more severe rejection. The paper is technically careful: three independent pipelines recover the O3 outlier, the O4 searches are shown to be sensitive to injected standard-CW signals at the candidate amplitude, and data-quality checks find no instrumental or correlated-noise explanation. The absence of a reproducibility package and the local-only p-values are secondary issues, but they reinforce the conditional status. The proposed injection campaign directly tests whether the O4 null result would still hold under plausible non-standard but astrophysically motivated source evolution; if the pipelines recover most such injections, the concern is resolved, whereas if they do not, the central claim's scope must be narrowed further. No change to the reader's verdict is needed because the reader already conditioned on this assumption and cited the same Section 5.2 caveat.","tokens_in":23720,"tokens_out":4707,"duration_ms":51944,"concrete_test":"Inject simulated signals matching the O3 candidate parameters into O4a and O4b data, adding a glitch at epochs drawn between O3 and O4b with fractional frequency jumps in [1e-9, 1e-6] and exponential post-glitch recovery timescales spanning days to years, following Antonopoulou et al. 2022. Run the same CWInPy, Weave, and PyFstat O4 analyses used in Section 5 and record the fraction of injections recovered above the loudest-template thresholds reported there. If a substantial fraction of astrophysically plausible glitch/recovery cases is missed, the O4 absence cannot be used to exclude a real neutron-star source, and the conclusions should state that limitation explicitly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central negative claim — no standard CW signal persists across O3 and O4 — rests on the assumption that the O4 searches would have found a real source if one were present. The paper acknowledges in Section 5.2 that a sufficiently large glitch, or a prolonged post-glitch recovery phase with exponential relaxation, would reduce the recovered 2F and allow an astrophysical signal to remain undetected. The enlarged O4 parameter ranges are substantial but finite: Weave covers Delta f0 = ±1e-4 Hz, Delta fdot = ±1.5e-12 Hz/s, Delta fddot = ±2e-20 Hz/s^2, and PyFstat includes a third spin-down derivative with sigma = 4e-29 Hz/s^3, while CWInPy allows glitch phase jumps only in the O3+O4a analysis. None of the three pipelines searches a post-glitch exponential recovery model, and the chosen ranges are motivated by O3 posterior widths and computational cost rather than by a physical prior over plausible glitch sizes and recovery timescales for a roughly 1.6 kyr old neutron star. Thus the O4 null result has no constraining power for that class of signals. Because the conclusion is explicitly restricted to 'under the assumptions adopted in this analysis' (Section 6), this is not an internal inconsistency; it is the load-bearing boundary of the paper's main claim, and it can be over-read as excluding a real astrophysical source.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a multi-pipeline follow-up of the continuous-gravitational-wave outlier identified by the Einstein@Home directed search toward the supernova remnant G347.3-0.5. The authors analyze O3, O4a, and the newly released O4b data from the LIGO H1 and L1 detectors, after extensive data-quality investigations that include spectral artifact identification, auxiliary-channel coherence checks with Fscan and STAMP-PEM, and inter-detector coherence studies with Stochmon. Three independent analysis pipelines are used: CWInPy (time-domain Bayesian), Weave (grid-based F-statistic), and PyFstat (stochastic F-statistic). In O3 data, all three pipelines recover the outlier with parameters consistent with the original report and with local false-alarm probabilities of order 10^-8 to 10^-6. In O4a and O4b, the maximum recovered 2F lies in the range 31-50, far below the value of about 222 predicted from the O3 parameters, and the observed template statistics are consistent with Gaussian noise; injection tests confirm that the pipelines would have recovered a signal of the O3 amplitude. The paper concludes that the outlier is not consistent with a long-lived, phase-coherent continuous wave under the standard signal model, while explicitly leaving open non-standard scenarios such as large glitches or post-glitch recovery.","tokens_in":23958,"tokens_out":13300,"duration_ms":126169,"significance":"This is a useful and well-executed case study for the continuous-wave follow-up workflow. Its strengths are the three independent pipelines with consistent conclusions, the injection tests that calibrate the O4 null result, the off-source noise distributions used for false-alarm estimation, and the explicit falsifiable prediction of the expected O4 detection statistic. The O4 null result is meaningful because the expected statistic is much larger than what is observed. The paper is appropriately cautious about restricting its conclusion to the standard continuous-wave model, but the boundary of the claim (no coverage for large glitches or post-glitch exponential recovery) is stated only in Section 5.2 and should be carried into the abstract. If the results hold, the paper provides a useful template for adjudicating future continuous-wave candidates.","major_comments":[{"comment":"The central negative claim is explicitly conditional on the standard continuous-wave signal model and on the searched parameter ranges, but the abstract and title do not carry that qualifier. The O4 searches cover finite ranges: Weave uses Delta f0 = ±1e-4 Hz, Delta fdot = ±1.5e-12 Hz/s, and Delta fddot = ±2e-20 Hz/s^2 (Section 5.2); PyFstat's third-derivative prior has sigma = 4e-29 Hz/s^3 (Section 5.3); and the CWInPy glitch analysis is limited to O3+O4a and models phase jumps rather than a post-glitch exponential recovery (Section 5.1 and Appendix A). As the authors themselves state in Section 5.2, a sufficiently large glitch or a prolonged post-glitch recovery could keep an astrophysical signal undetected. Because the published version will be read through its abstract, I request that the abstract and the first paragraph of Section 6 state explicitly that the null result applies to the standard, phase-coherent continuous-wave model and does not exclude large-glitch or post-glitch-recovery scenarios.","section":"Abstract and Section 6 (with Section 5.2)"}],"minor_comments":[{"comment":"The local false-alarm probabilities quoted for O3 (Weave ~1e-8, PyFstat ~6e-6) are not corrected for the multiple pipelines, prior configurations, or earlier search stages; the paper states this, but the abstract phrase \"robustly recover the outlier\" could be read as implying a global significance, so I suggest using wording such as \"recover the outlier with parameters consistent with those of the original search\" instead.","section":"Section 4.4"},{"comment":"The statement that the local pfa for each O4 search is \"approximately unity\" is correct only after accounting for the very large number of templates searched; I recommend adding the template count or a one-sentence explanation so that the reader does not infer that a 2F value near 46 is itself a small statistic.","section":"Section 5.2"},{"comment":"The y-axis labels in the text version contain placeholder boxes for the power-of-ten exponents; please check the final rendering of the superscripts.","section":"Figure 2"},{"comment":"The predicted O4ab 2F value of about 222 is mentioned only in Section 5.4; reporting it alongside the observed maxima in Section 5.3 would make the falsification more immediate for the reader.","section":"Sections 5.3 and 5.4"},{"comment":"The reported Bayes factor of 3.9 depends on the reweighting lower bound of 10^-33; since the flat-prior value is 0.11, a brief statement on the sensitivity of the Bayes factors to this bound would help the reader judge the strength of the evidence.","section":"Section 4.1"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is sound and within the journal's scope. The central negative result is supported by consistent multi-pipeline evidence, and the limitations are acknowledged in the text. My only substantive request is to carry the standard-model qualifier into the abstract and conclusions; the remaining points are presentational. I do not see grounds for further technical review beyond the minor items listed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a careful, well-scoped follow-up of a low-significance CW outlier, and the new material is the coordinated three-pipeline comparison plus the first look at O4b. I read it as a solid negative result within the standard CW model.\n\nThe paper does several things right. The O3 recovery across CWInPy, Weave, and PyFstat is consistent; the O4a/O4b results fall far short of the predicted 2F (~222 expected vs observed ~32–50), the off-source noise fits and injection tests are used properly, and the data-quality checks are genuinely extended. The authors also flag their own local p-values and Bayes factors clearly, without overclaiming: 'robust recovery' is parameter recovery, and the p_fa values do not include global trials factors.\n\nWhere are the soft spots? The main one is the boundary of the claim, which the authors actually state in Sec. 5.2: a sufficiently large glitch or a prolonged post-glitch recovery could hide an astrophysical signal. The O4 search ranges are wide but finite, and the CWInPy glitch test only covers O3+O4a with phase jumps, not an exponential relaxation. So the conclusion 'no standard, phase-coherent CW persisting across O3 and O4' is well supported; 'this source is not a neutron star' is not. That is a scope limitation the paper is explicit about, not a hidden flaw.\n\nTwo smaller issues. No code or data are shipped—everything is 'upon reasonable request'—which makes independent verification harder. And the braking-index estimate in Sec. 2.3 (n~19) is presented as a consistency check, but it is hard to take seriously for an outlier at p_fa~10%; it does not affect the main analysis, but it reads as over-interpretation.\n\nOverall: the central negative result holds up, and the multi-pipeline workflow is genuinely useful for future follow-ups. I would send it to a serious referee, and I would cite it for the O4b constraint and the methodology, even though I expect the final 'no detection' conclusion.","headline":"A solid, well-scoped multi-pipeline follow-up that confirms the O3 outlier and finds no O4 signal; the negative claim is strong within the standard CW model, and the paper is honest about its limits.","tokens_in":24731,"tokens_out":2041,"would_cite":true,"duration_ms":20320,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["04.30.-w","04.80.Nn","95.55.Ym"],"model":"deepseek-v4-flash","headline":"Three independent pipelines confirm a gravitational-wave outlier in O3 data and find no trace of it in O4, disfavouring a long-lived signal.","keywords":["continuous gravitational waves","gravitational-wave outlier","follow-up search","F-statistic","Bayesian inference","LIGO O4 data","supernova remnant G347.3-0.5","signal versus noise"],"falsifier":"Whether the negative conclusion is right can be settled by the next data release: if a standard phase-coherent signal with the O3 parameters exists, the fully coherent F-statistic in O4 should grow to the predicted values of about 91 for O4a and about 222 for O4ab and appear as a clear excess over the Gaussian-noise template distribution. Repeating the grid or stochastic searches on O4c data, expected in December 2026, and finding either no excess or an excess at the wrong parameters would confirm the paper's verdict; finding the predicted excess would refute it.","tokens_in":23441,"feed_emoji":"📡","tokens_out":6367,"duration_ms":59760,"temperature":0.7,"pith_summary":"This paper tests whether a persistent continuous gravitational-wave (CW) candidate reported toward the young supernova remnant G347.3-0.5 is a real signal or a noise artifact. Three independent search pipelines, using different detection statistics and sampling methods, all consistently recover the candidate in the third observing run (O3) of the LIGO detectors, with parameters matching the original report. When the search is extended to the first and second parts of the fourth observing run (O4a and O4b), none of the three pipelines finds any evidence of a phase-coherent signal at the extrapolated parameters; the loudest statistics are fully consistent with Gaussian noise. The paper argues that this absence of a clear O4 recovery disfavours a long-lived, phase-coherent CW source under the standard signal model, while leaving room for more exotic scenarios such as a large glitch or timing noise. The value of the study is procedural: it is a test case for how the field can independently corroborate or refute CW candidates before a first detection is claimed.","feed_headline":"Outlier gravitational-wave candidate vanishes in newer data","feed_subtitle":"Multi-pipeline follow-up finds the G347.3 candidate indistinguishable from noise in O4.","key_machinery":"The load-bearing object is the standard continuous-wave signal model: a slow frequency evolution f(t)=f0+fdot0(t-tref)+... truncated at second order, combined with the detector-response model that folds in Doppler and relativistic phase shifts. Against this model, the paper deploys three matched-filtering engines that share the same signal hypothesis but differ in inference: a time-domain Bayesian heterodyne with nested sampling, a grid-based fully coherent F-statistic search using a metric-template lattice, and a stochastic-sampling F-statistic search. The F-statistic analytically maximises over the four amplitude parameters, leaving a search over frequency-evolution parameters, and the semi-coherent averaged statistic 2Fhat is used for combined-run searches. The consistent absence of any high-statistic excess in O4, compared with injection-controlled expectations, is what carries the negative conclusion.","core_discovery":"Grounded in the standard CW signal model, a Taylor expansion of the signal frequency with up to second-order spin-down and Doppler/relativistic phase corrections, the paper finds that the G347.3-0.5 outlier is a consistent feature of O3 data but does not persist into O4a or O4b. In O3, the time-domain Bayesian analysis returns a log10 Bayes factor of 3.9 in favour of a coherent signal over noise, the grid-based F-statistic search recovers 2Fmax=74.6 with local pfa approximately 1e-8, and the stochastic F-statistic analysis recovers 2F approximately 78 with local pfa approximately 6e-6; all agree with the original candidate parameters, including a small sky offset reproduced by injection. In O4a and O4b, the same pipelines find loudest statistics (2F in the range of roughly 31-50, and 2Fhat approximately 33 for the semi-coherent search) that track the Gaussian-noise expectation instead of the predicted values of 2F approximately 91 for O4a and approximately 222 for O4ab, and Bayesian evidence shifts to disfavour the signal, with log10 Bayes factors around -0.2 to -0.5. The conclusion is that the candidate is not consistent with a standard CW signal persisting across the full available data, though no known instrumental artifact or correlated noise can be blamed either.","pith_inferences":["The pattern of a strong O3 recovery and an O4 absence is exactly what a transient noise event, or a signal whose phase changed between runs, would look like; the paper's glitch tests are not conclusive but point in that direction.","If the candidate is eventually classed as noise, search pipelines can still produce false-alarm survivors at the roughly ten-percent level, and only multi-run disfavouring can retire them; this argues for building follow-up protocols around later observing runs.","A testable extension is to hunt specifically for a glitched version of this signal in O4c data, using glitch-epoch priors informed by the paper's O3+O4a posteriors rather than only a standard signal model.","The candidate's implied braking index of about 19 is far outside standard expectations of n=3 or n=5, which independently weakens the astrophysical interpretation of the O3 outlier. "],"forward_implications":["If the paper is right, the G347.3-0.5 outlier should not be treated as a candidate CW source: no long-lived phase-coherent signal exists at those parameters in the available data.","CW candidates should require consistent recovery in later observing runs, not only in the discovery run, before being considered astrophysical.","Multi-pipeline cross-checks, combining Bayesian time-domain inference, grid-based F-statistics, and stochastic F-statistics, can produce a convergent verdict on a single candidate.","No known instrumental line or correlated detector noise explains the O3 outlier, so its origin remains unidentified.","For a genuine persistent signal, adding data should raise the detection statistic and evidence; the observed decrease is itself a discriminating test. "],"supporting_citations":[{"why":"The original directed search that identified the outlier and supplied the O3 parameters and the initial O4a follow-up result.","marker":"[11]"},{"why":"The time-domain Bayesian pipeline (CWInPy) used for coherent single- and joint-detector analyses.","marker":"[19]"},{"why":"The grid-based F-statistic pipeline (Weave) used for fully coherent and semi-coherent searches over O3, O4a, and O4ab.","marker":"[20]"},{"why":"The stochastic-sampling F-statistic pipeline (PyFstat) used for posterior estimation and injection-based significance tests.","marker":"[22]"},{"why":"The matched-filter F-statistic formalism on which the grid and stochastic searches are built.","marker":"[26]"},{"why":"The heterodyne time-domain Bayesian method underlying the CWInPy analysis.","marker":"[72]"},{"why":"The data-quality tools used to rule out instrumental artifacts and correlated noise around the candidate frequency.","marker":"[61]"},{"why":"The glitch and post-glitch relaxation scenario used to motivate tests beyond the standard CW model.","marker":"[24]"}],"fun_headline_variants":["Gravitational wave candidate fades in new data","O3 signal not confirmed in O4: CW candidate likely noise","Einstein@Home outlier: no persistence beyond O3","Supernova remnant signal? Not in newer LIGO data","G347 candidate: a flash in the pan for CW searches"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion rests on the assumption that a real signal from the same source would follow the standard CW model and lie within the enlarged parameter ranges searched in O4; a sufficiently large glitch or prolonged post-glitch recovery could make a genuine astrophysical signal undetectable in these follow-ups.","fun_headline_variants_meta":{"raw":{"variants":["Gravitational wave candidate fades in new data","O3 signal not confirmed in O4: CW candidate likely noise","Einstein@Home outlier: no persistence beyond O3","Supernova remnant signal? Not in newer LIGO data","G347 candidate: a flash in the pan for CW searches"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000257,"raw_usage":{"total_tokens":1625,"prompt_tokens":1041,"completion_tokens":584,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":657,"completion_tokens_details":{"reasoning_tokens":500}},"tokens_in":657,"tokens_out":584,"duration_ms":5842,"temperature":1.0,"reasoning_tokens":500,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T00:25:47.104002+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Whether the negative conclusion is right can be settled by the next data release: if a standard phase-coherent signal with the O3 parameters exists, the fully coherent F-statistic in O4 should grow to the predicted values of about 91 for O4a and about 222 for O4ab and appear as a clear excess over the Gaussian-noise template distribution. Repeating the grid or stochastic searches on O4c data, expected in December 2026, and finding either no excess or an excess at the wrong parameters would confirm the paper's verdict; finding the predicted excess would refute it.","supporting_citations":[{"cited_title":"Implementing a semicoherent search for continuous gravitational waves using optimally-constructed template banks","cited_arxiv_id":"1804.03392","evidence_quote":"The grid-based F-statistic pipeline (Weave) used for fully coherent and semi-coherent searches over O3, O4a, and O4ab."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The matched-filter F-statistic formalism on which the grid and stochastic searches are built."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The heterodyne time-domain Bayesian method underlying the CWInPy analysis."}],"review_version":1}