{"id":"de8f673d-74de-4777-9f58-8354fc8cf4fa","arxiv_id":"2507.12161","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A new low-frequency atlas of continuous gravitational wave upper limits is released, with worst-case strain limits below 1e-25 in the most sensitive region.","lead":"The authors release a public atlas of upper limits on continuous gravitational waves from LIGO's third observing run, covering 20 to 200 Hz. The atlas includes the most sensitive all-sky low-frequency results to date, with worst-case strain limits below 1e-25 in the best bands, plus a new sensitivity proxy.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The per-point frequentist upper limits are credible, but the headline population-average sensitivity curve rests on a lock-level proxy validated only at 100-200 Hz and known to fail in one contaminated band; until validated across the full band, the 'most sensitive all-sky search' claim exceeds…","rationale":"I read the paper as a data release whose primary deliverable is the public atlas of per-point upper limits. The direct upper-limit machinery appears credible: the universal-statistics approach is published, the strict frequentist design is stated, and the three hardware injections that fall inside the searched parameter space all satisfy UL/h0 greater than 100%, providing an end-to-end check. I therefore do not see a decisive flaw in the per-point upper limits themselves. The vulnerability is in how the paper and the reader convert these per-point results into a population-average sensitivity claim. The all-sky population-average curve is not a standard injection-based upper limit; it is a 95% quantile of location-specific lock-level proxies, labelled as 90% confidence. Its validation is limited to 100-200 Hz test signals and a 100-101 Hz spot-check, and the paper itself says that full test-signal simulations would be required to establish the confidence in each band. Since Figure 1 uses this curve for pipeline comparisons and Section V concludes 'most sensitive all-sky search to date', the headline sensitivity advantage over LVK/Einstein@Home results is exactly as strong as the proxy, not as strong as the per-point upper limits. The concrete test I propose extends the Table II spot-check across the full band. If uncontaminated bands at 20-100 Hz recover at or above 90%, the proxy concern is resolved and the claims can stand. If not, the population-average comparison should be labelled provisional or removed until proper injection-based estimates are produced. Because this matches the reader's stated rationale for a conditional verdict, I do not recommend changing the verdict.","tokens_in":12093,"tokens_out":12977,"duration_ms":163607,"concrete_test":"Run the Table II spot-check procedure on representative uncontaminated bands spanning 20-100 Hz (for example 20.003-20.054, 40.003-40.054, 60.003-60.054, and 80.003-80.054 Hz) and at 150-160 Hz, using 200 simulated signals per band at the published population-average proxy strain. If the frequency-plus-polarization lock rate falls below 90% in any uncontaminated band, the proxy curve cannot support the population-average sensitivity comparison in Figure 1; if all bands meet 90%, the population-average sensitivity claim is substantially strengthened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The per-entry frequentist upper limits are likely sound: the universal-statistics algorithm is published, and the three hardware injections that fall inside the searched parameter space (Table I rows 3, 5, and 11) all have UL/h0 greater than 100%, providing an end-to-end check. The load-bearing weak point is the simultaneous claim that this atlas is the most sensitive all-sky search in 20-200 Hz. That comparison is made through the new population-average curve ('Falcon pop avg' in Figure 1), which is derived from location-specific lock-level proxies rather than from direct upper-limit measurements. The paper validates the proxy only with test signals in the 100-200 Hz band (Figure 2), and the only published spot-check of the resulting all-sky population-average estimate (Table II) covers just 100-101 Hz, where one band (100.003-100.054 Hz) fails to reach the 90% lock rate because of a detector artifact. The paper itself concedes that full test-signal simulations would be required to attach confidence to the proxy in each band. If the proxy underestimates the true level needed for parameter recovery at 20-100 Hz or in other artifact-contaminated bands, the population-average curve in Figure 1 overstates sensitivity, and the 'most sensitive to date' conclusion is not supported. This does not invalidate the per-point upper limits, but it weakens the headline sensitivity comparison made in Section V.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper releases the low-frequency atlas of continuous gravitational waves covering 20-200 Hz, frequency derivatives |f1| <= 5e-11 Hz/s and |f2| <= 1e-20 Hz/s^2, built from the full LIGO O3 data with two-day coherence lengths. For each 50 mHz frequency band and sky-position Voronoi cell, the atlas provides frequentist 95% confidence upper limits on strain, computed with the universal statistics algorithm, together with the SNR peaks and their recovered parameters. The paper also introduces a location-specific polarization-average lock-level proxy and uses its 95% quantile over the sky to define an all-sky population-average sensitivity curve. The atlas data and example R scripts are publicly released.","tokens_in":12347,"tokens_out":5925,"duration_ms":61972,"significance":"The per-point frequentist upper limits are credible: they use the established universal statistics method, and the three hardware injections inside the searched parameter space (Table I rows 3, 5, and 11) all satisfy UL/h0 > 100%, providing an end-to-end validation of the upper-limit pipeline. The public data release with example scripts is a valuable community resource. The population-average lock-level proxy is a practically useful and clearly motivated concept, but it is not a direct upper limit and its validation is incomplete. The headline claim of being the most sensitive all-sky search to date rests on this proxy, so the result is significant if the proxy is reliable across the full band, but the current evidence only directly supports it for part of the band.","major_comments":[{"comment":"The conclusion that this is 'the most sensitive all-sky search to date' rests on the Falcon population-average proxy curve in Figure 1, not on direct upper-limit measurements. The proxy is validated only with test signals in the 100-200 Hz band (Figure 2) and spot-checked only in the 100-101 Hz band (Table II), where one band (100.003-100.054 Hz) fails to reach the 90% lock rate because of detector artifacts. The paper itself states that 'computationally expensive test-signal recovery simulations would be required to determine the actual confidence associated with the values in each band.' Without such validation across 20-100 Hz and in other artifact-contaminated bands, the wide-band comparison with other pipelines may overstate sensitivity. Please either validate the proxy over the full band or explicitly qualify the 'most sensitive' claim as provisional and restricted to the validated range.","section":"Section V and Figure 1"},{"comment":"The population-average proxy is defined as the strain at which 95% of signals have correctly recovered parameters at a given sky location, and the all-sky curve is the 95% quantile of these location-specific proxies, described as a 90% confidence level. This is a different statistical quantity from the frequentist 95% confidence upper limits reported by the other pipelines (e.g., refs. [15-18]) with which it is compared in Figure 1. Directly overlaying these curves and using them to claim superiority conflates a proxy for 'lock-on' with a rigorous upper limit. The text should state explicitly what this comparison can and cannot establish, and should avoid using the phrase 'most sensitive' as if it were established by an apples-to-apples measurement.","section":"Section IV and Figure 1"}],"minor_comments":[{"comment":"The caption states that the frequency match criterion is 0.2 mHz, while Section IV and Table I use 0.5 mHz as the expected frequency offset tolerance. Please align these tolerances or explain the difference.","section":"Figure 2"},{"comment":"The claim that higher-order frequency derivatives are covered by the 'intrinsic robustness' of a 2-day coherence search is asserted without supporting tests or citations. Since this is an assumption about search sensitivity, a brief justification or reference would help.","section":"Section II"},{"comment":"Reference [12] contains the placeholder text 'number will be inserted by publisher'; this must be resolved before final publication.","section":"References"},{"comment":"The phrase 'most sensitive all-sky search to date' appears without the caveat, stated in Section IV, that this comparison is based on the population-average proxy rather than on direct upper limits. Adding that qualifier in the abstract and conclusion would prevent misreading.","section":"Abstract and Section V"},{"comment":"The caption says 'The upper limits reflect the sensitivity of the search,' but the Falcon pop avg curve is not an upper limit; it is a proxy. The caption should distinguish upper-limit curves from proxy curves to avoid ambiguity.","section":"Figure 1 caption"}],"recommendation":"major_revision","confidential_remarks":"The per-point upper limits appear sound and the data release is valuable. The main issue is the gap between the headline 'most sensitive' claim and the current validation of the population-average proxy. I would advise the editor that a revision that either supplies full-band test-signal validation or carefully qualifies the claim would likely make the paper publishable. I see no reason to doubt the core upper-limit results."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a data release paper, and judged as that, it's a good one. The per-point 95% frequentist upper limits on continuous-wave strain from 20-200 Hz are derived with the universal statistics method, which has a published track record, and the hardware injections in Table I provide the right end-to-end check: every in-range injection has UL/h0 > 100%. The public atlas itself, with SNR peaks and per-sky-point coefficients, looks like a genuine community resource. The improved sky resolution (6x over the predecessor) and the sub-1e-25 worst-case limits in the most sensitive band are plausible and, as far as I can tell, correctly computed.\n\nThe genuinely new element is the location-specific polarization-average lock-level proxy, and the idea of using it to construct a population-average sensitivity curve for comparison with other pipelines. That is clever and useful, but it is the soft spot. The proxy is validated only on 100-200 Hz injection studies, and the spot-check in Table II covers just one band, 100-101 Hz, where one sub-band fails the 90% lock criterion due to a detector artifact. The paper is honest about this: it states that full test-signal simulations would be needed to attach confidence to the proxy in each band. So the conclusion's phrase 'most sensitive all-sky search to date' is a bit ahead of the evidence. It may well be true, but the evidence right now supports the per-point upper limits, not the population-average comparison curve.\n\nThe citation pattern is standard and appropriate; the authors cite their own prior atlas papers and the relevant LIGO/Virgo and Einstein@Home searches. No issue there.\n\nWho should read this: anyone planning a follow-up search in 20-200 Hz, or anyone who wants to compare sensitivity methods across pipelines. The paper deserves a serious referee; the proxy validation will presumably be tightened in the final analysis, but the release itself is valuable now. My verdict: accept the per-point atlas as sound, and treat the population-average curve as provisional until the full validation appears. I'd send it to review, and I'd cite it.","headline":"Solid per-point upper limits and a genuinely useful data release; the flashier population-average comparison rests on a proxy that the paper itself tells you is not fully validated.","tokens_in":12870,"tokens_out":2141,"would_cite":true,"duration_ms":22797,"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":"A public atlas now sets the tightest all-sky upper limits on 20–200 Hz continuous gravitational waves, reaching worst-case strain below $10^{-25}$.","keywords":["continuous gravitational waves","all-sky search","upper limits","atlas","neutron stars","O3 data","spin-down","lock-level proxy"],"falsifier":"For a frequency band where the worst-case upper limit is claimed below $10^{-25}$, inject simulated signals with strain $h_0$ above that limit, random sky positions, and worst-case linear polarization, re-run the search on the same O3 data, and count how many produce an SNR peak with the injected frequency recovered within 0.5 mHz; if fewer than 95% do, the frequentist coverage claim fails.","tokens_in":11870,"feed_emoji":"🔭","tokens_out":9293,"duration_ms":97129,"temperature":0.7,"pith_summary":"This paper releases a public atlas of upper limits on continuous gravitational-wave strain for the 20–200 Hz band, computed from the full third observing run of the LIGO detectors. It claims that, for every sky position and every 50 mHz frequency band, the atlas supplies strict frequentist 95% confidence upper limits, with the worst-case all-sky limits falling below $10^{-25}$ in the most sensitive region for the first time. The release also introduces a location-specific \"lock-level\" proxy that tells users the signal strength needed for the search to correctly report a signal's frequency and polarization at a given sky position. The data is made public before the completion of the full follow-up analysis, so it can be used to rank candidates and plan follow-up searches.","feed_headline":"All-sky gravitational-wave atlas sets strict limits below 1e-25","feed_subtitle":"Public 20-200 Hz maps give 95% confidence upper limits at every sky position, guiding follow-up searches.","key_machinery":"The central object is the atlas itself: a partition of the search space into 50 mHz frequency bands and Voronoi cells on the celestial sphere, each cell carrying an upper limit on $h_0$ and an SNR peak. The upper limits are computed with the universal statistics algorithm, which gives strict frequentist coverage without assuming a particular noise distribution. The new \"lock-level\" proxy is computed from a background-noise estimate for each cell and represents the signal strain required for the search to correctly report frequency and polarization for 95% of signals at that sky position. The proxy is deliberately insensitive to signals in the band, so it can be used to decide whether a parameter-space follow-up of an SNR peak is worthwhile.","core_discovery":"The paper's central claim is that it has carried out the most sensitive all-sky search for continuous gravitational waves in the 20–200 Hz band to date, using the full O3 dataset and a 2-day coherence length, and that the results are released as a public atlas. For each 50 mHz frequency band and each sky-position Voronoi cell, the atlas gives strict frequentist 95% confidence upper limits on the strain amplitude $h_0$, together with the SNR peak's frequency and polarization. The worst-case all-sky upper limits now drop below $10^{-25}$ at the most sensitive frequencies, about 30% better than the previous release, while the sky grid is six times finer. The atlas also defines a location-specific \"lock-level\" proxy: the strain at which 95% of signals from that sky location would have their frequency and polarization correctly reported in the atlas's SNR peak. Taking the 95% quantile of these proxies over the sky gives a population-average upper-limit estimate at 90% confidence, intended for comparison with other pipelines and not as a formal upper limit.","pith_inferences":["If the atlas's coverage extends to boson-cloud and wide-binary sources as the authors suggest, the same released upper limits could be reinterpreted as constraints on those exotic source classes without rerunning the search.","A natural next step would be to use the sky-position-dependent polarization coefficients to produce upper limits for arbitrary source orientations on demand, turning the static atlas into a queryable sensitivity service.","The 100 Hz band's artifact contamination, already visible in the spot-check, implies that per-cell lock-level maps could serve as a detector-health diagnostic, flagging bands where instrumental lines compromise parameter recovery.","The early-release workflow itself is testable: if a loud candidate found in the released SNR peaks is later confirmed by the follow-up analysis, that validates the practice of using the pre-follow-up atlas for prioritisation."],"forward_implications":["The public atlas lets any search team directly read off, for any sky position and 50 mHz band, the strain above which a continuous gravitational wave would have been loud enough to appear in the SNR peaks.","The sub-$10^{-25}$ worst-case limits in the most sensitive band are the tightest all-sky bounds reported there, so any future claim of a signal in this band must face this constraint.","The lock-level proxy gives a per-cell map of follow-up worthiness, allowing limited computational resources to be spent where parameter recovery is reliable rather than on noise-dominated cells.","Because the data are released before the full follow-up is finished, candidate follow-up and population studies can proceed in parallel with the final analysis.","The population-average proxy provides a common basis for sensitivity comparisons with other pipelines, while the formal upper limits remain the correct frequentist quantities."],"supporting_citations":[{"why":"Defines the earlier atlas release that this work extends and supplies the sensitivity baseline for comparison.","marker":"[1]"},{"why":"Provides the O3a dataset used in the search.","marker":"[4]"},{"why":"Provides the O3b dataset used in the search.","marker":"[5]"},{"why":"Describes the Advanced LIGO detectors whose data the search uses.","marker":"[6]"},{"why":"Supplies the method for deriving upper limits for arbitrary polarizations from sky-position-dependent coefficients.","marker":"[13]"},{"why":"Provides the universal statistics algorithm used to compute strict frequentist upper limits.","marker":"[14]"},{"why":"Einstein@Home bucket search result used as a comparison baseline for the sensitivity claims.","marker":"[16]"},{"why":"LIGO/Virgo/KAGRA all-sky search result used as a comparison baseline for the sensitivity claims.","marker":"[18]"}],"fun_headline_variants":["Most sensitive CW atlas to date: limits below 1e-25","All-sky GW atlas tops sensitivity, worst-case limits <1e-25","Public atlas tightens continuous-wave upper limits to <1e-25","20-200 Hz atlas: strict limits, 30% better than prior release","CW atlas release: sky-grid 6x finer, limits below 1e-25"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The search assumes every plausible continuous-wave source's frequency evolution over a year is described by $|f_1|\\le 5\\times10^{-11}$ Hz/s and $|f_2|\\le 10^{-20}$ Hz/s$^2$, and that a 2-day coherence length safely absorbs any higher-order frequency derivatives; a source with stronger spin-down would lie outside the atlas's coverage.","fun_headline_variants_meta":{"raw":{"variants":["Most sensitive CW atlas to date: limits below 1e-25","All-sky GW atlas tops sensitivity, worst-case limits <1e-25","Public atlas tightens continuous-wave upper limits to <1e-25","20-200 Hz atlas: strict limits, 30% better than prior release","CW atlas release: sky-grid 6x finer, limits below 1e-25"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000253,"raw_usage":{"total_tokens":1513,"prompt_tokens":845,"completion_tokens":668,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":461,"completion_tokens_details":{"reasoning_tokens":564}},"tokens_in":461,"tokens_out":668,"duration_ms":6330,"temperature":1.0,"reasoning_tokens":564,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:51:28.156635+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a frequency band where the worst-case upper limit is claimed below $10^{-25}$, inject simulated signals with strain $h_0$ above that limit, random sky positions, and worst-case linear polarization, re-run the search on the same O3 data, and count how many produce an SNR peak with the injected frequency recovered within 0.5 mHz; if fewer than 95% do, the frequentist coverage claim fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the earlier atlas release that this work extends and supplies the sensitivity baseline for comparison."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the O3a dataset used in the search."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the O3b dataset used in the search."},{"cited_title":"[number will be inserted by publisher] for numerical values of upper limits","cited_arxiv_id":null,"evidence_quote":"Supplies the method for deriving upper limits for arbitrary polarizations from sky-position-dependent coefficients."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the universal statistics algorithm used to compute strict frequentist upper limits."},{"cited_title":"Steltner, M","cited_arxiv_id":null,"evidence_quote":"Einstein@Home bucket search result used as a comparison baseline for the sensitivity claims."}],"review_version":1}