{"id":"498c1224-e618-4119-9361-9b4fb63df045","arxiv_id":"1908.05162","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"No significant steady neutrino emission was found from the directions of 81 IceCube alert events in 10.5 years of archival muon-track data.","lead":"Using 10.5 years of IceCube archival data, researchers searched for steady neutrino emission from the directions of 81 high-energy neutrino alert events and found no excess, with results compatible with background. The work tests whether alert-producing neutrinos mark persistent cosmic neutrino sources, a central question for multi-messenger astronomy.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline p-values are not clearly defined: 0.83 is reported as the 'best' individual p-value without stating whether it is the minimum local p across 81 alerts or a trials-corrected global value, and the paper does not say whether the alert events are removed from the archival search sample.","rationale":"The reader's conditional verdict is appropriate. My stress-test identifies the same region of the argument but puts more weight on the undefined trial structure: the individual p-value is not reproducible from the text. A p-value of 0.83 as the 'best' (most significant) of 81 independent tests is statistically impossible if the local p-values are well calibrated; therefore either the p-value has already been corrected for the 81 trials and the paper omits that correction, or the local p-values are not calibrated. Both possibilities undermine the precise numerical claim, though not necessarily the qualitative conclusion that no steady excess is seen. The self-inclusion question is a separate but related omission; the alert list is data-driven, so the null model must state how the alert-selected events are treated in the archival sample. If those events are included, the null is not pure background; if excluded, that should be explicit. The concrete check I propose would settle both ambiguities with one Monte Carlo run. I do not recommend rejection: the reported p-values are large, and no evidence points toward an actual source; the problem is that the analysis as written does not demonstrate the calibration needed for the central claim. Thus CONDITIONAL remains the right verdict.","tokens_in":3918,"tokens_out":10920,"duration_ms":110545,"concrete_test":"Run background-only pseudo-experiments using the archival through-going muon sample and the actual alert-selection rule: draw 81 alert directions from high-energy background tracks, compute the 81 local per-alert p-values, and record the minimum. Compare the observed 'best p-value of 0.83' to this null distribution. In the same Monte Carlo run, repeat the procedure twice, once with the 81 alert events included in the archival sample and once with them removed. This settles whether 0.83 is a local or post-trial p-value and whether self-inclusion changes the result; if removing the alert events leaves the p-values unchanged, the self-triggering concern is not load-bearing.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central result is the two p-values, 0.83 (individual) and 0.856 (stacking). For the stacking p-value the definition is standard, but for the individual p-value the text is ambiguous. Section 2.3 says 'we determine a p-value for each of the 81 alerts' and then reports 'a best p-value of 0.83'. If 0.83 is the smallest of the 81 local p-values, the null calibration cannot be correct: for 81 independent uniform p-values the minimum has expectation about 0.012 and P(min > 0.83) is vanishingly small. If 0.83 is instead the post-trial p-value of the most significant alert, the paper should state the number of trials and the local p-value, but no such correction is described. The same ambiguity affects the null model. Section 2.1 (Eq. 2.7) assumes ns=0 as background; however, the alert directions are not an external catalog. They were selected from IceCube events, and the text never states whether the triggering alert events are part of the 10.5-year through-going muon sample. If they are included, the null hypothesis for that direction is not pure background and the p-value does not measure an independent steady source excess. If they are excluded, the paper should say so. Because the entire paper is a null result resting on these two p-values, the missing calibration and sample-overlap statements leave the central claim unverified as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper (ICRC2019 proceedings) searches for time-integrated neutrino emission from the directions of 81 IceCube high-energy track-like alert events, using 10.5 years of archival through-going muon data. The analysis uses an unbinned likelihood ratio with spatial and energy terms, performs a 0.1-degree grid scan within the 90% error contour of each alert, and reports a best individual p-value of 0.83 and a stacking p-value of 0.856, concluding that the data are compatible with background. The paper also indicates that a transient search is planned.","tokens_in":4232,"tokens_out":3117,"duration_ms":31349,"significance":"If the analysis were fully specified and correctly calibrated, a null result from archival IceCube data toward the directions of IceCube's own alerts would be a useful consistency check for the alert program and for the possible existence of steady sources in those directions. The paper is clearly written and uses a standard likelihood formalism; the stacking test statistic is explicitly defined in Eq. (2.8). However, as presented, the central null conclusion rests on p-value definitions and a background model that are not sufficiently described to be verified.","major_comments":[{"comment":"The meaning of the reported 'best p-value of 0.83' is ambiguous. If 0.83 is the minimum of the 81 individual p-values, then the value is not trials-corrected; for 81 independent uniform p-values the expected minimum is about 0.012, and the probability that the minimum exceeds 0.83 is vanishingly small under the stated null. If 0.83 is instead a post-trial p-value for the most significant alert, the number of trials and the local p-value are not reported. The manuscript must state which quantity is being quoted and describe the trial-correction procedure explicitly.","section":"Sections 2.2 and 3"},{"comment":"The null hypothesis is defined as background only (ns=0) for each alert direction, but the alert directions are selected from IceCube's own events, and the manuscript never states whether the triggering alert events are included in the 10.5-year archival sample used in the search. If an alert event is present in the data used to test its own direction, the null hypothesis is not pure background and the reported p-values do not measure an independent steady source excess. If the alert events are removed, that exclusion should be stated; if they are not removed, the analysis should account for the self-triggered contribution to the null model.","section":"Section 2.1, Eq. (2.7)"},{"comment":"The p-value calibration for the position-grid scan is not described. The text states that 'based on the expected distribution of the best-fit test statistic we determine a p-value for each of the 81 alerts,' but it does not specify how the scan over 0.1-degree grid points (which may number many per alert) enters the null distribution. Without knowing whether and how the number of scanned positions is incorporated into the background-only test-statistic distribution, the reported p-values cannot be reproduced or checked.","section":"Section 2.2, position scan"},{"comment":"The paper reports only p-values and does not give the best-fit number of signal events, spectral index, or upper limits for the most significant alert or for the stacking result. These quantities are needed to assess the sensitivity of the search and the physical meaning of the null result, especially because the analysis is presented as a search rather than a limit-setting procedure.","section":"Section 3, Results"}],"minor_comments":[{"comment":"The abstract states that the search 'will be automated for all future high energy track-like neutrino alerts,' while the body reports results for 81 historic alerts; please clarify whether the current analysis already includes all alerts up to the stated data cutoff or only a subset, and where the automation is described.","section":"Abstract"},{"comment":"The event labeled 'IceCube-172209A' in the second paragraph appears to be a typo for 'IceCube-170922A' given the accompanying reference to TXS 0506+056; please correct the identifier.","section":"Section 2.2"},{"comment":"The sentence 'Therefore a trial factor of close to two would apply if those results were combined' is cryptic: it is unclear which results are being combined and why the factor is 'close to two.' Please expand this remark or remove it.","section":"Section 2.3"},{"comment":"The two panels of Figure 2 lack visible axes labels and a color scale for the test-statistic values; adding these would make the illustration of the position scan easier to interpret.","section":"Figure 2"},{"comment":"The most significant alert is described only as 'close to the galactic plane,' with no identifier or coordinates; provide the alert name and its fitted right ascension and declination so that the result is reproducible.","section":"Section 3"}],"recommendation":"major_revision","confidential_remarks":"This is a two-page ICRC proceedings paper, and one might be tempted to treat the omitted details as deferred to a future full paper. However, the ambiguities in the p-value definition and in the sample-overlap issue directly affect the validity of the central null claim as stated in this manuscript. The authors should either correct and fully specify the calibration or explicitly frame this as a preliminary methods presentation with the full analysis to appear elsewhere. I would also advise the editor to ask the authors to state whether the alert events are part of the archival data sample, since this is a common source of subtle self-triggering bias in such searches."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this before anything else: the paper reports a null result—no steady neutrino excess from the directions of 81 IceCube alerts in 10.5 years of archival muon data—but the headline p-values are not clearly defined. The individual p-value of 0.83 and the stacking p-value of 0.856 look fine as a backdrop, but the text is ambiguous about what 0.83 actually is. If it is the smallest of the 81 local p-values, then the null calibration cannot be correct: for 81 independent uniform p-values, the minimum should be around 0.012, and a minimum of 0.83 is astronomically unlikely. If it is a trials-corrected global p-value, the correction is not described anywhere. The stress-test note is right on this. The paper also never states whether the triggering alert events are included in the archival sample; if they are, the null hypothesis is not pure background for those directions, and the p-values lose their straightforward interpretation. These are not cosmetic issues for a null result whose entire force rests on those p-values.\n\nWhat is genuinely new here is the target list: using IceCube’s own alerts as a source catalog, rather than an external list of candidate sources. That is a legitimate extension of the standard unbinned likelihood point-source search, and the paper is honest about its preliminary status. The stacking test for a population hypothesis is standard, and the citation pattern looks solid. The plan to automate for future alerts is sensible, though it is not yet a result.\n\nThe soft spots are exactly the p-value definition, the sample overlap, and the absence of upper limits. For a null result, upper limits would let the reader assess how constraining the search really is; without them, the utility is limited. The paper is a short ICRC proceedings, so some brevity is expected, but these omissions are central, not peripheral.\n\nWho gets value from this? Researchers who follow IceCube’s alert program and are curious whether alert directions show long-term excesses. The idea is worth taking seriously, but the result in this draft is not fully verifiable. I would send it to peer review because the underlying analysis is important and the issues are fixable with clarification—not with new physics. Ask the authors to state the number of trials, define the p-values explicitly, and say whether alert events are in the sample. That revision would make this a useful, if quiet, result.","headline":"The null result is plausible but the two headline p-values are not clearly defined; as written, the individual p-value of 0.83 cannot be a simple minimum of 81 uniform p-values, and the paper should clarify the p-value calibration and whether alert events are in the archival sample.","tokens_in":4756,"tokens_out":4468,"would_cite":false,"duration_ms":45511,"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":"Ten and a half years of IceCube archival data show no steady neutrino excess from the directions of the detector's own high-energy alert events.","keywords":["IceCube","neutrino astronomy","point source search","IceCube alerts","unbinned likelihood","stacking analysis","through-going muons","null result"],"falsifier":"Check the archival data set for the 81 alert events and rerun both the individual and stacking searches with those events excluded; if any p-value changes materially, the background-compatible conclusion depends on self-triggering rather than an independent steady excess.","tokens_in":3716,"feed_emoji":"🔭","tokens_out":12349,"duration_ms":109389,"temperature":0.7,"pith_summary":"Every year IceCube broadcasts roughly eight track-like neutrino alerts, high-energy events with good directional accuracy, and other telescopes follow them up looking for counterparts. This paper asks whether those alert directions also hold a steady excess of neutrinos in 10.5 years of archival muon-track data, which would make the alerts tracers of persistent cosmic neutrino sources. In this preliminary time-integrated search the answer is no: the most significant single direction gives $p = 0.83$ and the combined stacking analysis gives $p = 0.856$, both consistent with pure background. The paper concludes that IceCube's alert events as a population show no evidence of continuous neutrino emission in archival data, and it announces a transient flare search as the next step.","feed_headline":"IceCube alert directions show no steady neutrino source","feed_subtitle":"A 10.5-year archival search finds no steady neutrino excess behind IceCube's alerts.","key_machinery":"The engine is an unbinned likelihood-ratio test statistic, $$TS = -2\\log\\left[\\frac{L(n_s=0)}{L(n_s=\\hat{n}_s)}\\right] = 2\\sum_i \\log\\left[\\frac{\\hat{n}_s}{N_{\\mathrm{obs}}}\\left(\\frac{S_i}{B_i}-1\\right)+1\\right],$$ with per-event signal and background densities split into spatial and energy factors. The signal spatial density is a two-dimensional Gaussian around the candidate position scaled by the event's reconstruction uncertainty; background is uniform in right ascension and a measured function of declination, with energy terms taken from Monte Carlo. Because the alert positions themselves are uncertain, the search re-fits the source position on a $0.1^\\circ$ grid inside the 90% error contour and converts the best-fitting $TS$ under the background hypothesis into a p-value.","core_discovery":"On its own terms, the paper claims that 10.5 years of through-going muon data contain no significant continuous neutrino emission from the reconstructed directions of 81 IceCube alert events. Using an unbinned likelihood ratio, the search scans each alert's 90% uncertainty region on a 0.1-degree grid, fits the expected number of signal events and the spectral index at every grid point, and ranks the alerts individually and as a stacked population. The best individual alert yields a p-value of 0.83 and the stacked population yields 0.856, so the paper concludes the observations are compatible with the background-only hypothesis. It also notes that the most significant alert lies close to the galactic plane, with a blazar, PKS 0723-008, about 0.33 arc minutes from the fitted source position.","pith_inferences":["The paper does not state whether the triggering alert event itself remains in the archival sample used to test its own direction; if it does, the null hypothesis is not pure background and the p-values would need to be recomputed with that event removed.","A natural extension is to stack the same archival data on other candidate source lists, such as gamma-ray-loud blazars, to see whether the null result reflects IceCube's alert selection or a broader property of candidate neutrino source directions.","The reported p-values are far enough from rejection that even a trial factor of about two for combining with the all-sky point-source search would leave the background-only conclusion unchanged."],"forward_implications":["If the result stands, IceCube's alert directions do not, as a group, coincide with steady neutrino-emitting sources in the archival sky.","A population of sources emitting steadily at the level needed to produce the alerts would be incompatible with the reported p-values.","The non-detection makes transient or flaring emission the more plausible way for alert-producing sources to hide from time-integrated searches, which is exactly what the announced follow-up will test.","Automating the pipeline would let every future high-energy track alert be checked immediately against 10.5 years of archival data for a continuous source."],"supporting_citations":[{"why":"Defines the IceCube alert system whose emitted directions form the source catalog searched here.","marker":"[3]"},{"why":"Supplies the 10.5-year archival through-going muon sample and event selection used as the data set.","marker":"[7]"},{"why":"Provides the Monte Carlo energy probability densities for signal and background in the likelihood.","marker":"[8]"}],"fun_headline_variants":["No steady neutrino sources behind IceCube alerts","IceCube alerts: no cosmic neutrino excess in 10.5 years","Archival search finds no neutrino point sources at alert directions","Null result: 81 IceCube alerts show no steady neutrino emission","IceCube alert directions yield no significant neutrino signal"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The reported p-values assume that, under the null hypothesis, the archival events in each alert direction are pure background, but the paper does not say whether the alert event that selected that direction is kept out of the archival sample.","fun_headline_variants_meta":{"raw":{"variants":["No steady neutrino sources behind IceCube alerts","IceCube alerts: no cosmic neutrino excess in 10.5 years","Archival search finds no neutrino point sources at alert directions","Null result: 81 IceCube alerts show no steady neutrino emission","IceCube alert directions yield no significant neutrino signal"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000231,"raw_usage":{"total_tokens":1436,"prompt_tokens":849,"completion_tokens":587,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":465,"completion_tokens_details":{"reasoning_tokens":506}},"tokens_in":465,"tokens_out":587,"duration_ms":5641,"temperature":1.0,"reasoning_tokens":506,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:21:23.133541+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the archival data set for the 81 alert events and rerun both the individual and stacking searches with those events excluded; if any p-value changes materially, the background-compatible conclusion depends on self-triggering rather than an independent steady excess.","supporting_citations":[{"cited_title":"Aartsen et al., Science 342 (2013)","cited_arxiv_id":null,"evidence_quote":"Defines the IceCube alert system whose emitted directions form the source catalog searched here."}],"review_version":1}