{"id":"3bdba498-2be3-49cd-8826-e1cf0a224659","arxiv_id":"1908.08248","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 point-like cosmic neutrino source is found in 11 years of ANTARES data; the best full-sky cluster has a post-trial p-value of 23%, and the best IceCube-track correlation has a trial-corrected significance of 1.5% (2.4 sigma).","lead":"After 11 years of data, the ANTARES neutrino telescope finds no clear point-like source of cosmic neutrinos, with the most promising candidate at only 2.4 sigma after trial corrections. The new limits are the strongest from the Southern sky and will guide the next-generation KM3NeT telescope.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Time-randomized pseudo-experiments cannot calibrate the time-integrated searches, so the quoted p-values are unsupported by the stated method.","rationale":"The reader identified the pseudo-experiment procedure as part of the weakest assumption, and I agree that it is load-bearing. My concern is more specific: taken literally, the 'data randomised in time' description in Section 2 cannot generate a valid background distribution for any of the time-integrated searches, because the test statistic in Eq. (2.1) is time-independent when no transient term is included. This would invalidate the full-sky post-trial p-value of 23% and the IceCube-track post-trial p-value of 1.5%, which are central to the paper's conclusion that no significant point-like source is observed. The paper does provide a credible null result and useful upper limits, and the IceCube-track search is a pre-defined candidate test, so the underlying analysis may be correct once the pseudo-experiment procedure is clarified. However, the text as it stands is insufficient to verify the statistical claims. I therefore adjust the verdict from CONDITIONAL to UNVERDICTED pending a concrete demonstration that the time-integrated pseudo-experiments use spatial scrambling rather than time randomization only. This is a resolvable issue, not a rejection of the science.","tokens_in":12534,"tokens_out":7136,"duration_ms":77188,"concrete_test":"Inspect the pseudo-experiment generation code or the corresponding description in refs. [8] and [10] for the time-integrated searches. Then rerun the full-sky search with 1000 pseudo-experiments that scramble right ascension (or otherwise resample spatial coordinates) and compute the post-trial p-value of the cluster at (RA, Dec) = (343.7, 23.6). If the resulting post-trial p-value differs from the quoted 23% by more than the Monte Carlo uncertainty, the reported significance is not supported by the method as stated.","verdict_should_be":"UNVERDICTED","load_bearing_attack":"Section 2 states that background-only pseudo-experiments are 'pseudo-data sets of data randomised in time.' But the likelihood in Eq. (2.1) for the time-integrated searches (full-sky scan, candidate-list, IceCube-track) contains no time term. Randomizing arrival times therefore leaves the test statistic Q unchanged in every pseudo-experiment, so the quoted post-trial p-values (23% for the full-sky cluster, 1.5% for EHE track 3) cannot be produced by the procedure as described. For a steady point-source search, the background ensemble must instead scramble spatial coordinates (e.g., right ascension) while preserving the declination-dependent acceptance; the paper does not state this. This is not merely an external calibration issue but an internal inconsistency in the central statistical method. If the description is an imprecise shorthand, the paper needs to say explicitly that time-integrated searches use spatially scrambled pseudo-experiments and point to the actual procedure in refs. [8] and [10]. As written, the significance estimates that carry the central null result and the one 2.4-sigma correlation are not reproducible.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper, a contribution to the 36th ICRC (2019), reports updated ANTARES searches for point-like cosmic neutrino sources using 11 years of data (January 2007 to December 2017, 3125.4 days livetime per Section 1) for a final sample of 8754 track-like and 195 shower-like events. The method is an unbinned extended maximum likelihood ratio test (Eq. 2.1) over track and shower samples, with signal and background PDFs built from directional and energy terms; time-dependent Gaussian or externally provided time profiles are added for transient searches (Section 2). Five analyses are presented: a full-sky scan, a search at 112 astrophysical candidate positions, a time-integrated search at 75 IceCube-track directions, a time-dependent search at 54 IceCube tracks, and dedicated time-integrated and time-dependent searches toward TXS 0506+056 after IC170922A (Section 3). No significant excess is found: the most significant full-sky cluster (RA 343.7 deg, Dec 23.6 deg) has a post-trial p-value of 23% (1.2 sigma), and the strongest IceCube-track correlation (EHE event 3, at the same sky position) has a trial-corrected p-value of 1.5% (2.4 sigma). The paper also tabulates 90% CL Neyman upper limits on the E^-2 flux normalization for all investigated candidates and fluence limits for the transient analyses, and concludes that no point-like cosmic neutrino source is established.","tokens_in":12737,"tokens_out":13619,"duration_ms":123983,"significance":"If the reported numbers hold, this is the most complete ANTARES point-source statement available at energies below about 100 TeV, where the experiment's sub-degree angular resolution is complementary to IceCube. The statistical framework is standard for the field: a combined track+shower unbinned likelihood with MC-derived energy PDFs, Neyman 90% CL upper limits, and pseudo-experiment-based trial corrections used consistently across the full-sky, candidate-list, and IceCube-track searches. The coincidence of the most significant full-sky cluster with the position of IceCube EHE event 3 is a useful internal cross-check, and the overall null conclusion is consistent with the quoted significances. The dedicated TXS 0506+056 limits under several spectral and time-window hypotheses are a useful complement to the IceCube detection claim. The principal weaknesses are that the manuscript is a preliminary conference contribution (figures labeled PRELIMINARY, compressed methodology) and that the pseudo-experiment calibration is not reproducible as written (major comment below); the quoted p-values and limits are also conditional on MC-simulated detector response and the Barr et al.","major_comments":[{"comment":"The description of the background-only pseudo-experiments is internally inconsistent with the time-integrated searches. Section 2 states that PEs are \"pseudo-data sets of data randomised in time,\" but for the steady searches the likelihood in Eq. (2.1) contains no time term: the signal and background PDFs are products of directional and energy terms only, with the background spatial PDF built from the observed declination distribution of the data. Randomizing arrival times therefore leaves the test statistic Q in Eq. (2.2) unchanged in every pseudo-experiment, so the quoted post-trial p-values (23% for the full-sky cluster in Section 3.1, 1.5% for EHE track 3 in Section 3.2, and 87% for TXS 0506+056 in Section 3.3) cannot be produced by the procedure as stated. The time-integrated searches require a spatial scrambling that preserves the declination-dependent acceptance (e.g., randomization of right ascension), which is exactly what the construction of the background PDF implies. Please either describe that procedure explicitly in Section 2 as applying to the steady searches, or state that the time-integrated significances were obtained with the spatially scrambled PE procedure of refs. [8] and [10] and give the precise reference. As written, the significance estimates that carry the central null result are not reproducible from the stated method; this affects the central claim and must be corrected. Note that for the time-dependent searches the time randomization is legitimate, so the fix is a clarification rather than a re-analysis.","section":"Section 2, Eq. (2.1)-(2.2); Sections 3.1-3.3"}],"minor_comments":[{"comment":"The Abstract quotes a total livetime of 3136 days, while Section 1 quotes 3125.4 days for the same data period; these numbers should be reconciled in the final version.","section":"Abstract / Section 1"},{"comment":"The candidate-count bookkeeping is inconsistent: Section 3.2 states that the list of 106 candidates of [8] was updated with five new TeVCat sources, giving 112 analysed candidates, while Section 3.3 says TXS 0506+056 was added to the \"106 pre-selected sources\" of [8], yielding \"107 investigated sources.\" Please clarify the source counts, since the trial corrections for the candidate searches depend on the exact number of tested directions.","section":"Section 3.2 / Section 3.3"},{"comment":"Please state which trials enter each quoted \"trial-corrected\" or post-trial value: for the full-sky scan the grid of tested directions, for the 112-candidate search the number of candidates, and for the IceCube-track search the 75 candidates plus the freedom of the fitted position within twice the angular error cone. Also specify the convention (presumably two-sided Gaussian) used to convert p-values to sigma, given that 23% is quoted as 1.2 sigma.","section":"Section 3.1 / Section 3.2"},{"comment":"The text should state explicitly whether the quoted p-values and 90% CL limits include any systematic uncertainties (e.g., from the MC-based energy estimators and point spread function, or the Barr et al. (2004) atmospheric background normalization); if they do not, a sentence pointing to the companion paper [8] for the systematic treatment would avoid over-interpretation of the limits.","section":"Section 2 / Section 3.2"},{"comment":"The phrase \"post-trial significance of 90%\" for the transient EHE 15 search should read \"post-trial p-value of 90%\" (or the equivalent significance), since 90% is a p-value rather than a significance in the usual sigma sense; additionally, Table 2 is very dense and the interleaving of HESE, EHE, and AMON rows makes it difficult to parse, so separating the samples into distinct blocks or tables would improve readability.","section":"Section 3.2 / Table 2"}],"recommendation":"major_revision","confidential_remarks":"For the editor: this is a PoS/ICRC proceedings contribution rather than a full journal paper, and several figures are marked PRELIMINARY. The central result (no point-source discovery) is consistent with the ANTARES 9-year published analysis [8] and with the quoted significances, so I do not doubt the overall physics conclusion. The one point that must be fixed before the proceedings version is final is the Section 2 description of the pseudo-experiment procedure, which as written is inconsistent with the time-integrated searches; the authors should confirm in their response that time-integrated significances use spatially scrambled pseudo-experiments (e.g., RA randomization preserving declination) and state this explicitly. The revision should be a small one, essentially one clarifying paragraph, so I would treat the next version as potentially acceptable without another full external review. No concerns about novelty, attribution, or fit to scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is an update of the ANTARES point-source search from nine to eleven years of data, now with 8754 tracks and 195 showers, and it extends the candidate list by five TeVCat sources, adds a time-integrated scan of 75 IceCube tracks, and gives dedicated limits at TXS 0506+056. The headline is a null result — most significant full-sky cluster at about 1.2 sigma post-trial, and the largest IceCube-track correlation at 2.4 sigma post-trial. The paper is careful not to overclaim, and the unbinned likelihood with post-trial p-values is the right machinery. The limits on E^-2 fluxes for 112 candidates and the TXS constraints are a useful input for source population models and for KM3NeT sensitivity projections.\n\nThe soft spots. The biggest one is in Section 2. The text says background pseudo-experiments are 'pseudo-data sets of data randomised in time.' For the time-integrated searches (full-sky, candidate list, IceCube tracks), the likelihood has no time term, so shuffling arrival times leaves the test statistic Q unchanged for every pseudo-experiment. That cannot produce the quoted 23% and 1.5% p-values. The standard procedure for a steady source search is to scramble right ascension (or generate background via MC) while preserving the declination-dependent acceptance. I suspect that is what was actually done — the previous ANTARES papers certainly do it — but the description as written is wrong and the results are not reproducible from the text. This needs an explicit fix, either by correcting the PE description or by pointing to the exact procedure in refs [8] and [10]. As a secondary issue, the livetime is given as 3136 days in the abstract and 3125.4 days in the introduction; that discrepancy should be resolved. Also the figures are marked preliminary and the full analysis details live in earlier papers, which is fine for an ICRC proceeding but means the reader cannot independently verify the numbers.\n\nI think this deserves a serious referee. The analysis is mature and the new numbers are a genuine update, but the statistical calibration description is central to the analysis and must be repaired before the results can be taken at face value. After that, it would be a normal null-result publication.\n\nLet's talk at coffee if you want.","headline":"Useful ANTARES null result, but the description of the background pseudo-experiments is wrong for the steady-state searches and the quoted p-values are not reproducible as written.","tokens_in":13296,"tokens_out":6374,"would_cite":true,"duration_ms":61684,"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":"Eleven years of ANTARES data reveal no significant point-like cosmic neutrino sources, with the strongest cluster at only 23% post-trial probability.","keywords":["point-like neutrino sources","ANTARES","unbinned likelihood","atmospheric neutrino background","IceCube tracks","TXS 0506+056","90% upper limits","multi-messenger astronomy"],"falsifier":"Recompute the full-sky search and the 90% confidence limits after replacing the simulated energy-dependent acceptance with a measured calibration response, for example from in-situ light sources or a high-statistics atmospheric-muon control sample; if the post-trial probability of the $\\alpha=343.7^\\circ$, $\\delta=23.6^\\circ$ cluster moves from 23% to below about 1%, the paper's null conclusion would be overturned in that direction.","tokens_in":12332,"feed_emoji":"🔭","tokens_out":11589,"duration_ms":99751,"temperature":0.7,"pith_summary":"This paper searches 11 years of ANTARES neutrino-telescope data for point-like cosmic neutrino sources, trying to establish whether any single direction in the sky emits neutrinos above the atmospheric background. It finds none: the most significant cluster in a full-sky scan, at right ascension $343.7^\\circ$ and declination $23.6^\\circ$, has a post-trial $p$-value of 23% ($1.2\\sigma$), and the most significant correlation with 75 IceCube tracks has a trial-corrected significance of 1.5% ($2.4\\sigma$). In the absence of a detection, the paper sets 90% confidence upper limits on neutrino flux normalization for 112 astrophysical candidates, 75 IceCube tracks, and the blazar TXS 0506+056. A null result like this matters because it sharpens the constraints on which candidate sources can be neutrino emitters and calibrates the discovery reach of the next-generation cubic-kilometre neutrino telescope.","feed_headline":"No point-like neutrino sources seen in 11 years of ANTARES data","feed_subtitle":"Hottest sky cluster reaches only 1.2 sigma after trials; new 90% limits tighten 112 candidate sources.","key_machinery":"The machinery is the unbinned extended maximum-likelihood ratio test statistic $Q=\\log L_{\\max}-\\log L_{\\mathrm{bkg}}$, summed over track and shower samples. Each event contributes a signal-plus-background likelihood in which the signal and background probability densities factor into a directional and an energy term: the spatial signal PDF is a parametrized point-spread function, the spatial background PDF is the observed declination distribution of the data, and the energy PDFs come from Monte Carlo simulated signal events and from a simulated atmospheric neutrino background spectrum. For time-dependent searches, a Gaussian or externally provided time profile is added to both PDFs. Significance is calibrated by generating background-only pseudo-experiments in which event times are randomized, and the post-trial $p$-value is the fraction of pseudo-experiments whose best cluster is at least as signal-like as the observed one.","core_discovery":"The central claim is that no point-like cosmic neutrino source is detected in the ANTARES data recorded between January 2007 and December 2017, analysed as 8754 track-like and 195 shower-like events. Using an unbinned extended maximum-likelihood ratio, the search scans the full visible sky, tests a list of 112 astrophysical candidates and 75 IceCube track directions, and performs dedicated time-integrated and time-dependent searches toward TXS 0506+056. The hottest full-sky cluster ($\\alpha=343.7^\\circ$, $\\delta=23.6^\\circ$) contains 18 tracks within $5^\\circ$ and 1 shower, but its post-trial significance is only 23%; the same position coincides with IceCube EHE event 3 and yields the strongest IceCube-track correlation at $2.4\\sigma$ after trials. For the candidate list, HESS J0632+057 is the most signal-like with a post-trial significance of $1.4\\sigma$. From these null observations the paper derives 90% confidence upper limits on the flux normalization for an assumed $E^{-2}$ spectrum, including special cut-off-spectrum limits for Eta Carinae, and fluence limits for transient emission toward IceCube tracks.","pith_inferences":["The paper does not pursue a stacked test, but the positional coincidence between the full-sky hottest cluster and IceCube EHE track 3 suggests a combined ANTARES plus IceCube likelihood at that direction could decide whether the $2.4\\sigma$ hint grows or disappears.","The limits assume a steady unbroken $E^{-2}$ spectrum, so the null result constrains hard-spectrum sources most strongly; extending the cut-off-spectrum treatment used for Eta Carinae to the full candidate list would quantify how much weaker the constraints are for softer sources.","If the diffuse astrophysical neutrino flux observed by IceCube were produced by a handful of bright steady sources in the ANTARES visible sky, this search should have seen them; the null result therefore points toward many faint sources or transient or hidden populations, an inference the paper leaves implicit."],"forward_implications":["No previously reported IceCube track, including EHE event 3, reaches more than $2.4\\sigma$ trial-corrected significance as a steady ANTARES point source.","The 90% confidence upper limits exclude steady $E^{-2}$ neutrino fluxes at the few $\\times 10^{-8}\\,\\mathrm{GeV}\\,\\mathrm{cm}^{-2}\\,\\mathrm{s}^{-1}$ level for most of the 112 candidate directions, so any real source among these candidates must be fainter than the quoted bound.","For TXS 0506+056, neither the steady search nor the searches matching IceCube's 2014 flare time windows finds a significant excess; the Gaussian-window limit on the flux normalization at 100 TeV is $4.6\\times 10^{-18}\\,\\mathrm{GeV}^{-1}\\,\\mathrm{cm}^{-2}\\,\\mathrm{s}^{-1}$ for an $E^{-2}$ spectrum.","The authors expect the same likelihood and pseudo-experiment procedure, scaled to a larger cubic-kilometre telescope, to detect the neutrino flux reported by IceCube within months of operation and to make firm statements about Galactic candidates."],"supporting_citations":[{"why":"Describes the ANTARES detector whose 11 years of data constitute the analysis sample.","marker":"[1]"},{"why":"Provides the event selection, energy estimators, candidate list, and likelihood approach that this paper updates and extends.","marker":"[8]"},{"why":"Supplies the atmospheric neutrino spectrum used for the background energy probability density.","marker":"[9]"},{"why":"Define the IceCube High-Energy Starting Event sample among the 75 track directions tested for correlations.","marker":"[2, 3, 4]"},{"why":"Define the IceCube Extremely High-Energy Event sample, including EHE event 3 that yields the strongest correlation.","marker":"[5, 6]"},{"why":"Reports IceCube's neutrino evidence from TXS 0506+056, motivating the dedicated time-integrated and time-dependent ANTARES searches.","marker":"[7]"},{"why":"Defines the earlier ANTARES time-dependent search whose 54-track sample and fluence calculation are used here.","marker":"[10]"},{"why":"Catalogue of TeV sources used to add five new astrophysical candidates to the 112-object list.","marker":"[11]"},{"why":"Gives the confidence-interval method used to set the 90% CL flux upper limits.","marker":"[12]"},{"why":"Provides predicted neutrino fluxes with energy cut-offs for Eta Carinae, used for the special limits on that source.","marker":"[13]"}],"fun_headline_variants":["ANTARES 11y scan yields no point-like neutrino sources","No cosmic neutrino point sources in 11 years of ANTARES","ANTARES: 11 years, no point-source neutrino detections","Null result: ANTARES 11y search finds no neutrino point sources"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quoted significances and upper limits assume the Monte Carlo simulation of the detector response and the adopted atmospheric neutrino background model describe the real 11-year data accurately, and that randomizing event times reproduces all background fluctuations.","fun_headline_variants_meta":{"raw":{"variants":["ANTARES 11y scan yields no point-like neutrino sources","No cosmic neutrino point sources in 11 years of ANTARES","ANTARES: 11 years, no point-source neutrino detections","Null result: ANTARES 11y search finds no neutrino point sources"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000285,"raw_usage":{"total_tokens":1729,"prompt_tokens":1044,"completion_tokens":685,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":660,"completion_tokens_details":{"reasoning_tokens":607}},"tokens_in":660,"tokens_out":685,"duration_ms":6821,"temperature":1.0,"reasoning_tokens":607,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:44:58.477972+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the full-sky search and the 90% confidence limits after replacing the simulated energy-dependent acceptance with a measured calibration response, for example from in-situ light sources or a high-statistics atmospheric-muon control sample; if the post-trial probability of the $\\alpha=343.7^\\circ$, $\\delta=23.6^\\circ$ cluster moves from 23% to below about 1%, the paper's null conclusion would be overturned in that direction.","supporting_citations":[{"cited_title":"Ageron et al., Nucl","cited_arxiv_id":null,"evidence_quote":"Describes the ANTARES detector whose 11 years of data constitute the analysis sample."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the atmospheric neutrino spectrum used for the background energy probability density."},{"cited_title":"Albert et al., Astrophys","cited_arxiv_id":null,"evidence_quote":"Defines the earlier ANTARES time-dependent search whose 54-track sample and fluence calculation are used here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Catalogue of TeV sources used to add five new astrophysical candidates to the 112-object list."},{"cited_title":"Neyman, Phil","cited_arxiv_id":null,"evidence_quote":"Gives the confidence-interval method used to set the 90% CL flux upper limits."},{"cited_title":"Gupta and S","cited_arxiv_id":null,"evidence_quote":"Provides predicted neutrino fluxes with energy cut-offs for Eta Carinae, used for the special limits on that source."}],"review_version":1}