{"id":"3032bd19-1329-426d-82f1-c1e5c3c72564","arxiv_id":"1908.04869","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A new IceCube event selection identifies neutrino tracks starting inside the detector, yielding a high-purity southern-sky sample above 10 TeV and a proposed realtime alert stream.","lead":"IceCube physicists describe a new way to pick out neutrino collisions that start inside the detector, which should reveal lower-energy neutrinos from the southern sky. The method is being turned into a realtime alert system for multimessenger astronomy.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. 2.1 defines pmiss as a product of log probabilities, which is not a probability and changes sign with the number of dark DOMs; the 10^-3 and 10^-5 cuts are therefore not meaningful as written.","rationale":"The reader's weakest_assumption identifies exactly the same load-bearing concern: Eq. 2.1 defines pmiss as a product of log probabilities, which is not a probability and can be negative. I verified the algebra: each log(p(lambda,0)) = -lambda, so the product alternates in sign with the number of dark DOMs and is not bounded to [0,1]. The thresholds 10^-3 and 10^-5 therefore have no clear probabilistic meaning, and the event selection described in Section 2.2 is not well-defined as written. This is a concrete correctness issue, not a disagreement with consensus. Because the paper is a proceedings contribution with no accompanying code or data, the reader cannot determine whether this is a typographical error in the writeup or a real flaw in the analysis. However, the central claim of high astrophysical purity depends on the pmiss-based veto, so the issue must be resolved before the claimed event rates and sensitivities can be taken at face value. I therefore recommend CONDITIONAL rather than REJECT: if the intended pmiss is the product of Poisson probabilities, the analysis may be sound, but the paper must be corrected and the thresholds re-derived. The proposed toy calculation is a minimal, decisive check that can be done analytically without IceCube simulation.","tokens_in":5818,"tokens_out":3652,"duration_ms":39526,"concrete_test":"Take the published formula literally for two toy events. Event A has N=1 dark DOM with lambda=5; Event B has N=2 dark DOMs each with lambda=5. Eq. 2.1 gives pmiss=-5 for A and +25 for B, so A passes the 10^-3 cut and B fails, despite identical per-DOM missing-charge probabilities. If the intended definition is pmiss = prod(p(lambda_i,0)), both give e^{-10} = 4.5e-5 and pass. The collaboration should confirm which formula is used in the IceCube software and rerun the event selection with the corrected definition; if the set of surviving events changes at the 10^-3 and 10^-5 stages, Table 1 and the quoted purities and sensitivities must be recomputed.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim is that the starting-track selection achieves high astrophysical purity by rejecting events with pmiss below 10^-3 and 10^-5 (Section 2.2). But Eq. 2.1 defines pmiss as the product over dark-region DOMs of log(p(lambda_i, k=0)). Since p(lambda_i,0) = e^{-lambda_i}, each factor is -lambda_i. The product is (-1)^N * prod(lambda_i), where N is the number of dark-region DOMs. This is not a probability, can be negative, and its magnitude depends on N in a way that makes a fixed threshold ill-defined: with one dark DOM and lambda=5, 'pmiss' is -5, which satisfies the <10^-3 cut, while with two such DOMs it is +25, which fails. The surrounding text says pmiss is the probability that the dark-region DOMs observed zero charge, which would instead be prod(p(lambda_i,0)); the 'log' in Eq. 2.1 looks like a typographical intrusion. But as written, the cuts are not a well-defined probability threshold, and the claimed event rates and purities (Table 1) rest on this selection. This is load-bearing because Section 2.2 applies pmiss cuts at three successive veto stages before the BDT; even if the BDT later reshapes the sample, the initial event survival is determined by an undefined quantity. The paper provides no code or data release that would let the reader verify the intended formula.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a new IceCube event selection for southern-sky starting muon-neutrino tracks. The method defines an event-specific veto region around each reconstructed track, computes a quantity pmiss intended to be the probability that the dark-region DOMs saw zero charge under the hypothesis that the event is an incoming muon, and applies successive pmiss thresholds (10^-3, 10^-5) before a boosted decision tree. The authors report expected event rates, effective areas, and pre-trial point-source and galactic-plane sensitivities, and they propose a realtime alert stream for declinations δ ≤ -20° using a modified online selection. The central claimed result is a high-purity astrophysical neutrino sample in the southern sky above roughly 10 TeV, competitive with through-going tracks at mid-TeV energies.","tokens_in":6152,"tokens_out":8820,"duration_ms":88317,"significance":"The idea of using track morphology to build an event-by-event veto is a genuinely useful contribution, and the comparison with through-going tracks indicates a real improvement in southern-sky sensitivity at 1-100 TeV. The paper gives explicit, falsifiable predictions from simulation, and the proposed alert stream is of immediate multimessenger interest. However, the central selection is defined through Eq. (2.1), which is incorrect as written; until that is fixed, the quantitative claims in Table 1 and the sensitivities in Figs. 3 and 4 cannot be assessed reliably.","major_comments":[{"comment":"The quantity pmiss is defined as the product over dark-region DOMs of log(p(λ_i, k=0)). Since p(λ_i, 0) = e^{-λ_i}, each factor is -λ_i, so pmiss = (-1)^N ∏ λ_i, which is not a probability, is not bounded to [0,1], and changes sign with the number N of dark-region DOMs. The thresholds 10^-3 and 10^-5 used in Section 2.2 are therefore not meaningful as written; for example, one dark DOM with λ=5 gives pmiss = -5 and passes the cut, while two such DOMs give pmiss = +25 and fail. The surrounding sentence states the intended definition as a product of the probabilities ∏ p(λ_i, 0), so this is likely a typographical intrusion of 'log', but the equation must be corrected and the thresholds (and Table 1) re-derived from the corrected quantity before the selection is well-defined.","section":"Section 2.1, Eq. (2.1)"},{"comment":"The abstract's headline claim of 'high astrophysical neutrino purity above 10 TeV at declinations less than -30°' is not directly supported by the numbers as presented. Table 1 lists 8 astrophysical versus 33 atmospheric neutrinos per year for δ ≤ -20° (plus 0.8 atmospheric muons), an integrated purity of about 20%; the energy-differential purity that would justify the headline claim is not shown. Please add a purity-versus-reconstructed-energy curve (or the equivalent) for the final selection, and state explicitly the declination and energy range to which the claim refers.","section":"Section 2.2, Table 1"}],"minor_comments":[{"comment":"There are repeated typographical errors that should be corrected, including 'Chernkov' (Cherenkov), 'Cummulative' (Cumulative), 'Therfore' (Therefore), and 'supression' (suppression).","section":"Throughout"},{"comment":"The sentence 'This event selection has the largest neutrino effective area at declinations of less that 30°' is unclear; it should presumably read 'declinations less than -30°' or 'δ < -30°'.","section":"Section 2.2"},{"comment":"The phrase 'pre-run track reconstruction' is unclear; it should likely read 'preliminary track reconstruction' or specify the reconstruction algorithm used for the first pmiss calculation.","section":"Section 2.2"},{"comment":"The comparison in Table 2 uses 8 years for the starting-track selection and 7 years for the through-going-track selection; the livetime difference should be stated, and the sensitivities should be quoted in a way that makes the livetime scaling transparent.","section":"Section 3, Table 2"},{"comment":"The sensitivities in Fig. 4 are pre-trial and no systematic uncertainties are included; this is acceptable for a proceedings contribution, but a sentence explicitly stating what is not included would help avoid overinterpretation.","section":"Section 3"},{"comment":"The sentence 'approximately 17.9 atmospheric neutrinos per year and 5.5 astrophysical neutrinos per year with 50% signalness or greater' is ambiguous: it should be clarified whether 5.5 is the number of astrophysical neutrinos passing the modified selection, or only those with signalness ≥ 50%.","section":"Section 4"}],"recommendation":"major_revision","confidential_remarks":"The central problem appears to be a typographical error in Eq. (2.1) rather than a fundamental conceptual flaw; the surrounding text gives the intended definition as a product of probabilities. With the equation corrected and an explicit purity-versus-energy curve added, the paper would be publishable as a proceedings contribution. I would not insist on new data comparisons or systematics at this stage, but the authors should confirm that the quoted rates and sensitivities are stable under the corrected pmiss definition."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nTwo things to know. The idea is good: replacing the fixed outer-layer veto with an event-specific veto that uses track timing and scans the alleyways between strings is a sensible way to lower the energy threshold for southern-sky starting tracks. But Eq. 2.1, the central definition of pmiss, is not a probability. It's the product of log Poisson zero-probabilities, which flips sign with the number of dark DOMs and can be negative. The 10^-3 and 10^-5 cuts on it are ill-defined as written.\n\nThe novelty is real. Previous starting-track selections used a fixed or charge-scaled veto layer. This one tailors the veto region to each track's expected light deposition and explicitly handles muons entering through the spaces between strings. The realtime southern-sky alert stream is also new, and the projected effective areas and galactic template sensitivities look competitive. The paper is honest about being a proceedings contribution and uses the external astrophysical flux from [5] consistently.\n\nThe Eq. 2.1 problem is load-bearing: Section 2.2 uses pmiss cuts at three stages before the BDT, and Table 1's rates depend on those cuts. The text says pmiss should be the product of zero-charge probabilities, i.e. prod(e^{-λ_i}); the 'log' in Eq. 2.1 looks like a typo. But as written, one dark DOM with λ=5 gives pmiss=-5, which passes a <10^-3 cut, while two such DOMs give +25, which fails. So the threshold is meaningless. This needs fixing first. The paper also lacks systematic uncertainties on effective areas and purities, and the sensitivities are pre-trial—minor for a proceedings paper, but worth noting.\n\nThis deserves a serious referee. The method is novel enough to merit careful scrutiny, and the flaw is repairable. I'd send it back with a request to correct Eq. 2.1 and add a note on systematics. Once fixed, the selection and its sensitivities would be a solid contribution.\n\nFor you: quick read if you follow IceCube's southern sky program, but don't cite it until the equation is corrected.","headline":"Clever new starting-track selection for IceCube's southern sky, but the central veto probability is mis-defined as written and the claimed event rates rest on that equation.","tokens_in":6654,"tokens_out":6401,"would_cite":false,"duration_ms":54938,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"By requiring muon tracks to start inside IceCube, this paper isolates southern-sky astrophysical neutrinos above 10 TeV and lays the basis for a new realtime alert stream.","keywords":["starting tracks","IceCube","neutrino astronomy","southern sky","realtime alerts","atmospheric muon veto","point source search","multimessenger astrophysics"],"falsifier":"Rerun the selection with $p_{\\mathrm{miss}}$ replaced by the actual Poisson probability of observing zero photoelectrons in the dark region, or by a proper log-likelihood, and compare the resulting event rates and purity with Table 1; if the quoted numbers change materially, the central claim is not robust.","tokens_in":5634,"feed_emoji":"🧊","tokens_out":11966,"duration_ms":106653,"temperature":0.7,"pith_summary":"This paper introduces a starting-track event selection for IceCube that aims to isolate astrophysical neutrinos in the southern sky by requiring each muon track to begin inside the detector volume. Instead of a fixed outer veto layer, the selection builds an event-specific veto region from the reconstructed track's timing and light pattern, rejecting both incoming cosmic-ray muons and atmospheric neutrinos accompanied by muons from the same air shower. The resulting sample has high astrophysical neutrino purity above $10\\,\\mathrm{TeV}$ at declinations below $-30^{\\circ}$, with an effective area that surpasses the through-going track sample below roughly $200\\,\\mathrm{TeV}$. The authors conclude that this makes the sample competitive for southern-sky point-source and galactic-plane searches and suitable as a realtime alert stream at $10$--$200\\,\\mathrm{TeV}$.","feed_headline":"Starting neutrino tracks open the southern sky to IceCube alerts","feed_subtitle":"High-purity astrophysical neutrinos above 10 TeV make a new realtime alert stream possible.","key_machinery":"The load-bearing object is the event-specific veto region. Given a reconstructed muon track, DOMs within $350\\,\\mathrm{m}$ of the track are divided by a Cherenkov cone into a 'muon region' (hits consistent with the track timing) and a 'dark region' (where an incoming muon would have been expected to leave light). The track's average luminosity is scaled from the muon-region hits through a likelihood, and $p_{\\mathrm{miss}}$ -- a product over dark-region DOMs of the log Poisson probability of observing zero photoelectrons -- is used to decide whether the event could be an entering muon. Because the detector's strings are spaced $125\\,\\mathrm{m}$ apart, the selection also tests 'alleyway' paths through the gaps, splits the best tracks into 131 segments for timing, and feeds the results into a boosted decision tree whose key variables are the position of the first reconstructed energy loss and the distance of the track start from the detector edge.","core_discovery":"The central claim is that a neutrino-induced muon whose track starts inside IceCube can be separated from an entering atmospheric muon by using the track's good angular resolution to create, for each event, a veto region tailored to that event. Hits consistent with the reconstructed track define a 'muon region', while DOMs behind the Cherenkov cone that should have detected an entering muon define a 'dark region'; the quantity $p_{\\mathrm{miss}}$ is meant to give the probability that the dark region stayed dark under the incoming-muon hypothesis. Combined with alleyway test tracks and a boosted decision tree, this suppresses atmospheric neutrinos in the $10$--$100\\,\\mathrm{TeV}$ range, giving an expected rate of less than one atmospheric muon per year and high astrophysical neutrino purity above $10\\,\\mathrm{TeV}$ for $\\delta < -30^{\\circ}$. The paper further shows pre-trial sensitivities competitive with through-going tracks in the southern sky and proposes a realtime stream that would send southern-sky starting-track alerts at energies of $10$--$200\\,\\mathrm{TeV}$.","pith_inferences":["Editorial inference: if the $p_{\\mathrm{miss}}$ definition is corrected to a true probability, the event rates in Table 1 may shift, but the event-specific veto concept could still hold with retuned thresholds.","Editorial inference: the same track-based veto idea could be transferred to other sparse neutrino detectors with kilometer-scale lattices, where a fixed outer veto wastes fiducial volume at low energies.","Editorial inference: the strong suppression of atmospheric neutrinos at TeV energies should also sharpen measurements of the diffuse astrophysical spectrum and the galactic diffuse emission, since the self-veto matters most below 100 TeV.","Editorial inference: the 16.8 triggers per day from the online stream imply that realtime latency and background rejection at the South Pole will need tight control if the stream is to yield prompt multimessenger alerts."],"forward_implications":["In the southern sky, the starting-track sample gives point-source sensitivities that become competitive with through-going tracks at mid-TeV energies, especially for source spectra with index 2.5 or 3.","The galactic plane template search with starting tracks reaches a sensitivity of $2.45\\times10^{-11}\\,\\mathrm{TeV}^{-1}\\mathrm{cm}^{-2}\\mathrm{s}^{-1}$ for the Fermi $\\pi^{0}$ template, improving on the previous result for the KRA$\\gamma$ models.","A realtime starting-track alert stream would add roughly 5.5 astrophysical neutrinos per year with 50% or greater signalness in the 10--200 TeV range from the southern sky, a lower energy band than current alerts.","The sample's neutrino energy resolution of $0.25$ in $\\log_{10}(E_{\\nu})$ is better than the through-going muon energy resolution because the starting cascade adds information.","The starting-track and through-going selections can be combined to increase future sensitivity."],"supporting_citations":[{"why":"Identifies the self-veto of atmospheric neutrinos accompanied by muons, the background mechanism the selection exploits.","marker":"[3]"},{"why":"Defines the earlier outer-layer veto event selections that this event-specific approach extends.","marker":"[4]"},{"why":"Supplies the astrophysical neutrino flux assumption used to compute expected rates, effective areas, and sensitivities.","marker":"[5]"},{"why":"Provides the boosted decision tree training used in the final selection step.","marker":"[6]"},{"why":"Supplies the Monte Carlo air-shower simulation used to model the atmospheric neutrino self-veto in the southern sky.","marker":"[7]"},{"why":"The through-going track point-source search and effective area that serve as the comparison baseline.","marker":"[8]"},{"why":"Provides the unbinned-likelihood signal-subtraction method used for the galactic plane template fit.","marker":"[11]"},{"why":"Defines the Fermi pi-zero template used in the galactic plane sensitivity comparison.","marker":"[12]"},{"why":"Defines the KRA-gamma galactic diffuse emission template used in the template sensitivity.","marker":"[13]"},{"why":"The existing realtime alert stream whose higher energy range motivates the new lower-energy starting-track stream.","marker":"[14]"}],"fun_headline_variants":["Southern sky neutrinos get a realtime alert stream","IceCube starting tracks filter atmospheric noise for southern alerts","Realtime southern-sky neutrino alerts via IceCube starting tracks","High-purity southern neutrinos: IceCube starting-track alerts","IceCube starting tracks spot southern-sky neutrino alerts in realtime"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The selection depends on $p_{\\mathrm{miss}}$ being a valid probability that an incoming muon would leave the dark region unlit; as written, $p_{\\mathrm{miss}}$ is a product of log Poisson probabilities, which is not a probability and can be negative, so the $10^{-3}$ and $10^{-5}$ thresholds are ill-defined unless this is corrected.","fun_headline_variants_meta":{"raw":{"variants":["Southern sky neutrinos get a realtime alert stream","IceCube starting tracks filter atmospheric noise for southern alerts","Realtime southern-sky neutrino alerts via IceCube starting tracks","High-purity southern neutrinos: IceCube starting-track alerts","IceCube starting tracks spot southern-sky neutrino alerts in realtime"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000656,"raw_usage":{"total_tokens":3002,"prompt_tokens":939,"completion_tokens":2063,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":555,"completion_tokens_details":{"reasoning_tokens":1980}},"tokens_in":555,"tokens_out":2063,"duration_ms":13984,"temperature":1.0,"reasoning_tokens":1980,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:30:07.639153+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Rerun the selection with $p_{\\mathrm{miss}}$ replaced by the actual Poisson probability of observing zero photoelectrons in the dark region, or by a proper log-likelihood, and compare the resulting event rates and purity with Table 1; if the quoted numbers change materially, the central claim is not robust.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the astrophysical neutrino flux assumption used to compute expected rates, effective areas, and sensitivities."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the boosted decision tree training used in the final selection step."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Monte Carlo air-shower simulation used to model the atmospheric neutrino self-veto in the southern sky."},{"cited_title":"Ackermann et al., Astrophys","cited_arxiv_id":null,"evidence_quote":"Defines the Fermi pi-zero template used in the galactic plane sensitivity comparison."},{"cited_title":"Gaggero, D","cited_arxiv_id":null,"evidence_quote":"Defines the KRA-gamma galactic diffuse emission template used in the template sensitivity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The existing realtime alert stream whose higher energy range motivates the new lower-energy starting-track stream."}],"review_version":1}