Low-latency neutrino follow-up combining diverse IceCube selections

Christoph Raab; Jessie Thwaites (for the IceCube Collaboration); Sam Hori; Steve Sclafani

REVIEW 3 major objections 6 minor 13 references

Low-latency neutrino follow-up combining diverse IceCube selections

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read IceCube's real-time transient follow-up can be run as a single all-flavor GeV-to-PeV search.

desk verdict Solid IceCube methods paper: the multi-selection FRA is a genuine operational step forward, and the overlap double-counting concern is real but unlikely to overturn the main southern-sky result. read the letter →

arxiv 2507.08748 v1 pith:SPV6BOCU submitted 2025-07-11 astro-ph.HE

Christoph Raab , Sam Hori , Steve Sclafani , Jessie Thwaites (for the IceCube Collaboration) This is my paper · ORCID

classification astro-ph.HE

keywords neutrinoastronomyIceCubeFastResponseAnalysismulti-messengertransientfollow-upunbinnedlikelihoodall-flavorsearch

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

IceCube's Fast Response Analysis currently follows up transients using only TeV muon-neutrino tracks, which see the southern sky poorly because of the atmospheric muon background. This paper argues that the same analysis can instead combine several event selections at once—GFU tracks, GRECO's GeV-scale DeepCore events, and the all-flavor DNN Cascades—so that one fit covers a GeV-to-PeV energy range and all neutrino flavors. The combination works by sharing the source parameters across selections while letting each selection's detector acceptance set its expected signal count, and the paper demonstrates the gain with sensitivity curves for a point-source search. It also lays out a reduced-latency pipeline that makes the computationally expensive reconstructions of the new samples available on a day scale. If correct, the method turns real-time IceCube follow-up into an all-sky, all-flavor search, including the southern sky.

What carries the argument

The central object is the multi-selection unbinned maximum likelihood at the core of the Fast Response Analysis. In it, each event selection contributes its own sample of neutrino candidate events, and the source hypothesis is described by parameters shared across all samples—position, time window, spectral index, and flux normalization—with the expected number of signal events in each sample set by that selection's acceptance. This lets a well-reconstructed track and a cascade with coarser angular resolution constrain the same transient instead of being analysed separately. A second mechanism is the reduced-latency data pipeline: GFU is selected and reconstructed at the South Pole, while GRECO and DNN Cascades are reconstructed in the North from satellite-transmitted data, with a tracking database that lets the analysis retrieve on-time events by calendar day; a minute-scale variant would run the first DNN selection stage at the South Pole and transmit the reduced event stream north. The new instability scores, built from Z-scores of intermediate selection rates, extend GFU's data-quality heuristic to the cascade and DeepCore samples.

What would settle it

Recompute the 90% C.L. sensitivity for the combined point-source search after deleting every event that appears in more than one selection, using the same background model and the same southern-sky bins; if the combined curve no longer improves on GFU alone, the central claim is false.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that the Fast Response Analysis framework can be extended from a single sample of TeV muon tracks to an arbitrary set of event selections without changing the structure of the unbinned likelihood. The three example selections—GFU high-energy tracks, GRECO GeV tracks and cascades from DeepCore, and DNN Cascades high-energy cascades produced by all flavors—are fitted together with shared parameters for source position, emission time window, spectral index, and flux normalization, while each selection's acceptance determines its expected number of signal events. The sensitivity calculation shows the multi-selection combination recovering southern-hemisphere sensitivity that GFU alone loses to the atmospheric muon background, and the energy-differential sensitivity shows the samples covering complementary energy bands. The paper further claims that the more computationally intensive reconstructions of GRECO and DNN Cascades can be brought into low-latency use through a day-scale pipeline built on the existing satellite data flow, with data-quality monitoring via new instability scores.

Load-bearing premise

The result assumes that counting the same physical neutrino twice, when it passes more than one event selection, is rare enough to leave the test statistic and sensitivity estimates unchanged; the paper states that removing such duplicate events remains to be done.

Editorial extensions

If this is right

A single follow-up analysis can cover neutrino energies from roughly GeV to PeV and all flavors, instead of only TeV muon tracks.
The southern-hemisphere sensitivity that GFU loses to the atmospheric muon background is partially recovered by adding cascade and DeepCore selections.
GRECO and DNN Cascades events become available for real-time transient searches at day-scale latency through the proposed pipeline, with a minute-scale option under development.
Because all selections share source parameters, a combined fit can use every event type to constrain one transient hypothesis rather than running separate searches.
The framework supports arbitrary event selections, so future low-latency samples can be added without changing the likelihood structure.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the duplicate events turn out to cluster near real transient candidates, the double-counting could mimic or mask a source; the first validation step is therefore a duplicate-removed reanalysis, not just a check of the total overlap fractions.
The same shared-parameter likelihood could be applied to archival IceCube data, where the day-scale latency constraint does not apply, potentially sharpening all-sky point-source limits across the full energy range.
Should the minute-scale pipeline reach production, IceCube would be able to issue all-flavor alerts from GeV to PeV within tens of minutes, which would make fast-evolving phenomena such as GRB prompt emission accessible to real-time multi-messenger follow-up.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The genuinely new thing here is combining GFU tracks, GRECO GeV events, and DNN Cascades into a single Fast Response Analysis with shared source parameters. That is more than an incremental tweak: the previous real-time FRA used only TeV tracks, and the combination is what gives you broader energy coverage, all flavors, and better southern-sky sensitivity. The proposed day-scale latency pipeline for the computationally heavier reconstructions is also new and clearly described, with a plausible path to minute-scale in the future. The paper credits the original FRA and the likelihood method properly, and the validation is honest: simulated skymaps and sensitivity curves support the qualitative claims.

The main soft spot is the overlap between selections. The authors state that 1.8% of GFU events are also in GRECO, 0.03% in DNN Cascades, and GRECO–DNN Cascades overlap by 0.2%, and Section 6 lists removal of these duplicates as a remaining issue. In an unbinned likelihood, a duplicated event is counted twice, which does inflate its weight and can bias the test statistic. The stress-test is right to flag this as a load-bearing concern for the quantitative sensitivity curves. But let me put it in proportion: the overlap fractions are small, and the southern-sky gain is driven primarily by DNN Cascades and GRECO, which are largely disjoint from GFU. So the main conclusion—that combining selections regains southern sensitivity—would almost certainly survive a deduplicated rerun. What is at risk is the exact sensitivity values, not the big picture.

Other soft spots are minor for a proceedings paper: the sensitivity curves lack uncertainty bands, the common data-quality threshold is not fully specified, and no code or data are provided (typical for IceCube, but it limits reproducibility). The circularity burden is low—they are not fitting parameters and re-predicting them; self-citations are appropriate.

Who gets value: anyone building real-time follow-up pipelines, or using IceCube alerts for multi-messenger work. It is a conference proceedings, so the depth is limited, but the logic is coherent and the limitations are acknowledged rather than hidden. I would send this to a serious referee, mainly to push for a deduplicated sensitivity figure and some uncertainty quantification. Treat the numbers as preliminary until that is done.

Referee Report

3 major / 6 minor

Summary. The paper proposes extending IceCube's Fast Response Analysis (FRA) to combine multiple event selections—the existing GFU high-energy tracks, the GRECO DeepCore tracks and cascades, and the DNN Cascades high-energy cascade sample—into a single unbinned maximum-likelihood follow-up that shares source parameters across selections. It describes a day-scale latency pipeline for the more computationally expensive selections, introduces instability-score data-quality monitoring for the new samples, and validates the approach with a simulated background sky map/test-statistic map (Fig. 5) and with 90% C.L. sensitivity curves for individual and combined selections (Fig. 6). The central claim is that the multi-selection FRA broadens the accessible energy range, includes all neutrino flavors, and regains southern-sky sensitivity lost by the GFU muon background.

Significance. If the quantitative performance claims hold, this is a valuable and timely extension of IceCube's realtime follow-up program, enabling GeV-to-PeV, all-flavor searches for transient neutrino sources with day-scale latency. The paper is unusually transparent: it reports overlap fractions between selections, explicitly lists unresolved items (Section 6), and grounds the sensitivity estimates in archival data and Monte Carlo rather than fitting parameters to the same predictions. The main risk is not circularity but the unresolved treatment of multiply selected events, which can bias the combined sensitivity, together with an unspecified data-quality threshold. These issues are local and fixable, but they currently leave the magnitude of the southern-hemisphere gain and the broadened energy range quantitatively unproven.

major comments (3)

[Section 5, Fig. 6a] The combined-sensitivity curves are computed while intentionally including events that appear in multiple selections (1.8% of GFU events also in GRECO, 0.03% also in DNN Cascades, and 0.2% overlap between GRECO and DNN Cascades), and Section 6 states that their removal remains to be resolved. In the unbinned likelihood, a physical event present in two selections contributes a likelihood-ratio factor for each selection, effectively doubling its signal-vs-background weight; for the rare signal events that drive the sensitivity this can inflate the test statistic and make the combined curves appear better than they are. I request a quantitative estimate of this bias, or a rerun of Fig. 6a with deduplicated events, before the southern-hemisphere sensitivity gain is used as a central claim.
[Section 4, Fig. 6] The instability-score threshold used to define good data runs for the sensitivity calculations is never specified. Figure 4 shows ROC curves and marks a threshold value of 10, but the text does not state whether this threshold was applied to all three event selections when producing Figs. 5 and 6, nor whether the resulting sensitivity estimates are robust to the choice of threshold. Please state the exact threshold(s) used and justify the value, since the data-quality selection is a free parameter of the analysis.
[Section 5, Fig. 6b] The differential sensitivity curves in Fig. 6b are shown for each event selection individually, but no combined differential curve is presented. The abstract's claim that the multi-selection combination is 'sensitive to a broader energy range of a neutrino transient spectrum' is therefore not directly validated by the sensitivity estimates shown; please add a combined differential sensitivity curve, or explicitly state that the combined search inherits the union of the individual energy responses and quantify the expected gain.

minor comments (6)

[Fig. 6b] The axis label 'E^2' and the caption's 'E^{-2} per-flavor flux' are inconsistent; please clarify whether the plotted quantity is E^2 dN/dE or an E^{-2} flux.
[Fig. 2] The CPU-time axis in Fig. 2 is not labeled with units in the caption; please add units (e.g., seconds per event) or state them explicitly in the text.
[Abstract and Section 1] 'Moreso' should be 'Moreover' for standard usage.
[Fig. 5] The caption describes 'randomized events' while the text refers to a 'simulated background-only observation'; please use consistent terminology.
[Fig. 6a] The legend entry 'DNNC combined' is ambiguous; please distinguish the DNN Cascades-only curve from the multi-selection combined curve.
[Fig. 6] The sensitivity curves have no uncertainty bands; the 'IceCube preliminary' label is helpful, but a sentence stating which uncertainties (Monte Carlo statistics, reconstruction systematics, etc.) are included or omitted would aid interpretation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the combined-FRA sensitivity is computed from Monte Carlo and archival data with an established likelihood, not fitted from, or defined in terms of, the claimed output.

full rationale

The paper's central result is the 90% C.L. sensitivity of a multi-selection Fast Response Analysis (Section 5, Fig. 6). The sensitivity curves are obtained by injecting Monte Carlo signal into archival background and applying the previously published unbinned likelihood [13]; the event selections GFU, GRECO, and DNN Cascades are independent inputs with quoted astrophysical and atmospheric compositions [10]. No parameter is fitted to data and then renamed a prediction. The claim that adding GRECO and DNN Cascades regains southern-sky sensitivity is a computed consequence of those inputs, not an equivalence to them. Self-citations to IceCube selections, analyses, and the likelihood are legitimate external evidence: the selections and detector response are analysis products with stated astrophysical fractions, and the likelihood is a standard published method. The acknowledged overlap of events appearing in multiple selections (1.8%, 0.03%, 0.2%) is a residual systematic and correctness concern that could bias the combined sensitivity, but it is disclosed rather than concealed and does not make the derivation circular; the paper lists removing such duplicates as future work in Section 6. There is no imported uniqueness theorem, no ansatz smuggled in via citation, and no re-labeling of a known result as a new one.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central sensitivity claim relies on the inherited unbinned likelihood, on the accuracy of simulations and archival data, and on the rarity of duplicate events across selections. The only hand-chosen numeric parameter is the data-quality instability score threshold. No new physical entities are introduced.

free parameters (1)

Instability score threshold = 10 (indicated by filled circles in Fig. 4)
A common threshold applied to the three Z-score-based instability measures to flag bad runs. It is chosen from ROC curves rather than derived, and the paper says a common threshold or multivariate cut 'can be optimized' for a particular analysis.

assumptions (4)

standard math The unbinned maximum likelihood framework of ref. [13] is directly applicable to each event selection and to their combination with shared source parameters.
Section 5 states the combination uses 'the previously implemented unbinned likelihood method [13]' without re-derivation.
domain assumption Archival data and Monte Carlo simulation accurately model the background and signal distributions for GFU, GRECO, and DNN Cascades in the combined analysis.
Used to compute the sensitivities in fig. 6 and the simulated skymap in fig. 5; the paper provides no validation of simulation fidelity beyond citing prior uses.
domain assumption Events appearing in multiple selections can be treated as statistically independent (their overlap is negligible).
Section 6 explicitly lists removal of overlapping events as a remaining point to resolve, so the current implementation assumes overlaps are rare enough to ignore.
domain assumption The day-scale latency stream is complete for the analysis time window before the FRA is run (the 'Check completeness' step in fig. 3).
Section 3 describes that >99% of events arrive within about 1.5 days and half within 16 hours, but a ±1 day window could include events that have not yet arrived; the paper does not quantify the completeness-check success rate.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Low-latency neutrino follow-up combining diverse IceCube selections." pith.science (2026). https://pith.science/paper/SPV6BOCU

@misc{pith2026250708748,
  author       = {Pith},
  title        = {Pith review of: Low-latency neutrino follow-up combining diverse IceCube selections},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SPV6BOCU}},
  note         = {Machine review of arXiv:2507.08748}
}

read the original abstract

Neutrino observations are crucial for multi-messenger astronomy, but currently limited by effective area and atmospheric background. However, while other telescopes must point their limited field of view, IceCube can perform full-sky realtime follow-up of astrophysical transient sources with high uptime. For this, IceCube uses its Fast Response Analysis (FRA), which can provide results from an unbinned maximum likelihood analysis within tens of minutes of an astrophysical transient. Besides manually selected transient candidates, it also routinely scans areas of the sky compatible with gravitational wave alerts from LIGO/Virgo/KAGRA and the IceCube event singlets most likely to originate from an astrophysical source. Currently the analysis uses TeV muon neutrino candidate events whose signature permits especially precise angular reconstruction, selected and reconstructed at the South Pole and transmitted with low-latency over satellite. Recently, different event selections are also being included in IceCube analyses. These efforts include the follow-up of gravitational wave events with GeV neutrinos detected by IceCube-DeepCore and the observation of the Galactic plane with cascade events produced by all neutrino flavors. If made available on a day-scale latency, these event selections can also be used in FRAs. Moreso, multiple event samples can be combined in a Fast Response Analysis that is sensitive to a broader energy range of a neutrino transient spectrum and ensures the inclusion of all neutrino flavors. We present the analysis method and technical aspects of such an extension of the existing framework. This includes a proposed new pipeline allowing the inclusion of the more computationally-intensive reconstruction methods used by the aforementioned event selections. The extension is validated using example analyses implemented in this framework. (abbreviated)

Figures

Figures reproduced from arXiv: 2507.08748 by the authors.

**Figure 1.** Skymap zoomed around the FRA best-fit for the ±1 day search for IceCube-Cascade 240714A. The two coincident GFU events are shown (purple) with the FRA best-fit provided in real-time (star). In real-time, an error around the best-fit was also provided with a radius of 0.3 ◦ (shown as a black circle). The nearest Fermi-LAT source (top of plot) is shown in an orange triangle. Einstein Probe followed up our ATel in real… view at source ↗

**Figure 3.** Flow chart of reduced latency data stream methods. The upper path corresponds to the day-scale latency method proposed in section 3 for DNN Cascades and GRECO, while the lower path describes the one currently used by GFU. only rely on timing information to identify astrophysical activity. The goal of the current work is to allow a low-latency application of these event selections to transient searches. Some of these… view at source ↗

**Figure 2.** [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: ROC curves comparing possible thresholds to be set on the three instability scores. The x-axis is a "false positive" rejection rate during good runs, while the y-axis is the "true positive" rejection during bad runs. The threshold of 10 is indicated by filled circles …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 9 canonical work pages

[1]

IceCube collaboration, Neutrino emission from the direction of the blazar TXS 0506+056 prior to the IceCube-170922A alert, Science 361 (2018) 147 [1807.08794]

arXiv 2018
[2]

IceCube collaboration, Evidence for neutrino emission from the nearby active galaxy NGC 1068, Science 378(2022) 538 [2211.09972]

arXiv 2022
[3]

IceCube collaboration, The IceCube neutrino observatory: Instrumentation and online systems, JINST 12(2017) P03012 [1612.05093]

arXiv 2017
[4]

IceCube, MAGIC, VERITAScollaboration, Very high-energy gamma-ray follow-up program using neutrino triggers from IceCube, Journal of Instrumentation 11(2016)P11009 [1610.01814]

work page Pith review arXiv 2016
[5]

IceCube collaboration, New Public Neutrino Alerts for Clusters of IceCube Events, PoS ICRC2025(these proceedings) 949

work page
[6]

IceCubecollaboration, Results from IceCube Searches for High-energy Neutrinos Coincident with Gravitational-Wave Alerts in LVK O4, PoS ICRC2025(these proceedings) 1113

work page
[7]

IceCube collaboration, Follow-up of Astrophysical Transients in Real Time with the IceCube Neutrino Observatory, ApJ 910 (2021) 4

work page 2021
[8]

IceCube collaboration, Constraints on Populations of Neutrino Sources from Searches in the Directions of IceCube Neutrino Alerts, ApJ 951 (2023) 45

work page 2023

Show all 13 references

[9]

IceCube collaboration, Multi-Energy and Multi-Sample Searches for IceCube Neutrinos from LIGO/Virgo Gravitational Wave Events, PoS ICRC2025(these proceedings) 915
[10]

IceCube collaboration, Observation of high-energy neutrinos from the Galactic plane , Science 380 (2023) adc9818 [2307.04427]

2023
[11]

Murase, M

K. Murase, M. Mukhopadhyay, A. Kheirandish, S.S. Kimura and K. Fang,Neutrinos from the Brightest Gamma-Ray Burst?, ApJL 941(2022) L10

2022
[12]

IceCube collaboration, Probing neutrino emission at GeV energies from compact binary mergers detected during O4 with the IceCube Neutrino Observatory, PoS ICRC2025(these proceedings) 947
[13]

IceCube collaboration, All-sky search for time-integrated neutrino emission from astrophysical sources with 7 years of IceCube data, Astrophys. J. 835(2017) 151. 8 Multi-selection Fast Response Full Author List: IceCube Collaboration R. Abbasi16, M. Ackermann63, J. Adams17, S....

2017

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

IceCube collaboration, Neutrino emission from the direction of the blazar TXS 0506+056 prior to the IceCube-170922A alert, Science 361 (2018) 147 [1807.08794]

arXiv 2018

[2] [2]

IceCube collaboration, Evidence for neutrino emission from the nearby active galaxy NGC 1068, Science 378(2022) 538 [2211.09972]

arXiv 2022

[3] [3]

IceCube collaboration, The IceCube neutrino observatory: Instrumentation and online systems, JINST 12(2017) P03012 [1612.05093]

arXiv 2017

[4] [4]

IceCube, MAGIC, VERITAScollaboration, Very high-energy gamma-ray follow-up program using neutrino triggers from IceCube, Journal of Instrumentation 11(2016)P11009 [1610.01814]

work page Pith review arXiv 2016

[5] [5]

IceCube collaboration, New Public Neutrino Alerts for Clusters of IceCube Events, PoS ICRC2025(these proceedings) 949

work page

[6] [6]

IceCubecollaboration, Results from IceCube Searches for High-energy Neutrinos Coincident with Gravitational-Wave Alerts in LVK O4, PoS ICRC2025(these proceedings) 1113

work page

[7] [7]

IceCube collaboration, Follow-up of Astrophysical Transients in Real Time with the IceCube Neutrino Observatory, ApJ 910 (2021) 4

work page 2021

[8] [8]

IceCube collaboration, Constraints on Populations of Neutrino Sources from Searches in the Directions of IceCube Neutrino Alerts, ApJ 951 (2023) 45

work page 2023