{"id":"fbfc0cdd-9af0-4c29-806a-d01e87f15e25","arxiv_id":"2607.15130","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A new open-source pipeline recovers arcminute positions for gamma-ray bursts during Swift spacecraft slews, complementing NITRATES and potentially doubling Swift-BAT's precise localization rate.","lead":"BAT-GLIMPSE is a new autonomous software pipeline that finds the positions of gamma-ray bursts in Swift-BAT data, including while the Swift spacecraft is slewing—a time when onboard triggering is disabled. On 66 test bursts it recovered arcminute positions for 43, and it may roughly double Swift's precise-localization rate when combined with the existing NITRATES search.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Validation sample is selected on the outcome (prior arcmin localizations in GCNs); the 88% slew recovery and factor-of-two gain are conditional on detectability and likely optimistic for the full GUANO population.","rationale":"The reader's weakest_assumption identifies exactly the load-bearing concern: the validation sample is selected from GCN circulars that already reported arcminute localizations, so the recovery rates and sensitivity ratio are conditioned on sources already found on the ground. I agree with this assessment. The pipeline itself is credible engineering: it is open-source, demonstrated on archival data, used in real time, and cross-validated against NITRATES and afterglow detections. The concern is not about the existence of the capability but about the quantitative headline figures. The paper's own Sec 5.3 lower bound (+17%) and upper bound (+125%) bracket the 'double' claim, and the abstract selects the upper end without the caveats. This is a significant but not fatal issue: with explicit statements that the 88% and factor-of-two are conditional on already-detectable events and on optimistic assumptions, the central technological claim remains intact. Therefore the reader's CONDITIONAL verdict is appropriate; I would not change it.","tokens_in":23246,"tokens_out":7714,"duration_ms":87334,"concrete_test":"Rerun BAT-GLIMPSE on the 146 GUANO GRB triggers that were detected but NOT reported with arcminute positions in GCNs (i.e., the complement of the 68 selected in Sec 3), and on a set of simulated slew observations with injected GRBs spanning the partial-coding/flux plane. Compare the recovery fraction on the full trigger set and on low-PC (<10%) injections with the reported 43/66 and 14/16 rates. If the full-population recovery is significantly lower (e.g., <50% for PC<10%), then the 88% and 'double' claims are overestimates and the paper should report a conditional recovery rate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claims—88% slew recovery (Sec 4.2) and the estimated doubling of the arcminute-localization rate (abstract, Sec 7)—are built on a validation set that is not representative. The 66 GRBs were selected from GUANO GCN circulars that already reported an O(arcmin) localization (Sec 3). Thus all 16 slew events had already been found and localized on the ground by imaging/mosaic before BAT-GLIMPSE was run; the sample contains no slews where the source was too faint or too far off-axis to be localized. The 43/66 and 14/16 recovery fractions therefore measure reproducibility on a detectability-biased sample, not detection efficiency on the full trigger population. The paper acknowledges this only obliquely in Sec 5.3, as a reason the +17% lower bound is a lower limit, but the abstract's '88%' and 'double' omit the caveat. Moreover, λ_arcmin in Eq. (13) is inferred from Fig 9 using this same selected sample; the resulting ρ_arcmin≲2.25 is an upper bound under optimistic assumptions, not an expected value. If the GUANO population contains a substantial fraction of fainter/lower-partial-coding bursts, the true slew recovery and the true gain could be closer to the paper's own lower bound (+17%).","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents BAT-GLIMPSE, an open-source pipeline for ground-based localization of Swift-BAT TTE data using coded-mask imaging and mosaic techniques, with autonomous operation during both pointed observations and spacecraft slews. The pipeline is validated on 66 GUANO GRBs whose GCN circulars already reported an arcminute localization; 43 are recovered with angular separations ≲5′ and 14/16 events occurring during slews are recovered. The paper further estimates that BAT-GLIMPSE combined with NITRATES increases the arcminute-localized GRB detection rate relative to the onboard trigger rate by roughly +17% to +125%, and describes real-time use during LVK O4 and in the ULTRA-Swift early-warning program, together with upper-limit map products.","tokens_in":1541,"tokens_out":1537,"duration_ms":190991,"significance":"If the population-level claims held, this would be a valuable contribution: BAT-GLIMPSE addresses a real coverage gap created by disabled onboard triggering during slews, uses a transferable imaging/mosaic methodology, ships open-source code, and reports 29 events with independent afterglow or IPN confirmation. The offset distribution for recovered events and the practical SNR>7 reliability guidance are also useful. However, the headline recovery fractions and the factor-of-two gain estimate are currently built on a validation sample that was selected for previously published arcminute localizations. The paper is therefore more convincing as a demonstration of the pipeline on known, bright, already-localized events than as a measurement of detection efficiency on the full GUANO population. The central quantitative claims need to be re-based or explicitly re-framed as conditional before publication.","major_comments":[{"comment":"The validation sample is constructed from GCN circulars that already reported O(arcmin) GUANO localizations. The 43/66 and 14/16 recovery fractions therefore measure reproducibility on events that had already been found and localized on the ground, not detection efficiency for all GRBs occurring during slews. In particular, no slew event too faint or too far off-axis to have produced a GCN localization appears in the sample, so the abstract's 'approximately 88% of the GRBs occurring during spacecraft slews are recovered' is a population-level statement the design cannot support. The paper notes in Sec. 5.3 that the archival re-run was not done on the full GUANO trigger set, but this limitation is not carried into the abstract or Sec. 7. Please either rerun BAT-GLIMPSE on a sample of GUANO triggers selected independently of whether a localization was published—at least over all slews with","section":"Sec. 3, Sec. 4.2, abstract"},{"comment":"The gain estimate for arcminute-localized GRBs uses the same outcome-selected sample. lambda_arcmin is inferred from Fig. 9, whose points are all GRBs already localized by NITRATES with arcmin precision; the ratio of NITRATES-to-BAT-GLIMPSE detections is therefore not an unbiased estimate of g_inFOV lambda_arcmin for the full GUANO population. The text identifies the external-trigger sensitivity bias but does not address the more basic outcome-selection bias. Consequently, rho_arcmin less-or-equal 2.25 is an upper bound in an optimistic scenario, not an expected value, and the 'double' wording in the abstract and Sec. 7 overstates the support. The abstract should say, at most, that the measured lower limit is +17% and that the upper end of the estimated range is +125% under the assumption that the selected sample is representative. A full-GUANO re-run or a clearly conditional analysis is","section":"Sec. 5.3, Eq. (13), Fig. 9"}],"minor_comments":[{"comment":"'In 16 cases (~18%)' appears inconsistent with the stated totals: 16/68 is about 23.5% and 16/214 is about 7.5%. Please clarify the denominator and define the percentage.","section":"Sec. 3"},{"comment":"If R_BAT is the total onboard trigger rate per calendar year, the BAT-GLIMPSE slew term should scale as f_s/(1-f_s), not f_s, because R_BAT is accrued only during non-slew time. The numerical effect is small at f_s=0.15, but the formula as written is dimensionally inconsistent.","section":"Sec. 5.2, Eq. (7)"},{"comment":"The abstract says the pipelines 'are estimated to double' the onboard arcminute-localization rate, while Sec. 7 says 'by up to a factor of two' and Sec. 5.3 gives a range from +17% to +125% under optimistic assumptions. Please align the wording so the abstract does not present the upper end of the range as the expected value.","section":"Abstract and Sec. 7"},{"comment":"Several bold rows in Table 1 correspond to GCNs produced in real time by BAT-GLIMPSE itself after 2025 (Sec. 2.4). These are not independent archival validations. The operations are disclosed, but the validation section should either exclude or separately analyze the real-time detections.","section":"Sec. 4.2 and Table 1"},{"comment":"No uncertainties are reported for the recovery fractions. The 88% figure rests on 14/16 successes; a binomial 95% confidence interval is approximately 62 to 98 percent. Please report confidence intervals for 43/66 and 14/16.","section":"Sec. 4.2"},{"comment":"The SNR histogram comparison to a Gaussian assumes the null distribution; it would help to state the expected distribution under pure Poisson noise and to define the source-detection threshold of SNR>7 in the same section. Also 'reducing the the latency' in Sec. 7 should read 'reducing the latency.'","section":"Sec. 4.1, Fig. 4"}],"recommendation":"major_revision","confidential_remarks":"The pipeline is promising and the paper is generally well organized, but the headline quantitative claims currently outrun the validation sample. I would encourage the editor to request either a rerun on a representative set of GUANO triggers or a thorough re-framing of all recovery-rate and gain claims as conditional on prior localization. If the authors can do the full-GUANO re-run, the paper would be considerably stronger and could be suitable for acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth reading if you care about Swift-BAT or ground-based GRB localization. The real contribution is a working, open-source, automated pipeline (BAT-GLIMPSE) that localizes transients in BAT TTE data during slews using imaging and mosaics, and it is already running in real time for GUANO. That is genuinely new: Copete's thesis had the concept, but this is an operational implementation with a public repo and a 66-GRB validation set, 29 of which have independent afterglow or IPN confirmation. When it finds the source, the positions match published GCN positions to within ~5 arcmin. The engineering is solid and the paper is clearly written.\n\nThe soft spot is the quantitative framing. The 66 events were selected from GCN circulars that already reported arcminute localizations, so the recovery fractions—43/66 overall, 14/16 during slews—measure how often the pipeline reproduces a known position, not how often it would find a burst in the full GUANO trigger population. The two slew events it missed were at partial coding <10%, which hints that true slew recovery for faint, far-off-axis bursts is lower. The paper does acknowledge in Sec 5.3 that the archival rerun was not done on the full GUANO set, but the abstract's '88%' and 'double' drop that caveat. The body gives a range of +17% to +125% for the arcminute-localization gain, with an upper bound of ~2.25 under optimistic assumptions. So the abstract's 'double' is the top of the range, not the central expectation. I would want that reframed before publication.\n\nThe sensitivity model in Sec 5 has several free parameters (g_inFOV, g_outFOV, epsilon_glimpse, epsilon_ext, lambda_arcmin) and depends on the same selected sample for lambda_arcmin, so the +125% upper bound should be treated as a model-dependent ceiling, not a measured gain. The +17% lower bound, from actual GCN circular rates, is the more defensible number.\n\nWho is this for? Anyone doing Swift/GUANO follow-up or multi-messenger work will want to know this pipeline exists and that it works in real time. The ULTRA-Swift section is context, not the core. It deserves a serious referee; the main asks should be to reframe the abstract and to either run the pipeline on the full GUANO trigger set or explicitly label the recovery rates as conditional on already-localized events. I would send it to review.","headline":"Useful, well-engineered pipeline paper; the 88% and 'double' headlines are conditional on a validation sample selected for already being localized, so treat those as optimistic bounds.","tokens_in":24139,"tokens_out":2191,"would_cite":true,"duration_ms":24623,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"An autonomous pipeline called BAT-GLIMPSE localizes gamma-ray bursts during Swift spacecraft slews, when the onboard trigger is off, and together with the NITRATES search is estimated to double the number of arcminute-localized bursts Swift","keywords":["gamma-ray bursts","coded-mask imaging","mosaic imaging","Swift-BAT","arcminute localization","spacecraft slew","transient localization","multi-messenger astrophysics"],"falsifier":"Run BAT-GLIMPSE blindly on the full set of GUANO-triggered BAT time-tagged data from 2020 through 2026, including triggers that did not yield a published arcminute localization, and compare recovered positions with independent detections from other instruments; if the slew recovery rate on that unbiased sample falls well below 88%, or if the overall recovery fraction drops markedly, the central claims would not hold.","tokens_in":23145,"feed_emoji":"🔭","tokens_out":5966,"duration_ms":56534,"temperature":0.7,"pith_summary":"The Burst Alert Telescope on the Swift observatory can pinpoint gamma-ray bursts to a few arcminutes, but its onboard trigger is disabled whenever the spacecraft is slewing, so bursts that happen during a slew go unlocalized in real time. This paper presents BAT-GLIMPSE, a fully autonomous ground pipeline that takes the downlinked time-tagged BAT data and automatically chooses between coded-mask imaging (for very short time bins) and mosaic imaging (for longer bins overlapping the slew). Validated on 66 bursts that had already been reported, the pipeline recovers 43 arcminute positions with typical offsets under 5 arcminutes, including 88% of the bursts occurring during slews. If these rates hold more generally, BAT-GLIMPSE plus NITRATES would roughly double the rate at which Swift-BAT localizes bursts to arcminute precision, closing the slew gap and improving follow-up of gravitational-wave counterparts.","feed_headline":"Recovers 88% of gamma-ray bursts that fire during Swift slews","feed_subtitle":"Ground pipeline plus NITRATES could double Swift-BAT's arcminute localizations.","key_machinery":"The load-bearing method is coded-mask imaging combined with a mosaic step. For each candidate time bin, the detector-plane image is cross-correlated with the coded mask to form a sky image; if the bin is short enough (under 0.2 seconds) that the source would not move more than the instrument point-spread function during a slew, that single image is used. Otherwise the bin is split into 0.2-second sub-bins, each is imaged, and the resulting sky images are weighted by signal-to-noise and summed into a mosaic. Automatic selection between these two modes based on spacecraft attitude is what lets the pipeline work during the slews when the onboard trigger is switched off.","core_discovery":"The paper's central claim is that a single automated pipeline can recover arcminute positions for gamma-ray transients from Swift-BAT data both during stationary pointing and during spacecraft slews, by splitting long time bins into 0.2-second images and co-adding them into a motion-corrected mosaic that tracks the source across the detector. Running on 66 bursts that had already been localized on the ground, BAT-GLIMPSE reproduces published positions in 43 cases with offsets of about 5 arcminutes or less; for the 16 bursts that occurred during slews, it recovers 14, eight by mosaic and six by short imaging. The authors further argue, from the sensitivity of NITRATES and the roughly 15% of t","pith_inferences":["The headline recovery percentages are measured on bursts that were already localized and reported on the ground; on a complete, unbiased set of GUANO triggers the true slew recovery could be lower, especially for faint or off-axis bursts.","The 'doubling' estimate is a scaling argument the paper brackets between +17% and +125%; a full archival re-run over all GUANO triggers would pin down where in that range the real gain lies.","Because BAT-GLIMPSE works on downlinked data after the fact, it also enables archival searches for untriggered transients in slew intervals, effectively turning slew time into survey time.","The partial-coding-weighted light-curve diagnostic offers a generic template for predicting whether a source is inside the coded-mask field of view during a slew, useful for any future coded-mask instrument."],"forward_implications":["Swift-BAT can now produce arcminute localizations for bursts occurring during slews, a gap that previously left those events unlocalized.","Combined with NITRATES, the two pipelines are estimated to double the onboard arcminute-localization rate of Swift-BAT.","BAT-GLIMPSE provides localizations up to 30-60 times faster than NITRATES when a source is bright enough for imaging, enabling quicker follow-up.","The pipeline has operated in real time since 2025 and has already produced seven GCN circulars with arcminute positions, several confirmed by X-ray afterglow.","The method transfers to other coded-mask missions, such as the upcoming SVOM-ECLAIRs or THESEUS-XGIS, for localizing transients during slews."],"fun_headline_variants":["BAT-GLIMPSE recovers gamma-ray bursts during Swift slews","88% of slew-time GRBs recovered by new Swift pipeline","Ground software could double Swift-BAT burst localizations","New pipeline catches GRBs that Swift used to miss"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The performance metrics come from a sample of bursts that were already found and localized on the ground, so the recovery and doubling rates may not hold for the broader, fainter population of transients that occur during slews.","fun_headline_variants_meta":{"raw":{"variants":["BAT-GLIMPSE recovers gamma-ray bursts during Swift slews","88% of slew-time GRBs recovered by new Swift pipeline","Ground software could double Swift-BAT burst localizations","New pipeline catches GRBs that Swift used to miss"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000424,"raw_usage":{"total_tokens":2073,"prompt_tokens":870,"completion_tokens":1203,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":614,"completion_tokens_details":{"reasoning_tokens":1133}},"tokens_in":614,"tokens_out":1203,"duration_ms":12488,"temperature":1.0,"reasoning_tokens":1133,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T00:04:08.340709+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run BAT-GLIMPSE blindly on the full set of GUANO-triggered BAT time-tagged data from 2020 through 2026, including triggers that did not yield a published arcminute localization, and compare recovered positions with independent detections from other instruments; if the slew recovery rate on that unbiased sample falls well below 88%, or if the overall recovery fraction drops markedly, the central claims would not hold.","supporting_citations":[],"review_version":1}