{"id":"a1f161ca-1c33-4167-85e7-24363df92ad6","arxiv_id":"2505.05561","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Fitting the luminosity function of high-redshift Swift long bursts and extrapolating to low redshift suggests that 33 to 47 percent of low-redshift long bursts may not be produced by collapsing massive stars.","lead":"The authors reanalyze Swift long gamma-ray burst data and find that a large fraction of low-redshift bursts may come from mergers of compact objects rather than collapsing massive stars. The result matters because gamma-ray bursts are often used as cosmic distance indicators, and a mixed progenitor population would complicate that use.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"60% redshift completeness with a P-only correction θ_z(P) may bias the z≥2 LF and inflate the inferred non-collapsar fraction at z<2.","rationale":"The reader identified the assumption that all z≥2 bursts are collapsars and the extrapolation to z<2 as the weakest point. That is related but distinct from my concern: even if all z≥2 bursts are collapsars, the LF fitted to that subsample can be biased if the redshift measurement probability depends on z or luminosity. The paper's own selection recognizes a low completeness (60%) and introduces a peak-flux-only correction, but this correction cannot capture a z-dependent selection effect. Such a bias would systematically reduce the predicted low-z collapsar count and inflate the inferred non-collapsar excess, directly threatening the central claim. I therefore agree with the CONDITIONAL verdict: the paper's quantitative conclusions should be accepted only after the completeness model is validated or an alternative high-completeness analysis is shown to give similar results. The qualitative point that some low-z LGRBs are non-collapsars is independently supported by GRB 211211A/230307A and prior rate studies, so I do not recommend rejection, but the specific percentages should not yet be used in cosmological applications.","tokens_in":12244,"tokens_out":19952,"duration_ms":224367,"concrete_test":"Re-estimate the redshift completeness from the full Swift sample as a two-dimensional function θ_z(P, z), e.g., by binning all bursts (with and without redshifts) in P and z, or by modeling the follow-up probability with a logistic regression on both P and z. Refit the LF using Eq. (6) with θ_z(P, z) instead of θ_z(P), and recompute the predicted collapsar count at z<2. If the predicted count rises by more than ~15% (i.e., the non-collapsar fraction drops below ~30% or becomes consistent with zero), the paper's central claim is not robust. As a complementary check, redo the analysis using only the higher-completeness BAT6ext subsample (C≈82%) and compare the inferred non-collapsar fraction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim depends on the collapsar LF fitted to the 59 z≥2 bursts. The sample has only 60% redshift completeness, and the correction θ_z(P) = [1 + (1.28±0.21)×(0.95±0.01)^P]^{-1} in Eq. (3)–(6) is assumed to depend solely on peak flux P. If the probability of redshift measurement also depends on z or on luminosity at fixed P—e.g., because high-z events are harder to follow up spectroscopically—then the observed z≥2 subsample is biased toward the most luminous bursts. This bias would flatten the fitted faint-end slope, raise the break luminosity, and lower the efficiency η, because the true number of z≥2 bursts is underrepresented. Extrapolating this biased LF to z<2 (as in Eq. 6 with z_max = 2) would then underestimate the number of collapsar GRBs, artificially increasing the inferred non-collapsar fraction (72.67/57.28 predicted vs 108 observed). The paper never validates the z-independence of θ_z; it only quotes the overall 60% completeness. This is load-bearing because the quantitative conclusion (67.29% or 53.04% collapsar fraction at z<2) is directly proportional to the unbiasedness of the z≥2 LF.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses a Swift sample of 280 long GRBs with peak flux >=2.6 ph cm^-2 s^-1 (60% redshift completeness) to constrain the luminosity function of collapsar GRBs, assuming that all z>=2 LGRBs are collapsars. The authors fit a broken power-law LF with three evolutionary scenarios to the 59 z>=2 bursts using maximum likelihood/MCMC, including a peak-flux-dependent completeness correction theta_z(P). They find that no-evolution is strongly disfavored by AIC/BIC, and that luminosity evolution (delta=1.87) or density evolution (delta=1.10) is required. Extrapolating the best-fit models to z<2 predicts 72.67 or 57.28 collapsar bursts, compared to 108 observed z<2 LGRBs, implying that roughly 33-47% of low-redshift LGRBs have non-collapsar progenitors.","tokens_in":12517,"tokens_out":8230,"duration_ms":87785,"significance":"If the result is robust, it provides quantitative support for a non-collapsar component among low-redshift LGRBs, consistent with GRB 211211A and 230307A, and it would caution against using empirical GRB relations without accounting for progenitor diversity. The analysis is transparent: the likelihood formalism is standard, the MCMC fitting and AIC/BIC comparison are reproducible in principle, the sample is public, and the central assumption is explicitly stated. The predicted low-redshift deficit is a falsifiable number. However, the quantitative claim currently rests on a point prediction without uncertainties and on a completeness correction that is assumed to be redshift-independent, so the significance is not yet established at the level claimed.","major_comments":[{"comment":"The completeness correction theta_z is taken to be a function of peak flux only, with no test of redshift dependence. With only 60% redshift completeness (Section 2), selection effects at fixed P can bias the z>=2 subsample if high-z bursts are systematically harder to follow up; this would flatten the fitted faint-end slope and change the low-z extrapolation. Please validate theta_z(P) against redshift (e.g., via a binned completeness map in P and z) and show how the predicted counts 72.67 and 57.28 change under a z-dependent correction.","section":"Section 3, Eqs. (3)-(6)"},{"comment":"The predicted numbers 72.67 and 57.28 are reported without uncertainties. These are functions of the MCMC posterior for (eta, a, b, log Lc, delta); without a posterior or confidence interval for N_exp(0<z<2), the comparison to 108 observed bursts is unquantified and the conclusion of a 'substantial' non-collapsar fraction cannot be assessed for statistical significance. Please propagate the MCMC posterior through Eq. (6) and report the full distribution or at least an uncertainty interval.","section":"Section 4, Table 1"},{"comment":"The assumption that all z>=2 LGRBs are collapsars is load-bearing; the only support offered is that the high-z rate approximately tracks the SFR, which is itself subject to the same selection effects. Because the LF fitted to z>=2 is extrapolated to z<2, any non-collapsar contamination at z>=2 directly biases the inferred low-z non-collapsar fraction. Please add a robustness test, e.g., varying the high-z threshold (z>=2.5 or z>=3) or allowing a nuisance fraction of non-collapsars at z>=2, and state the resulting range of predicted low-z collapsar counts.","section":"Section 2 and Abstract"},{"comment":"The functional form and fitting procedure for theta_z(P) are described only in words, not documented in detail. Since theta_z(P) enters the likelihood and the N_exp prediction directly, the paper should provide the data used, the fitting method, the goodness of fit, and a figure or table showing the completeness as a function of P. It should also clarify whether theta_z(P) was fitted to the same 280-burst sample used for the LF fit, and if so, discuss any potential circularity in the effective exposure.","section":"Section 3, Eq. (3)"}],"minor_comments":[{"comment":"The AIC and BIC comparisons are reported for the high-z fit, but Figures 2 and 3 show comparisons to all 167 redshift-known bursts; please clarify the role of the low-z data in the model evaluation and whether the plotted model curves include the theta_z(P) correction.","section":"Section 4, Figures 2 and 3"},{"comment":"The caption says the shaded area indicates Poisson errors, but the error bars are not defined; please state the confidence level and how the Poisson errors were computed.","section":"Figure 1 caption"},{"comment":"The Akaike weight expression is fine, but the two-model comparison should state explicitly that the resulting probability is relative only to the two models considered, not an absolute probability of correctness.","section":"Section 4, Eq. (8)"},{"comment":"The text says the findings are in good agreement with Petrosian & Dainotti (2024), who found approximately 60% non-collapsars, but this paper's own estimates are 33-47%; please rephrase to reflect the partial overlap and the different model assumptions.","section":"Section 5"},{"comment":"The organization into three redshift groups is clear, but it would help to include a column with T90 values or a note describing how the LGRB selection (T90 > 2 s) was applied to the listed bursts.","section":"Table 2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's central claim is timely and of interest, and the analysis is transparent, but the two load-bearing issues (uncertainty on the predicted counts and the potentially redshift-dependent completeness correction) must be resolved before publication. I do not recommend rejection, as the direction of the result is plausible and the requested checks are feasible within the scope of a revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The quick take: this is a competent, standard luminosity-function analysis on a larger Swift sample that supports the emerging view that a non-trivial fraction of low-redshift long GRBs are not collapsars. The headline percentages are not yet something I would put in a review or a cosmology paper, but the paper deserves a serious referee.\n\nWhat is new: the authors revisit the selection criteria and build a 280-burst sample with peak flux above 2.6 ph cm^-2 s^-1, about 2.8 times larger than BAT6ext. They then fit a broken-power-law LF to the 59 z>=2 bursts under no-evolution, luminosity-evolution, and density-evolution models. The no-evolution model is cleanly rejected by AIC and BIC. The evolution parameters are consistent with earlier work. The larger sample and the explicit theta_z(P) completeness correction are genuine improvements, as is the transparency: full sample table, residual plots, and standard maximum-likelihood machinery.\n\nSoft spots, in order of importance. First, the two acceptable models give 72.67 versus 57.28 expected collapsars at z<2, i.e. 33% versus 47% non-collapsar fractions. Those numbers are reported without error bars even though the MCMC chains exist; the model-to-model spread alone is large enough that quoting one fraction overstates the precision. Second, the completeness correction theta_z(P) is assumed to depend only on peak flux. The paper never tests whether the probability of getting a redshift also depends on z or luminosity at fixed P. If high-z bursts are systematically harder to follow up, which is the usual worry, the fitted z>=2 LF is biased bright, and the extrapolation to z<2 will undercount collapsars and inflate the inferred non-collapsar fraction. That is a load-bearing assumption and it is not validated. Third, the whole exercise assumes all z>=2 LGRBs are collapsars. That is stated clearly, so it is not a hidden flaw, but it is a real one: contamination at high z would move the fitted LF and the downstream counts. These caveats weaken the exact percentages, not the qualitative conclusion.\n\nWho should read it: people working on GRB rates, progenitor channels, and the use of LGRBs as star-formation tracers. The paper should go to peer review. A referee should ask for propagated uncertainties on the predicted counts and for a check of the z-dependence of theta_z, but the analysis is coherent and the conclusion is broadly consistent with independent work.","headline":"A competent LF analysis on a larger Swift sample that reinforces the non-collapsar story at z<2; the exact fractions are model-dependent and lack error bars, so treat them as indicative.","tokens_in":13088,"tokens_out":4882,"would_cite":true,"duration_ms":54040,"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":"The paper argues that one-third to nearly half of low-redshift long gamma-ray bursts likely come from non-collapsar progenitors.","keywords":["long gamma-ray bursts","collapsars","luminosity function","redshift evolution","star formation rate","Swift","GRB progenitors","multiple progenitor scenarios"],"falsifier":"Perform a systematic census of supernova associations versus kilonova/merger signatures for the 280 Swift LGRBs with $P\\ge 2.6\\,\\mathrm{ph\\,cm^{-2}\\,s^{-1}}$: if the spectroscopically confirmed non-collapsar fraction among $z<2$ bursts is much smaller than 33-47%, the extrapolated collapsar luminosity function is wrong. A second test is to check whether the sample's $z\\ge 2$ bursts contain identifiable non-collapsars; even a few would break the calibration anchor and change the predicted low-z counts.","tokens_in":12016,"feed_emoji":"💥","tokens_out":10296,"duration_ms":87968,"temperature":0.7,"pith_summary":"This paper asks whether all long gamma-ray bursts really come from the collapse of massive stars (collapsars). It builds a new Swift sample of 280 bursts with peak flux above $2.6\\,\\mathrm{ph\\,cm^{-2}\\,s^{-1}}$, fits a broken power-law luminosity function to the 59 bursts at $z\\ge 2$ under the assumption that those are all collapsars, and extrapolates the fit to $z<2$ in three evolutionary scenarios. The no-evolution scenario is ruled out; strong luminosity evolution ($\\delta = 1.87^{+0.27}_{-0.31}$) or density evolution ($\\delta = 1.10^{+0.21}_{-0.20}$) is required. Extrapolating to low redshift predicts only 67.29% (luminosity evolution) or 53.04% (density evolution) of the observed $z<2$ bursts as collapsars, implying that roughly one-third to nearly one-half of low-redshift long GRBs have other progenitors. If correct, empirical GRB relations calibrated on a single collapsar population would mix two classes, biasing their use in cosmology.","feed_headline":"A third to half of low-redshift long GRBs may not be collapsars","feed_subtitle":"Fitting Swift burst luminosity to star formation leaves one-third to nearly half of z<2 long bursts unexplained.","key_machinery":"The load-bearing object is a broken power-law luminosity function $\\phi(L,z)$ combined with a star-formation-rate-proportional event rate $\\psi(z)=\\eta\\psi_*(z)$, where $\\psi_*(z)$ is the Hopkins & Beacom / Li star formation rate. An extra factor $(1+z)^{\\delta}$ is inserted either into the break luminosity $L_c(z)=L_{c,0}(1+z)^{\\delta}$ (luminosity evolution) or into the rate $\\psi(z)=\\eta\\psi_*(z)(1+z)^{\\delta}$ (density evolution). A maximum-likelihood fit to the 59 bursts at $z\\ge 2$ fixes the parameters, and the same integral over $0<z<2$, with a peak-flux detection efficiency and the luminosity threshold from a Band-function spectrum, gives the expected collapsar count at low redshift. The new sample itself is built by maximizing $F=N\\times C^3$, balancing sample size against redshift completeness.","core_discovery":"The paper's central claim is that the redshift and luminosity distributions of Swift long GRBs cannot be explained by a collapsar-only population with a non-evolving luminosity function. Fitting the $z\\ge 2$ bursts and extrapolating to $z<2$, the luminosity-evolution model predicts 72.67 collapsar GRBs with $z<2$ and $P\\ge 2.6\\,\\mathrm{ph\\,cm^{-2}\\,s^{-1}}$, which is 67.29% of the observed number; the density-evolution model predicts 57.28, or 53.04%. The paper concludes that a substantial fraction of low-redshift LGRBs are not collapsars, consistent with kilonova-associated long bursts such as 211211A and 230307A, and that the mixture challenges the universality of empirical GRB correlations used for cosmological applications.","pith_inferences":["If the non-collapsar fraction is real, its redshift dependence in this sample could be used to measure the delay-time distribution of the alternative progenitors by fitting the z<2 excess with a delayed star-formation kernel.","The calibration anchor itself is testable: if a non-negligible share of z>=2 bursts turn out to be non-collapsars (e.g., from host-galaxy or supernova signatures), the fitted evolution delta is overestimated and the inferred low-z non-collapsar fraction would need revision.","A direct census of supernova versus kilonova associations in the 280-burst sample would provide an independent check: finding merger-like signatures in roughly a third to a half of z<2 bursts would confirm the prediction, while finding almost none would point to a problem in the luminosity-function extrapolation."],"forward_implications":["The no-evolution model is rejected at a level that matters: its BIC is worse by 8.12 and its Akaike weight relative to luminosity evolution is 0.001, so a collapsar-only explanation needs strong redshift evolution.","Under either viable model, z<2 Swift LGRBs contain roughly 33% to 47% non-collapsars, making non-collapsar progenitors a common rather than rare channel.","Empirical GRB luminosity relations calibrated on the assumption of a single collapsar population would be polluted by non-collapsar events, weakening distance estimates and cosmological parameter inference.","The low-luminosity end of the observed luminosity distribution lies above the collapsar prediction, supporting the view that high- and low-luminosity LGRBs have different progenitors.","The reported triple power-law shape of the LGRB luminosity function can be interpreted as two overlapping broken power laws: bright collapsars plus fainter non-collapsars."],"supporting_citations":[{"why":"Defines the high-completeness criteria that the new selection procedure starts from and then relaxes.","marker":"Jakobsson et al. (2006)"},{"why":"Provides the earlier 58-burst complete sample and the five selection criteria whose declining completeness motivates the new sample.","marker":"Salvaterra et al. (2012)"},{"why":"Provides the BAT6ext sample, the benchmark that the new 280-burst sample is 2.8 times larger than.","marker":"Pescalli et al. (2016)"},{"why":"Supplies the star-formation-rate parametrization that sets the expected collapsar rate.","marker":"Hopkins & Beacom (2006)"},{"why":"Supplies the specific SFR functional form used in the rate model.","marker":"Li (2008)"},{"why":"Supplies the photon spectrum model used to convert peak flux into the luminosity threshold.","marker":"Band et al. (1993)"},{"why":"Reported that over 60% of z<2 LGRBs may be non-collapsars, the comparison that the paper's predicted fractions agree with.","marker":"Petrosian & Dainotti (2024)"},{"why":"Showed high-luminosity LGRBs track the SFR while low-luminosity ones do not, supporting distinct progenitors and matching the paper's luminosity-distribution result.","marker":"Dong et al. (2023)"},{"why":"Supplies the broken-power-law LF fitting framework and the detection-efficiency treatment used in the maximum-likelihood analysis.","marker":"Lan et al. (2021)"}],"fun_headline_variants":["A third to half of nearby long GRBs may not be collapsars","Swift data hint many low-z long GRBs are not collapsars","Collapsar model falls short for many low-redshift long GRBs","Low-redshift long GRBs: up to half may be non-collapsars"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that every long GRB at $z\\ge 2$ in the sample is a collapsar and that the luminosity function fitted to those bursts, together with a star-formation-rate-proportional rate and a $(1+z)^{\\delta}$ evolution term, remains valid when extrapolated to $z<2$.","fun_headline_variants_meta":{"raw":{"variants":["A third to half of nearby long GRBs may not be collapsars","Swift data hint many low-z long GRBs are not collapsars","Collapsar model falls short for many low-redshift long GRBs","Low-redshift long GRBs: up to half may be non-collapsars"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000817,"raw_usage":{"total_tokens":3605,"prompt_tokens":997,"completion_tokens":2608,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":613,"completion_tokens_details":{"reasoning_tokens":2526}},"tokens_in":613,"tokens_out":2608,"duration_ms":20184,"temperature":1.0,"reasoning_tokens":2526,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:02:01.577748+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Perform a systematic census of supernova associations versus kilonova/merger signatures for the 280 Swift LGRBs with $P\\ge 2.6\\,\\mathrm{ph\\,cm^{-2}\\,s^{-1}}$: if the spectroscopically confirmed non-collapsar fraction among $z<2$ bursts is much smaller than 33-47%, the extrapolated collapsar luminosity function is wrong. A second test is to check whether the sample's $z\\ge 2$ bursts contain identifiable non-collapsars; even a few would break the calibration anchor and change the predicted low-z counts.","supporting_citations":[],"review_version":1}