pith. machine review for the scientific record. sign in

arxiv: 2604.19182 · v1 · submitted 2026-04-21 · 🌌 astro-ph.HE

Recognition: unknown

Identifying Merger-Driven and Collapsar-Driven Gamma-Ray Bursts with Precursor based Solely on Prompt Emission

Si-Yuan Zhu , Pak-Hin Thomas Tam , Fu-Wen Zhang , Hui-Ying Deng , Bing Zhang
Authors on Pith no claims yet

Pith reviewed 2026-05-10 02:08 UTC · model grok-4.3

classification 🌌 astro-ph.HE
keywords gamma-ray burstsprecursor emissionmachine learningt-SNEUMAPprogenitor classificationType I and Type II GRBs
0
0 comments X

The pith

A new prompt-emission index separates merger-driven from collapsar-driven gamma-ray bursts with precursors.

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

Gamma-ray bursts are thought to arise from either compact binary mergers or the collapse of massive stars. Traditional separation by burst duration has exceptions, so this work focuses on those with precursor emission. Using t-SNE and UMAP clustering on Fermi data, distinct groups emerge that match the two origin types, highlighting precursor features. The authors define the EPI as a specific log combination of peak energy and time scales that places most merger events above 6.2 and collapsar events below. This matters because it permits classifying the progenitor using only the initial burst data, guiding faster follow-up observations.

Core claim

The paper shows that unsupervised machine learning applied to prompt emission properties of GRBs with precursors can distinguish Type I (merger, including Type I-L) from Type II (collapsar) events. Precursor emission is a key separator. This inspires the EPI diagnostic, log10(E_p,ME^2 / (T_100,PE * T_100,QE1^{1/2} * T_MVT,PE)), where values above 6.2 correspond to Type I GRBs and below to Type II. The method is validated with Swift observations.

What carries the argument

The E_p,ME-precursor index (EPI), a logarithmic ratio incorporating the main emission's peak energy and durations of the precursor and quiescent phases that acts as a separator between GRB progenitor classes.

If this is right

  • GRBs with EPI > 6.2 are predicted to be merger-driven and associated with kilonovae.
  • GRBs with EPI < 6.2 are predicted to be collapsar-driven and associated with supernovae.
  • The classification relies solely on prompt emission, enabling rapid origin diagnosis.
  • The parameter may apply to GRBs observed by instruments other than Fermi, as suggested by Swift validation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the EPI threshold generalizes, it could help identify more long-duration merger GRBs like Type I-L.
  • Similar machine learning approaches might be tested on GRB samples without requiring precursors.
  • Real-time computation of EPI could be integrated into alert systems for targeted multi-messenger follow-up.

Load-bearing premise

The clusters identified by t-SNE and UMAP accurately reflect the physical distinction between compact binary merger and massive star collapsar progenitors.

What would settle it

Finding a GRB with EPI above 6.2 that is unambiguously associated with a supernova, or one below 6.2 associated with a kilonova, would indicate the index does not reliably separate the classes.

Figures

Figures reproduced from arXiv: 2604.19182 by Bing Zhang, Fu-Wen Zhang, Hui-Ying Deng, Pak-Hin Thomas Tam, Si-Yuan Zhu.

Figure 1
Figure 1. Figure 1: The UMAP and t-SNE maps based on 9 parameters. The red, orange, and blue markers represent the three clusters classified by HDBSCAN. The hollow orange and blue markers show observationally confirmed Type I-L GRBs-KN and Type II GRBs-SN. tion based on the machine learning identifies 35 Type I-S GRBs, 9 Type I-L GRBs, and 323 Type II GRBs. To study the properties of these three classes of GRBs, we depict a c… view at source ↗
Figure 2
Figure 2. Figure 2: The UMAP and t-SNE maps based on 8 parameters beside T100,WE. The red and blue markers represent the two clusters classified by HDBSCAN. sents a subclass of Type I GRBs characterized by both a longer T100,ME and a longer T100,WE. Therefore, we sug￾gest an empirical classification for those Type I GRBs with T100,ME > 2 s and T100,WE > 10 s as Type I-L GRBs, while the remaining are classified as Type I-S GRB… view at source ↗
Figure 3
Figure 3. Figure 3: The P I–Ep,ME plane (left) and the EP I–T90 plane (right). The hollow red circles represent GRB 200522A with a lower limit of Ep at 40 keV. The purple dashed line P I = 10−6.2E 2 p,ME serves as the dividing line corresponding to EP I = 6.2. The shallow purple shaded region represents the range corresponding to the EP I = 6.2±0.2. In the right panel, the histograms along the top and right axes show the dist… view at source ↗
read the original abstract

Gamma-ray bursts (GRBs) are generally classified as Type~I GRBs, which originate from compact binary mergers, and Type~II GRBs, which originate from massive collapsars. The traditional correspondence between short--Type~I GRBs and long--Type~II GRBs, separated by a duration of 2 seconds, has been challenged by recent observations of long GRBs associated with kilonovae (i.e., Type~I-L GRBs) and a short GRB associated with a supernova. In this paper, we focus on GRBs with precursor emission (PE) and compile 366 GRBs detected by Fermi/GBM. Applying the unsupervised machine learning methods t-SNE and UMAP, we are able to distinguish Type~I (including subclass Type~I-L) and Type~II GRBs for the first time and identify PE as a key feature for distinguishing GRBs of different origins. Inspired by results of machine learning, we propose a diagnostic parameter, the $E_{\rm p,ME}$-precursor index ($EPI$), defined as ${\rm log_{10}}(E_{\rm p,ME}^{2}/(T_{\rm 100,PE}T_{\rm 100,QE1}^{1/2}T_{\rm MVT,PE}))$, where most Type~I GRBs have $EPI > 6.2$ and most Type~II GRBs have $EPI < 6.2$. This parameter can help the community to diagnose the origin of any GRB with PE based solely on its prompt emission and rapidly plan for follow-up observations. The validation using Swift GRBs provides illustrative evidence that our method may also be applicable to GRBs observed by instruments other than Fermi.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper compiles a sample of 366 Fermi/GBM GRBs exhibiting precursor emission (PE) and applies unsupervised t-SNE and UMAP dimensionality reduction to prompt-emission parameters. The authors report that the resulting embeddings separate Type I (merger-driven, including Type I-L) from Type II (collapsar-driven) GRBs, with PE identified as a key distinguishing feature. Motivated by these clusters, they define the EPI diagnostic as log10(Ep,ME² / (T100,PE × T100,QE1^{1/2} × TMVT,PE)) and state that most Type I GRBs satisfy EPI > 6.2 while most Type II satisfy EPI < 6.2. The parameter is proposed as a prompt-only classifier to guide follow-up, with an illustrative application to Swift GRBs.

Significance. If the reported separation robustly traces physical progenitor classes and generalizes, the EPI diagnostic would supply a practical, rapid tool for classifying GRBs with precursors using only prompt data, facilitating targeted multi-messenger follow-up. The large PE sample and first application of these embedding methods to this subclass constitute a concrete advance in GRB taxonomy.

major comments (3)
  1. [t-SNE/UMAP results] In the t-SNE/UMAP analysis section, the mapping of unsupervised clusters to Type I versus Type II progenitors is performed without reported quantitative metrics (e.g., silhouette score, adjusted Rand index on any labeled subset, or cross-validation scores). Because the central claim equates the observed embedding separation with physical origin, the absence of these diagnostics leaves the cluster-to-class assignment under-constrained.
  2. [EPI definition] The EPI definition (abstract and diagnostic section) is explicitly “inspired by” the ML clusters and the numerical threshold 6.2 is chosen to separate those same clusters in the Fermi sample. This post-hoc tuning on the discovery data introduces circularity that directly affects the claimed generality of the 6.2 cutoff.
  3. [Swift validation] The Swift validation is labeled “illustrative” with no quantitative success rate, confusion matrix, or error analysis on an independent set of spectroscopically confirmed progenitors. A load-bearing claim of applicability beyond Fermi therefore rests on an unquantified demonstration.
minor comments (2)
  1. [EPI equation] Notation for the EPI components (T100,PE, T100,QE1, TMVT,PE) should be standardized between the abstract, equation, and any tables to avoid subscript ambiguity.
  2. [Figures] The t-SNE/UMAP figures would benefit from explicit overlay of the small number of GRBs with independent progenitor labels to illustrate the cluster-to-class mapping.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We appreciate the referee's thoughtful review and the opportunity to clarify and strengthen our manuscript. Below we address each major comment in turn. We agree with several points and will incorporate revisions accordingly, while providing additional context for others.

read point-by-point responses

  1. Referee: In the t-SNE/UMAP analysis section, the mapping of unsupervised clusters to Type I versus Type II progenitors is performed without reported quantitative metrics (e.g., silhouette score, adjusted Rand index on any labeled subset, or cross-validation scores). Because the central claim equates the observed embedding separation with physical origin, the absence of these diagnostics leaves the cluster-to-class assignment under-constrained.

    Authors: We acknowledge that quantitative validation metrics for the clustering were not included in the original submission. While t-SNE and UMAP are primarily visualization tools and the separation is visually evident in the embeddings, we agree that reporting metrics such as the silhouette score for the labeled data in the embedding space would provide a more rigorous assessment. In the revised manuscript, we will compute and report the silhouette score using the known Type I and Type II labels on the t-SNE and UMAP embeddings. We note that since the dimensionality reduction is unsupervised, cross-validation is not directly applicable, but the silhouette score will quantify the separation. revision: yes

  2. Referee: The EPI definition (abstract and diagnostic section) is explicitly “inspired by” the ML clusters and the numerical threshold 6.2 is chosen to separate those same clusters in the Fermi sample. This post-hoc tuning on the discovery data introduces circularity that directly affects the claimed generality of the 6.2 cutoff.

    Authors: The EPI parameter is constructed from the prompt-emission features that the unsupervised methods identified as most discriminative between the two classes. The threshold of 6.2 is indeed selected to optimally separate the clusters in our Fermi/GBM sample of 366 events. We recognize the potential for circularity in this choice. To address this, we will revise the text to emphasize that the 6.2 threshold is empirical and derived from the current sample, and we will present the distribution of EPI values for Type I and Type II GRBs to allow readers to assess the separation. The Swift application serves as an independent check, though we agree it is limited. revision: partial

  3. Referee: The Swift validation is labeled “illustrative” with no quantitative success rate, confusion matrix, or error analysis on an independent set of spectroscopically confirmed progenitors. A load-bearing claim of applicability beyond Fermi therefore rests on an unquantified demonstration.

    Authors: We agree that the Swift validation is currently presented as illustrative without quantitative metrics. The number of Swift GRBs with precursors and confirmed progenitors is small, limiting the ability to compute robust statistics such as a confusion matrix. In the revision, we will expand the discussion to include a table listing the Swift GRBs analyzed, their EPI values, and the assigned types based on other indicators, and we will compute a simple success rate where possible. However, we maintain that the primary validation is the separation in the large Fermi sample, with Swift providing supporting evidence for broader applicability. revision: yes

Circularity Check

1 steps flagged

EPI diagnostic form and 6.2 threshold fitted to t-SNE/UMAP clusters on the same 366-GRB Fermi sample

specific steps
  1. fitted input called prediction [Abstract]
    "Inspired by results of machine learning, we propose a diagnostic parameter, the $E_{p,ME}$-precursor index ($EPI$), defined as $log_{10}(E_{p,ME}^{2}/(T_{100,PE}T_{100,QE1}^{1/2}T_{MVT,PE}))$, where most Type I GRBs have $EPI > 6.2$ and most Type II GRBs have $EPI < 6.2$."

    The functional form of EPI is chosen to capture the separation found by t-SNE/UMAP on the identical 366-GRB Fermi sample, and the specific 6.2 threshold is selected to separate the clusters after they are labeled as Type I/II. The diagnostic therefore reproduces the input clustering by construction on the data used to discover it.

full rationale

The paper applies unsupervised t-SNE and UMAP to the Fermi PE sample to discover clusters, interprets them as Type I vs Type II, then explicitly constructs the EPI functional form and selects the numerical cutoff to reproduce that separation. This step reduces the claimed diagnostic to a post-hoc fit on the discovery data rather than an independent or first-principles result. The Swift validation is described only as illustrative, with no held-out quantitative test against confirmed progenitors. The core ML clustering itself is not circular, but the load-bearing interpretive mapping and subsequent parameter tuning introduce moderate circularity.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on one fitted cutoff value and the domain assumption that progenitor class is the dominant driver of the observed clustering.

free parameters (1)
  • EPI threshold = 6.2
    Numerical cut of 6.2 chosen to separate most Type I from Type II events in the Fermi sample.
axioms (1)
  • domain assumption GRBs originate either from compact-object mergers (Type I) or from massive-star collapsars (Type II), and these classes produce distinguishable prompt-emission signatures when a precursor is present.
    Invoked to interpret the ML clusters and to motivate the diagnostic.

pith-pipeline@v0.9.0 · 5646 in / 1490 out tokens · 54796 ms · 2026-05-10T02:08:36.319553+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

78 extracted references · 74 canonical work pages · 2 internal anchors

  1. [1]

    2017, PhRvL, 119, 161101, doi: 10.1103/PhysRevLett.119.161101

    Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017, PhRvL, 119, 161101, doi: 10.1103/PhysRevLett.119.161101

    work page doi:10.1103/physrevlett.119.161101 2017

  2. [2]

    P., & Anand, S

    Ahumada, T., Singer, L. P., Anand, S., et al. 2021, Nature Astronomy, 5, 917, doi: 10.1038/s41550-021-01428-7

    work page doi:10.1038/s41550-021-01428-7 2021

  3. [3]

    2002, A&A, 390, 81, doi: 10.1051/0004-6361:20020722

    Amati, L., Frontera, F., Tavani, M., et al. 2002, A&A, 390, 81, doi: 10.1051/0004-6361:20020722 Astropy Collaboration, Price-Whelan, A. M., Lim, P. L., et al. 2022, ApJ, 935, 167, doi: 10.3847/1538-4357/ac7c74

    work page doi:10.1051/0004-6361:20020722 2002

  4. [4]

    Band, D. L. 1997, ApJ, 486, 928, doi: 10.1086/304566

    work page doi:10.1086/304566 1997

  5. [5]

    X., Ma, Y

    Cheng, L. X., Ma, Y. Q., Cheng, K. S., Lu, T., & Zhou, Y. Y. 1995, A&A, 300, 746

    1995

  6. [6]

    2023, ApJL, 954, L29, doi: 10.3847/2041-8213/acf21d

    Dichiara, S., Tsang, D., Troja, E., et al. 2023, ApJL, 954, L29, doi: 10.3847/2041-8213/acf21d

    work page doi:10.3847/2041-8213/acf21d 2023

  7. [7]

    Dimple, Misra, K., & Arun, K. G. 2023, ApJL, 949, L22, doi: 10.3847/2041-8213/acd4c4

    work page doi:10.3847/2041-8213/acd4c4 2023

  8. [8]

    Dimple, Misra, K., & Arun, K. G. 2024, ApJ, 974, 55, doi: 10.3847/1538-4357/ad6d6a

    work page doi:10.3847/1538-4357/ad6d6a 2024

  9. [9]

    , keywords =

    Fong, W., Laskar, T., Rastinejad, J., et al. 2021, ApJ, 906, 127, doi: 10.3847/1538-4357/abc74a

    work page doi:10.3847/1538-4357/abc74a 2021

  10. [10]

    and Dong, Yuxin and Berger, Edo and Paterson, Kerry and Chornock, Ryan and Levan, Andrew and Blanchard, Peter and Alexander, Kate D

    Fong, W.-f., Nugent, A. E., Dong, Y., et al. 2022, ApJ, 940, 56, doi: 10.3847/1538-4357/ac91d0

    work page doi:10.3847/1538-4357/ac91d0 2022

  11. [11]

    J., Vreeswijk, P

    Galama, T. J., Vreeswijk, P. M., van Paradijs, J., et al. 1998, Nature, 395, 670, doi: 10.1038/27150

    work page doi:10.1038/27150 1998

  12. [12]

    Z., & Butler, N

    Golkhou, V. Z., & Butler, N. R. 2014, ApJ, 787, 90, doi: 10.1088/0004-637X/787/1/90

    work page doi:10.1088/0004-637x/787/1/90 2014

  13. [13]

    Z., Butler, N

    Golkhou, V. Z., Butler, N. R., & Littlejohns, O. M. 2015, ApJ, 811, 93, doi: 10.1088/0004-637X/811/2/93

    work page doi:10.1088/0004-637x/811/2/93 2015

  14. [14]

    2018, MNRAS, 479, 588, doi: 10.1093/mnras/sty1462

    Gottlieb, O., Nakar, E., Piran, T., & Hotokezaka, K. 2018, MNRAS, 479, 588, doi: 10.1093/mnras/sty1462

    work page doi:10.1093/mnras/sty1462 2018

  15. [15]

    K., Severin, J

    Jespersen, C. K., Severin, J. B., Steinhardt, C. L., et al. 2020, ApJL, 896, L20, doi: 10.3847/2041-8213/ab964d

    work page doi:10.3847/2041-8213/ab964d 2020

  16. [16]

    1997, ApJ, 490, 92, doi: 10.1086/512791

    Kobayashi, S., Piran, T., & Sari, R. 1997, ApJ, 490, 92, doi: 10.1086/512791

    work page doi:10.1086/512791 1997

  17. [17]

    Kouveliotou,et al., Identification of two classes of gamma-ray bursts.Astrophys

    Kouveliotou, C., Meegan, C. A., Fishman, G. J., et al. 1993, ApJL, 413, L101, doi: 10.1086/186969

    work page doi:10.1086/186969 1993

  18. [18]

    2012, ApJL, 757, L3, doi: 10.1088/2041-8205/757/1/L3 Lemaˆ ıtre, G

    Lai, D. 2012, ApJL, 757, L3, doi: 10.1088/2041-8205/757/1/L3

    work page doi:10.1088/2041-8205/757/1/l3 2012

  19. [19]

    and Gompertz, Benjamin P

    Levan, A. J., Gompertz, B. P., Salafia, O. S., et al. 2024, Nature, 626, 737, doi: 10.1038/s41586-023-06759-1

    work page doi:10.1038/s41586-023-06759-1 2024

  20. [20]

    M., Zhang, Z

    Li, Q. M., Zhang, Z. B., Han, X. L., et al. 2023, MNRAS, 524, 1096, doi: 10.1093/mnras/stad1648

    work page doi:10.1093/mnras/stad1648 2023

  21. [21]

    2020, ApJ, 897, 154, doi: 10.3847/1538-4357/ab96b8 L¨ u, H.-J., Liang, E.-W., Zhang, B.-B., & Zhang, B

    Li, Y., Zhang, B., & Yuan, Q. 2020, ApJ, 897, 154, doi: 10.3847/1538-4357/ab96b8 L¨ u, H.-J., Liang, E.-W., Zhang, B.-B., & Zhang, B. 2010, ApJ, 725, 1965, doi: 10.1088/0004-637X/725/2/1965

    work page doi:10.3847/1538-4357/ab96b8 2020

  22. [22]

    2023, ApJ, 959, 44, doi: 10.3847/1538-4357/ad03ec

    Luo, J.-W., Wang, F.-F., Zhu-Ge, J.-M., et al. 2023, ApJ, 959, 44, doi: 10.3847/1538-4357/ad03ec

    work page doi:10.3847/1538-4357/ad03ec 2023

  23. [23]

    E., et al

    Maccary, R., Guidorzi, C., Camisasca, A. E., et al. 2025, A&A, 702, A95, doi: 10.1051/0004-6361/202555418

    work page doi:10.1051/0004-6361/202555418 2025

  24. [24]

    I., & Woosley, S

    MacFadyen, A. I., & Woosley, S. E. 1999, ApJ, 524, 262, doi: 10.1086/307790

    work page doi:10.1086/307790 1999

  25. [25]

    & Kroupa, P

    MacFadyen, A. I., Woosley, S. E., & Heger, A. 2001, ApJ, 550, 410, doi: 10.1086/319698

    work page doi:10.1086/319698 2001

  26. [26]

    A., Shenoy, A., Sonbas, E., et al

    MacLachlan, G. A., Shenoy, A., Sonbas, E., et al. 2013, MNRAS, 432, 857, doi: 10.1093/mnras/stt241

    work page doi:10.1093/mnras/stt241 2013

  27. [27]

    URLhttps://doi.org/10.21105/joss.00205

    McInnes, L., Healy, J., & Astels, S. 2017, The Journal of Open Source Software, 2, 205, doi: 10.21105/joss.00205

    work page doi:10.21105/joss.00205 2017

  28. [28]

    UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

    McInnes, L., Healy, J., & Melville, J. 2018, arXiv e-prints, arXiv:1802.03426. https://arxiv.org/abs/1802.03426

    work page internal anchor Pith review arXiv 2018

  29. [29]

    2018, The Journal of Open Source Software, 3, 861

    McInnes, L., Healy, J., Saul, N., & Grossberger, L. 2018, The Journal of Open Source Software, 3, 861

    2018

  30. [30]

    N., et al

    Meegan, C., Lichti, G., Bhat, P. N., et al. 2009, ApJ, 702, 791, doi: 10.1088/0004-637X/702/1/791 M´ esz´ aros, P., Ramirez-Ruiz, E., Rees, M. J., & Zhang, B. 2002, ApJ, 578, 812, doi: 10.1086/342611 M´ esz´ aros, P., & Rees, M. J. 2000, ApJ, 530, 292, doi: 10.1086/308371

    work page doi:10.1088/0004-637x/702/1/791 2009

  31. [31]

    J., & Papathanassiou, H

    Meszaros, P., Rees, M. J., & Papathanassiou, H. 1994, ApJ, 432, 181, doi: 10.1086/174559

    work page doi:10.1086/174559 1994

  32. [32]

    Y., & Pozanenko, A

    Minaev, P. Y., & Pozanenko, A. S. 2020, MNRAS, 492, 1919, doi: 10.1093/mnras/stz3611

    work page doi:10.1093/mnras/stz3611 2020

  33. [33]

    & Sari, R.\ 2012, , 747, 2, 88

    Nakar, E., & Sari, R. 2012, ApJ, 747, 88, doi: 10.1088/0004-637X/747/2/88

    work page doi:10.1088/0004-637x/747/2/88 2012

  34. [34]

    , eprint =

    Narayan, R., Paczynski, B., & Piran, T. 1992, ApJL, 395, L83, doi: 10.1086/186493

    work page doi:10.1086/186493 1992

  35. [35]

    Neill, D., Tsang, D., van Eerten, H., Ryan, G., & Newton, W. G. 2022, MNRAS, 514, 5385, doi: 10.1093/mnras/stac1645

    work page doi:10.1093/mnras/stac1645 2022

  36. [36]

    P., Share, G

    Norris, J. P., Share, G. H., Messina, D. C., et al. 1986, ApJ, 301, 213, doi: 10.1086/163889

    work page doi:10.1086/163889 1986

  37. [37]

    The Astrophysical Journal , author =

    Nugent, A. E., Fong, W.-F., Dong, Y., et al. 2022, ApJ, 940, 57, doi: 10.3847/1538-4357/ac91d1 O’Connor, B., Troja, E., Dichiara, S., et al. 2021, MNRAS, 502, 1279, doi: 10.1093/mnras/stab132

    work page doi:10.3847/1538-4357/ac91d1 2022

  38. [38]

    1986, ApJL, 308, L43, doi: 10.1086/184740 Pe’er, A

    Paczynski, B. 1986, ApJL, 308, L43, doi: 10.1086/184740

    work page doi:10.1086/184740 1986

  39. [39]

    2013, PhRvL, 111, 061105, doi: 10.1103/PhysRevLett.111.061105

    Palenzuela, C., Lehner, L., Ponce, M., et al. 2013, PhRvL, 111, 061105, doi: 10.1103/PhysRevLett.111.061105

    work page doi:10.1103/physrevlett.111.061105 2013

  40. [40]

    2011, Journal of Machine Learning Research, 12, 2825

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825

    2011

  41. [41]

    2013, ApJ, 763, 15, doi: 10.1088/0004-637X/763/1/15

    Qin, Y., Liang, E.-W., Liang, Y.-F., et al. 2013, ApJ, 763, 15, doi: 10.1088/0004-637X/763/1/15

    work page doi:10.1088/0004-637x/763/1/15 2013

  42. [42]

    , keywords =

    Ramirez-Ruiz, E., MacFadyen, A. I., & Lazzati, D. 2002, MNRAS, 331, 197, doi: 10.1046/j.1365-8711.2002.05176.x

    work page doi:10.1046/j.1365-8711.2002.05176.x 2002

  43. [43]

    and Gompertz, Benjamin P

    Rastinejad, J. C., Gompertz, B. P., Levan, A. J., et al. 2022, Nature, 612, 223, doi: 10.1038/s41586-022-05390-w 13

    work page doi:10.1038/s41586-022-05390-w 2022

  44. [44]

    J., & Meszaros, P

    Rees, M. J., & Meszaros, P. 1994, ApJL, 430, L93, doi: 10.1086/187446

    work page doi:10.1086/187446 1994

  45. [45]

    D., Norris, J

    Scargle, J. D., Norris, J. P., Jackson, B., & Chiang, J. 2013, ApJ, 764, 167, doi: 10.1088/0004-637X/764/2/167

    work page doi:10.1088/0004-637x/764/2/167 2013

  46. [46]

    2018, A&A, 616, A48, doi: 10.1051/0004-6361/201731475

    Selsing, J., Kr¨ uhler, T., Malesani, D., et al. 2018, A&A, 616, A48, doi: 10.1051/0004-6361/201731475

    work page doi:10.1051/0004-6361/201731475 2018

  47. [47]

    L., Mann, W

    Steinhardt, C. L., Mann, W. J., Rusakov, V., & Jespersen, C. K. 2023, ApJ, 945, 67, doi: 10.3847/1538-4357/acb999

    work page doi:10.3847/1538-4357/acb999 2023

  48. [48]

    G., & Kokkotas, K

    Suvorov, A. G., & Kokkotas, K. D. 2020, PhRvD, 101, 083002, doi: 10.1103/PhysRevD.101.083002

    work page doi:10.1103/physrevd.101.083002 2020

  49. [49]

    2025, ApJ, 993, 89, doi: 10.3847/1538-4357/ae0456

    Tan, W.-J., Wang, C.-W., Zhang, P., et al. 2025, ApJ, 993, 89, doi: 10.3847/1538-4357/ae0456

    work page doi:10.3847/1538-4357/ae0456 2025

  50. [50]

    2013, ApJ, 777, 103, doi: 10.1088/0004-637X/777/2/103

    Tsang, D. 2013, ApJ, 777, 103, doi: 10.1088/0004-637X/777/2/103

    work page doi:10.1088/0004-637x/777/2/103 2013

  51. [51]

    2012, PhRvL, 108, 011102, doi: 10.1103/PhysRevLett.108.011102 Van Der Maaten, L

    Bondarescu, R. 2012, PhRvL, 108, 011102, doi: 10.1103/PhysRevLett.108.011102 Van Der Maaten, L. 2014, Journal of Machine Learning Research, 15, 3221 Van Der Maaten, L., & Hinton, G. 2008, Journal of Machine Learning Research, 9, 2579

    work page doi:10.1103/physrevlett.108.011102 2012

  52. [52]

    N., Burns, E., et al

    Veres, P., Bhat, P. N., Burns, E., et al. 2023, ApJL, 954, L5, doi: 10.3847/2041-8213/ace82d von Kienlin, A., Meegan, C. A., Paciesas, W. S., et al. 2020, ApJ, 893, 46, doi: 10.3847/1538-4357/ab7a18

    work page doi:10.3847/2041-8213/ace82d 2023

  53. [53]

    2025, ApJ, 979, 73, doi: 10.3847/1538-4357/ad98ec

    Wang, C.-W., Tan, W.-J., Xiong, S.-L., et al. 2025, ApJ, 979, 73, doi: 10.3847/1538-4357/ad98ec

    work page doi:10.3847/1538-4357/ad98ec 2025

  54. [54]

    2017, ApJL, 851, L18, doi: 10.3847/2041-8213/aa9e08

    Wang, H., Zhang, F.-W., Wang, Y.-Z., et al. 2017, ApJL, 851, L18, doi: 10.3847/2041-8213/aa9e08

    work page doi:10.3847/2041-8213/aa9e08 2017

  55. [55]

    2018, ApJ, 868, 19, doi: 10.3847/1538-4357/aae531

    Wang, J.-S., Peng, F.-K., Wu, K., & Dai, Z.-G. 2018, ApJ, 868, 19, doi: 10.3847/1538-4357/aae531

    work page doi:10.3847/1538-4357/aae531 2018

  56. [56]

    2020, ApJL, 902, L42, doi: 10.3847/2041-8213/abbfb8

    Zhang, B. 2020, ApJL, 902, L42, doi: 10.3847/2041-8213/abbfb8

    work page doi:10.3847/2041-8213/abbfb8 2020

  57. [57]

    The Deep and Transient Universe in the SVOM Era: New Challenges and Opportunities - Scientific prospects of the SVOM mission

    Wei, J., Cordier, B., Antier, S., et al. 2016, arXiv e-prints, arXiv:1610.06892, doi: 10.48550/arXiv.1610.06892

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1610.06892 2016

  58. [58]

    Woosley, S. E. 1993, ApJ, 405, 273, doi: 10.1086/172359

    work page doi:10.1086/172359 1993

  59. [59]

    E., & Bloom, J

    Woosley, S. E., & Bloom, J. S. 2006, ARA&A, 44, 507, doi: 10.1146/annurev.astro.43.072103.150558

    work page doi:10.1146/annurev.astro.43.072103.150558 2006

  60. [60]

    2023, ApJS, 268, 5, doi: 10.3847/1538-4365/ace77c

    Xiao, S., Yang, J.-J., Luo, X.-H., et al. 2023, ApJS, 268, 5, doi: 10.3847/1538-4365/ace77c

    work page doi:10.3847/1538-4365/ace77c 2023

  61. [61]

    2015, Nature Communications, 6, 7323, doi: 10.1038/ncomms8323

    Yang, B., Jin, Z.-P., Li, X., et al. 2015, Nature Communications, 6, 7323, doi: 10.1038/ncomms8323

    work page doi:10.1038/ncomms8323 2015

  62. [62]

    2022, Nature, 612, 232, doi: 10.1038/s41586-022-05403-8

    Yang, J., Ai, S., Zhang, B.-B., et al. 2022, Nature, 612, 232, doi: 10.1038/s41586-022-05403-8

    work page doi:10.1038/s41586-022-05403-8 2022

  63. [63]

    Nature , author =

    Yang, Y.-H., Troja, E., O’Connor, B., et al. 2024, Nature, 626, 742, doi: 10.1038/s41586-023-06979-5

    work page doi:10.1038/s41586-023-06979-5 2024

  64. [64]

    T., Mellema, G., Pen, U.-L., et al

    Yi, T., Liang, E., Qin, Y., & Lu, R. 2006, MNRAS, 367, 1751, doi: 10.1111/j.1365-2966.2006.10083.x

    work page doi:10.1111/j.1365-2966.2006.10083.x 2006

  65. [65]

    2025, Science China

    Yuan, W., Dai, L., Feng, H., et al. 2025, Science China

    2025

  66. [66]

    Physics, Mechanics, and Astronomy, 68, 239501, doi: 10.1007/s11433-024-2600-3

    work page doi:10.1007/s11433-024-2600-3

  67. [67]

    2006, Nature, 444, 1010, doi: 10.1038/4441010a

    Zhang, B. 2006, Nature, 444, 1010, doi: 10.1038/4441010a

    work page doi:10.1038/4441010a 2006

  68. [68]

    2025, Journal of High Energy Astrophysics, 45, 325, doi: 10.1016/j.jheap.2024.12.013

    Zhang, B. 2025, Journal of High Energy Astrophysics, 45, 325, doi: 10.1016/j.jheap.2024.12.013

    work page doi:10.1016/j.jheap.2024.12.013 2025

  69. [69]

    J., et al

    Zhang, B., Zhang, B.-B., Virgili, F. J., et al. 2009, ApJ, 703, 1696, doi: 10.1088/0004-637X/703/2/1696

    work page doi:10.1088/0004-637x/703/2/1696 2009

  70. [70]

    B., Liu, Z

    Zhang, B. B., Liu, Z. K., Peng, Z. K., et al. 2021, Nature Astronomy, 5, 911, doi: 10.1038/s41550-021-01395-z

    work page doi:10.1038/s41550-021-01395-z 2021

  71. [71]

    2012, ApJ, 750, 88, doi: 10.1088/0004-637X/750/2/88

    Zhang, F.-W., Shao, L., Yan, J.-Z., & Wei, D.-M. 2012, ApJ, 750, 88, doi: 10.1088/0004-637X/750/2/88

    work page doi:10.1088/0004-637x/750/2/88 2012

  72. [72]

    2022, ApJL, 939, L25, doi: 10.3847/2041-8213/ac9b55

    Zhang, Z., Yi, S.-X., Zhang, S.-N., Xiong, S.-L., & Xiao, S. 2022, ApJL, 939, L25, doi: 10.3847/2041-8213/ac9b55

    work page doi:10.3847/2041-8213/ac9b55 2022

  73. [73]

    B., Jiang, M., Zhang, Y., et al

    Zhang, Z. B., Jiang, M., Zhang, Y., et al. 2020, ApJ, 902, 40, doi: 10.3847/1538-4357/abb400

    work page doi:10.3847/1538-4357/abb400 2020

  74. [74]

    Zhu, S.-Y., Deng, H.-Y., Zhang, F.-W., Mo, Q.-Z., & Tam, P.-H. T. 2025a, MNRAS, 541, 3236, doi: 10.1093/mnras/staf954

    work page doi:10.1093/mnras/staf954

  75. [75]

    2023, ApJ, 950, 30, doi: 10.3847/1538-4357/acc83b

    Zhu, S.-Y., Liu, Z.-Y., Shi, Y.-R., et al. 2023, ApJ, 950, 30, doi: 10.3847/1538-4357/acc83b

    work page doi:10.3847/1538-4357/acc83b 2023

  76. [76]

    T., & Zhang, F.-W

    Zhu, S.-Y., Shao, L., Tam, P.-H. T., & Zhang, F.-W. 2025b, A&A, 702, A173, doi: 10.1051/0004-6361/202556274

    work page doi:10.1051/0004-6361/202556274

  77. [77]

    2024, MNRAS, 532, 1434, doi: 10.1093/mnras/stae1594

    Zhu, S.-Y., Sun, W.-P., Ma, D.-L., & Zhang, F.-W. 2024, MNRAS, 532, 1434, doi: 10.1093/mnras/stae1594

    work page doi:10.1093/mnras/stae1594 2024

  78. [78]

    Zhu, S.-Y., & Tam, P.-H. T. 2024, ApJ, 976, 62, doi: 10.3847/1538-4357/ad813a

    work page doi:10.3847/1538-4357/ad813a 2024