Pith. sign in

REVIEW 4 major objections 6 minor 73 references

A gated recurrent unit trained on simulated galaxy clusters can recover true total mass profiles from X-ray gas profiles alone, with ~5% scatter, and indicates hydrostatic masses of massive observed clusters are biased low by about 10%.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review

2026-08-01 09:32 UTC pith:ZCU7UXP5

load-bearing objection Solid simulation-side study of GRU mass profiles; the X-COP hydrostatic-bias claim is suggestive but not yet validated. the 4 major comments →

arxiv 2607.20721 v1 pith:ZCU7UXP5 submitted 2026-07-22 astro-ph.CO

Probing the baryonic--dark matter connection in galaxy clusters using X-rays with gated recurrent unit neural networks

Asif Iqbal , Subhabrata Majumdar , Weiguang Cui , Elena Rasia , Gabriel W. Pratt , Daniel de Andres This is my paper
classification astro-ph.CO
keywords galaxy clustersmass estimationhydrostatic equilibrium biasgated recurrent unitintracluster mediumX-ray observationsmachine learningcosmology
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper claims that the radial temperature, density, pressure, and enclosed gas mass profiles of a galaxy cluster's hot gas encode the cluster's true total mass profile well enough that a sequence-learning neural network can recover it without assuming hydrostatic equilibrium. Trained on two hydrodynamical simulation suites and tested on held-out simulated clusters, the model achieves unbiased mass profiles with roughly 5% scatter over most of the cluster, beating conventional hydrostatic estimates that show radially growing bias. Applied to X-ray observations of the REXCESS and X-COP samples, the model finds hydrostatic masses are on average consistent with its predictions for the lower-mass REXCESS clusters but about 10% lower for the more massive X-COP clusters, suggesting a mass-dependent hydrostatic bias. If right, this offers a data-driven route to cluster mass calibration independent of the hydrostatic assumption, with implications for cluster cosmology.

Core claim

The central claim is that the three-dimensional total matter mass profile M(<R) of a cluster is learnable from its spherically averaged intracluster medium profiles (temperature, density, pressure, and enclosed gas mass) treated as a sequence. A two-layer bidirectional GRU, trained with Gaussian negative-log-likelihood and an ensemble of 50 initialisations, predicts log10 M(<R) at each radius with a median fractional dispersion of about 5% across most of the cluster region and no systematic bias, both for relaxed and disturbed clusters and for observation-like coarse radial binning. Enclosed gas mass is the dominant predictor, with pressure and temperature shaping the radial structure. When

What carries the argument

The key object is a gated recurrent unit (GRU) — a recurrent neural network that processes the radial ICM profiles as ordered sequences, with update and reset gates that let it retain dependencies across radii while accepting variable-length inputs. The network is conditioned on redshift and outputs mean and log-variance of a log-normal total-mass distribution at each radius; an ensemble of 50 models provides calibrated, radius-dependent uncertainties. The GRU is the mechanism that learns the nonlinear mapping from observable thermal profiles to the unobservable total mass profile, replacing the hydrostatic equilibrium equation.

Load-bearing premise

The network's training labels are simulation truth, so the entire method assumes the simulated relation between hot-gas profiles and total mass is the same as the real one; if real clusters' gas behaves differently from the simulations, both the predicted masses and the claimed hydrostatic bias reflect simulation bias, not reality.

What would settle it

Take the same REXCESS and X-COP clusters with measured weak-lensing mass profiles. If the lensing masses agree with the hydrostatic masses rather than the GRU predictions, or scatter symmetrically between the two, then the simulation-to-observation transfer assumed by the model fails and the claimed ~10% hydrostatic bias is a simulation artefact.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • Cluster mass measurements no longer need to assume hydrostatic equilibrium; the network learns the mapping directly from simulated gas–mass relationships.
  • The model's sparsity estimates (S200,500 and S500,2500) scatter at 0.025–0.060 dex versus 0.13–0.19 dex for hydrostatic masses, sharpening a cosmological probe.
  • The observed X-COP offset suggests hydrostatic masses of massive clusters are systematically low by roughly 10%, which would propagate into X-ray mass–observable scaling relations.
  • Because the method uses only spherically averaged gas profiles, it can be extended to Sunyaev–Zel'dovich and optical data, broadening the samples whose masses can be calibrated without HSE.

Where Pith is reading between the lines

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

  • If the GRU predictions are unbiased on real clusters, the X-COP offset implies the missing hydrostatic support in massive clusters is comparable to the non-thermal pressure fractions predicted by simulations; a direct weak-lensing comparison on the same clusters could confirm or refute this.
  • The model could be used as a diagnostic of baryonic physics: clusters whose learned mass profiles disagree strongly with hydrostatic estimates may be the ones with the most extreme feedback or dynamical disturbance, effectively separating equilibrium from non-equilibrium systems without a dynamical-state classifier.
  • A strong test would be to train on a third simulation suite with substantially different feedback prescriptions and check whether the GRU-vs-HSE offset for X-COP is stable; if it varies with the training physics, the observed offset is simulation-driven rather than evidence about real clusters.
  • Applying the same GRU to the CHEX-MATE sample, analysed uniformly, would test whether the mass-dependent offset persists when REXCESS and X-COP are re-reduced with identical procedures.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper presents a GRU-based deep-learning framework that maps spherically averaged ICM radial profiles (temperature, gas density, pressure, enclosed gas mass, and radius) to total 3D cumulative mass profiles M(<R) of galaxy clusters. The model is trained on The Three Hundred Project simulations (GIZMO-SIMBA and GADGET-X) with a region-based train/test split, data augmentation to mimic observational binning, redshift conditioning, an uncertainty-aware NLL loss, and a 50-model ensemble. The authors report ~5% scatter over most of the cluster region, ~17% in the core, robustness to dynamical state and radial binning, and unbiased cross-simulation performance when trained jointly on the two simulation suites. SHAP analysis identifies enclosed gas mass as the dominant feature. The model is then applied to REXCESS and X-COP X-ray samples, finding HSE masses systematically lower than GRU predictions by about 10% for X-COP, while REXCESS shows near-zero average offset.

Significance. If the simulation-side results are taken at face value, this is a useful contribution: a data-driven mass-profile estimator with well-calibrated uncertainties, tested on held-out regions and across two independent hydrodynamical codes. The region-based split, the MC uncertainty treatment, the robustness tests for dynamical state and binning, and the explicit comparison with HSE masses are all strengths. The cross-simulation generalization test is particularly valuable. However, the observational claim of a ~10% HSE mass bias in X-COP is not secured by the evidence presented, and the paper itself concedes that a uniform reanalysis is needed. The central scientific advance is therefore the simulation methodology rather than the observational bias measurement. Reproducibility is limited because the model is only 'available upon request'.

major comments (4)
  1. [§5 (Fig. 11; Tables F.1/F.2)] The abstract's claim that HSE masses are systematically lower by ~10% for X-COP rests on an unvalidated simulation-to-observation transfer. Joint training on GIZMO-SIMBA and GADGET-X does not validate this transfer because the two suites share the same MDPL2 parent N-body simulation and initial conditions. Figure 6 shows that single-suite training already produces 5–10% cross-suite bias, the same magnitude as the claimed HSE offset. The GNN comparison in Fig. F.3 is not independent because that GNN was trained on GADGET-X. The Planck SZ masses tabulated in Tables F.1/F.2 are not used as a benchmark. The paper's own Sec. 6 states that REXCESS and X-COP were analyzed with different pipelines and that the result 'will therefore need to be checked against a sample with sufficient mass leverage, and which has been subjected to a uniform analysis procedure.' The observational offset should be
  2. [§4.6 (Fig. 10, Eq. 4)] The enclosed gas mass M_g(<R) is a cumulative integral, and the target M(<R) is also cumulative; the two-feature model (radius + M_g) nearly matches the five-feature model. Since the gas fraction varies slowly with radius and mass, M_g is essentially a rescaled proxy for total mass in the simulations. The SHAP dominance of M_g is therefore unsurprising and does not by itself reveal a physical baryon–dark matter connection. The sentence in §4.6 claiming that this 'reflects its fundamental role... rather than the presence of circular or redundant information' is asserted rather than demonstrated. This does not invalidate the supervised prediction, but the transfer of this feature to real clusters requires that the simulated gas-fraction–mass relation matches reality. Please quantify the sensitivity of the predicted M(<R) to assumed gas fraction and temper the physical interpretation.
  3. [§3.5 and §4.1] The hyperparameter optimization description says Optuna minimizes 'test loss'. If the same test split used for reporting headline errors is used for hyperparameter selection, the test-set residuals in Fig. 4 are not fully independent. Please clarify whether an inner validation split was used. Relatedly, §3.4 states that data augmentation is applied to both training and testing samples; if multiple augmented copies of the same cluster enter the test set, the effective number of independent test objects is reduced. The paper should report metrics on the unaugmented test sample as the primary evaluation or demonstrate that augmentation does not inflate apparent performance.
  4. [§5 and §6] The training data are 3D spherically averaged simulation profiles, while the observational REXCESS/X-COP profiles are deprojected under spherical symmetry assumptions from X-ray surface brightness and spectroscopy. No test is provided that these two representations are statistically equivalent. In addition, REXCESS and X-COP were reduced with different pipelines, so the mass-dependent offset seen in Fig. 11 could be an artifact of differing analysis procedures. The manuscript should either include a deprojection systematics test (e.g., forward-modeling simulated 3D profiles into observed-like 2D data and redeprojecting) or explicitly limit the abstract and conclusions to the simulation-validated part of the work.
minor comments (6)
  1. [Table F.1] RXC J1044.5-0704 appears twice with different redshifts and masses; this is likely a duplicate/typo and should be corrected.
  2. [§4.4] Typo: 'GADGET-X and and GIZMO-SIMBA' — duplicate 'and'.
  3. [Abstract vs. §6] The abstract says HSE masses are 'systematically lower' for X-COP, while §6 says 'exceed them by up to ∼10%'. Please harmonize the wording to avoid overstating the effect.
  4. [Appendix E captions] Captions for Figs. E.2 and E.3 state 'up to 0.75 R500', while §4.3 describes 14 bins up to R500 and 7 bins up to 0.75 R500. The captions are inconsistent with the text and with each other.
  5. [§3.3, §3.5] The notation for logarithms is confusing: the paper declares log = natural log, but Eq. (6) and the loss use log10 target values. Please define the base unambiguously at each use.
  6. [Data Availability] The model is 'available upon request'. For reproducibility in a machine-learning paper, please release code, trained weights, and configuration files under an open license, or provide a clear reason for withholding them.

Circularity Check

0 steps flagged

No significant circularity: held-out simulation test is a real prediction, and the observational HSE-bias claim is explicitly tentative.

full rationale

The central mass-profile result is a genuine supervised prediction. Section 3.4 splits the 324 zoom-in regions into 274 training and 50 test regions, with test regions reserved exclusively for final evaluation, and Section 4.1 compares predictions with ground-truth simulated mass profiles. The target M(<R) includes dark matter and is not computed from the input features, so the model must learn a physical ICM-to-mass mapping. The high SHAP importance of enclosed gas mass and the near-equivalent 2-feature model (Sec. 4.6, Fig. 10) reflect the slowly varying gas fraction; this is a learned scaling relation rather than a definitional identity, and the paper explicitly argues against 'circular or redundant information.' No fitted parameter is renamed as a prediction. The observational HSE-bias claim is presented with an explicit transfer caveat: Sec. 6 states the result 'will therefore need to be checked against a sample with sufficient mass leverage, and which has been subjected to a uniform analysis procedure,' so it is not put forward as a closed derivation. The GNN cross-check (Appendix F, Fig. F.3) is a self-citation from overlapping authors and is not an independent benchmark, but it is auxiliary rather than load-bearing for the paper's central claims; under the review criteria this is a minor non-circularity issue, hence score 1 rather than 0.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 0 invented entities

The ledger is dominated by training choices and the simulation-to-observation transfer assumption. No new physical entities, particles, forces, or dimensions are introduced. The most consequential unverified premise is that the simulated ICM–mass relation transfers to real X-ray clusters.

free parameters (5)
  • GRU network weights (W_u, W_r, W_h, biases, redshift projection, output head) = not released
    All predictive power comes from weights fitted by minimizing the NLL loss on ~16,495 training clusters. The trained weights are not public, so the exact fitted mapping cannot be inspected.
  • GRU hyperparameters = lr=7e-4, weight decay=1e-3, hidden dim=120, dropout=0.2, batch size=128, 2 bidirectional layers
    Selected with Optuna by 'minimising test loss' (Sec. 3.5). Using the test set for model selection means reported test scatter may be optimistic.
  • Pressure smoothing parameter f for HSE baseline = 0.10–0.20, changing from outer to inner regions
    Chosen manually for the local linear weighted regression used to compute hydrostatic masses. It affects the HSE baseline the ML model is compared against (Sec. 2.3).
  • Data augmentation ranges = inner radius drawn from [0.05,0.3] R500, outer from [0.75,2] R500, 7–40 radial bins
    Hand-chosen to mimic observational sampling. Reported robustness to resolution depends on these ranges (Sec. 3.4).
  • Ensemble combination weights = inversely proportional to validation loss
    The 50-model ensemble is combined with weights fit to validation loss (Eq. 7), a minor but additional fitted ingredient.
axioms (6)
  • domain assumption The Three Hundred GIZMO-SIMBA and GADGET-X simulations represent the real cluster ICM–total-mass relation closely enough for transfer to observations.
    Training labels are simulation truth; if the simulated baryon–dark matter connection is biased, all predicted mass profiles and the inferred HSE bias inherit that bias (Sec. 2.1, Sec. 5).
  • domain assumption Deprojected observational X-ray profiles (REXCESS/X-COP) are comparable to the 3D spherically averaged simulated profiles used in training.
    The model is trained on 3D simulation profiles; the observed profiles must be statistically equivalent after deprojection for the transfer to be valid (Sec. 5).
  • domain assumption Cumulative enclosed gas mass M_g(<R) is a valid observable input and is not itself constructed from the total mass being predicted.
    M_g is derived from the gas density profile and used as a predictor; its strong correlation with total mass is what the model exploits, as shown by the two-feature model (Eq. 4, Sec. 4.6).
  • ad hoc to paper A Gaussian NLL loss in log-mass space is an appropriate statistical model for mass-profile prediction.
    The log-normal likelihood (Eq. 9) is a modeling choice; the reported uncertainty calibration depends on this assumption.
  • standard math The hydrostatic equilibrium equation is a legitimate baseline for comparison.
    Used to define the conventional HSE mass and its bias; not central to the ML method itself (Eq. 2).
  • standard math Planck 2016 flat ΛCDM cosmology.
    Sets H0=67.77 km/s/Mpc, Ωm=0.30, ΩΛ=0.70 for mass definitions and overdensity radii.

reviewed 2026-08-01 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Probing the baryonic--dark matter connection in galaxy clusters using X-rays with gated recurrent unit neural networks." pith.science (2026). https://pith.science/paper/ZCU7UXP5

@misc{pith2026260720721,
  author       = {Pith},
  title        = {Pith review of: Probing the baryonic--dark matter connection in galaxy clusters using X-rays with gated recurrent unit neural networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZCU7UXP5}},
  note         = {Machine review of arXiv:2607.20721}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Accurate cluster mass measurements are crucial for cosmology, yet conventional hydrostatic equilibrium (HSE) methods can suffer from systematic biases, particularly in dynamically disturbed systems. We present a gated recurrent unit (GRU) based deep learning framework for predicting three-dimensional mass profiles of galaxy clusters from spherically averaged intra-cluster medium (ICM) radial profiles. By treating ICM profiles as sequential data, the GRU captures radial dependencies and naturally handles profiles with different radial samplings. We train and validate the model using high-resolution hydrodynamical simulations from The Three Hundred Project, achieving unbiased mass predictions with a typical 1$\sigma$ scatter of $\sim$5% over most of the cluster region, significantly improving upon HSE estimates. The model provides radius-dependent uncertainty estimates and remains robust against variations in data quality and cluster morphology. When trained jointly on independent simulation suites (GIZMO-SIMBA and GADGET-X), it successfully generalises across both simulations. Feature importance analysis shows that enclosed gas mass is the dominant predictor, with pressure and temperature providing additional information on the radial mass distribution. We further apply the GRU model to X-ray observations of the REXCESS and X-COP cluster samples from XMM-Newton and compare the inferred mass profiles with HSE estimates. The HSE masses are systematically lower than the GRU predictions for the higher-mass X-COP sample, while the REXCESS sample shows mass differences that are close to zero on average. This work provides a data-driven framework for cluster mass inference that bridges simulations and observations and can be extended to multi-wavelength datasets, including Sunyaev-Zel'dovich and optical observations.

Figures

Figures reproduced from arXiv: 2607.20721 by Asif Iqbal, Daniel de Andres, Elena Rasia, Gabriel W. Pratt, Subhabrata Majumdar, Weiguang Cui.

Figure 1
Figure 1. Figure 1: The distribution of galaxy clusters in The Three Hundred Project as a function of true mass MTrue 500 for the GIZMO-SIMBA and GADGET-X simulation suites in the redshift range 0 ≲ z ≲ 1, contain￾ing 9997 and 9590 clusters, respectively. tion merger trees as well as other companion objects present in the resimulated region. For each simulated cluster, we ex￾tracted spherically averaged three dimensional prof… view at source ↗
Figure 3
Figure 3. Figure 3: Cluster sparsity derived from hydrostatic mass profiles, SHSE, compared with the corresponding true sparsity values for the overdensity pairs STrue 500,2500 (left) and STrue 200,500 (right), using a randomly selected sample of 3000 galaxy clusters from the GIZMO-SIMBA simulations. Points are color-coded by the dynamical state parameter χ, with lower sparsity values associated with more relaxed systems (sma… view at source ↗
Figure 4
Figure 4. Figure 4: GRU mass profile reconstructions and uncertainty estimates as a function of normalised radius. The left panel shows the GRU mass profile reconstructions for 50 randomly selected clusters from the test sample, while the lower sub-panel displays the corresponding fractional residuals. The blue line and blue shaded region represent, respectively, the median and the 84th − 16th percentile range of the fraction… view at source ↗
Figure 5
Figure 5. Figure 5: Median fractional residuals of the mass profiles reconstructed with the GRU model for GIZMO-SIMBA test-sample clusters [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of GRU predicted and HSE masses for the test sample as a function of normalised radius. The solid lines represent the median fraction residuals and the shaded regions indicate the 84th −16th percentile interval. The GRU derived masses are largely unbiased, with a scatter of ∼5%. In contrast, HSE masses exhibit a radially increas￾ing bias, reaching ∼23% at 2 R500 for the GADGET-X sample, while th… view at source ↗
Figure 8
Figure 8. Figure 8: GRU derived cluster sparsity for the GIZMO-SIMBA test sample. The GRU predicts SGRU 200,500 and SGRU 500,2500 with scatters of 0.027 and 0.063 dex, respectively, with the larger scatter at smaller radii reflecting the more complex inner cluster structure. R ng Tg Pg Mg 0.00 0.05 0.10 0.15 0.20 0.25 0.30 S H A P f e a t u r e im p o r t a n c e [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of the performance of the fiducial 5-feature model and the reduced 2-feature model (radius and integrated gas mass) is pre￾sented in terms of the median fractional residual and the 84th − 16th percentile range. The inset shows the distribution of fractional residuals in M500 for both models, with vertical dashed lines marking the cor￾responding medians. Horizontal dotted black lines represent ±… view at source ↗
Figure 11
Figure 11. Figure 11: Cumulative mass profile reconstruction for the REXCESS sample (left panel) and the X-COP sample (right panel) using the GRU model. For clarity, the error envelopes of the individual profiles are omitted here and shown instead in Figs. F.1 & F.2 of Apps. F. In both panels, the lower subplots show the fractional residuals of the hydrostatic masses relative to the GRU masses, (MHSE − MGRU)/MGRU. The thick bl… view at source ↗
Figure 12
Figure 12. Figure 12: Comparison between HSE and GRU derived cluster sparsity estimates. The left panel shows S500,2500 for the combined REXCESS (blue) and X-COP (red) samples, while the right panel presents S200,500 for the X-COP sample alone. Error bars denote the 1σ uncertainties. The dashed black line indicates the one-to-one relation and the solid lines represent the best-fitting linear relations. data. This demonstrates … view at source ↗

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

73 extracted references · 1 linked inside Pith

  1. [1]

    2019, in The 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2623--2631

    Akiba, T., Sano, S., Yanase, T., Ohta, T., & Koyama, M. 2019, in The 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2623--2631

    2019

  2. [2]

    W., Evrard , A

    Allen , S. W., Evrard , A. E., & Mantz , A. B. 2011, , 49, 409

    2011

  3. [3]

    2023, , 675, A188

    Angelinelli , M., Ettori , S., Dolag , K., Vazza , F., & Ragagnin , A. 2023, , 675, A188

    2023

  4. [4]

    2020, , 634, A113

    Ansarifard , S., Rasia , E., Biffi , V., et al. 2020, , 634, A113

    2020

  5. [5]

    W., Piffaretti , R., et al

    Arnaud , M., Pratt , G. W., Piffaretti , R., et al. 2010, , 517, A92

    2010

  6. [6]

    2025, , 703, L18

    Aymerich , G., Douspis , M., Battaglia , N., et al. 2025, , 703, L18

    2025

  7. [7]

    W., et al

    Aymerich , G., Douspis , M., Pratt , G. W., et al. 2024, , 690, A238

    2024

  8. [8]

    2014, , 437, 2328

    Balm \`e s , I., Rasera , Y., Corasaniti , P.-S., & Alimi , J.-M. 2014, , 437, 2328

    2014

  9. [9]

    W., D \'e mocl \`e s , J., & Lovisari , L

    Bartalucci , I., Arnaud , M., Pratt , G. W., D \'e mocl \`e s , J., & Lovisari , L. 2019, , 628, A86

    2019

  10. [10]

    M., Murante , G., Arth , A., et al

    Beck , A. M., Murante , G., Arth , A., et al. 2016, , 455, 2110

    2016

  11. [11]

    E., et al

    Bocquet , S., Grandis , S., Bleem , L. E., et al. 2024, , 110, 083509

    2024

  12. [12]

    W., et al

    B \"o hringer , H., Schuecker , P., Pratt , G. W., et al. 2007, , 469, 363

    2007

  13. [13]

    2024, , 533, 2656

    Braspenning , J., Schaye , J., Schaller , M., et al. 2024, , 533, 2656

    2024

  14. [14]

    M., Alatalo , K., et al

    Cappellari , M., McDermid , R. M., Alatalo , K., et al. 2013, MNRAS, 432, 1862

    2013

  15. [15]

    & Moss , A

    Charnock , T. & Moss , A. 2017, , 837, L28

    2017

  16. [16]

    Chen , C. M. H., Arnaud , M., Pointecouteau , E., Pratt , G. W., & Iqbal , A. 2024, , 688, A219

    2024

  17. [17]

    2021, , 650, A104

    CHEX-MATE Collaboration . 2021, , 650, A104

    2021

  18. [18]

    2014, in Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing ( EMNLP ), ed

    Cho, K., van Merri \"e nboer, B., Gulcehre, C., et al. 2014, in Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing ( EMNLP ), ed. A. Moschitti, B. Pang, & W. Daelemans (Doha, Qatar: Association for Computational Linguistics), 1724--1734

    2014

  19. [19]

    2018, , 477, 139

    Cialone , G., De Petris , M., Sembolini , F., et al. 2018, , 477, 139

    2018

  20. [20]

    S., Ettori , S., Rasera , Y., et al

    Corasaniti , P. S., Ettori , S., Rasera , Y., et al. 2018, , 862, 40

    2018

  21. [21]

    S., Richardson , T

    Corasaniti , P. S., Richardson , T. R. G., Ettori , S., et al. 2025, , 697, A33

    2025

  22. [22]

    H., Pratt , G

    Croston , J. H., Pratt , G. W., B \"o hringer , H., et al. 2008, , 487, 431

    2008

  23. [23]

    2022, , 514, 977

    Cui , W., Dave , R., Knebe , A., et al. 2022, , 514, 977

    2022

  24. [24]

    2018, , 480, 2898

    Cui , W., Knebe , A., Yepes , G., et al. 2018, , 480, 2898

    2018

  25. [25]

    2017, Monthly Notices of the Royal Astronomical Society, 464, 2502

    Cui , W., Power , C., Borgani , S., et al. 2017, Monthly Notices of the Royal Astronomical Society, 464, 2502

    2017

  26. [26]

    2022, Nature Astronomy, 6, 1325

    de Andres , D., Cui , W., Ruppin , F., et al. 2022, Nature Astronomy, 6, 1325

    2022

  27. [27]

    2024, , 528, 1517

    de Andres , D., Cui , W., Yepes , G., et al. 2024, , 528, 1517

    2024

  28. [28]

    2021, , 504, 5383

    De Luca , F., De Petris , M., Yepes , G., et al. 2021, , 504, 5383

    2021

  29. [29]

    2024, , 687, A58

    Dupourqu \'e , S., Clerc , N., Pointecouteau , E., et al. 2024, , 687, A58

    2024

  30. [30]

    2017, Astronomische Nachrichten, 338, 293

    Eckert , D., Ettori , S., Pointecouteau , E., et al. 2017, Astronomische Nachrichten, 338, 293

    2017

  31. [31]

    Eckert , D., Ettori , S., Pointecouteau , E., van der Burg , R. F. J., & Loubser , S. I. 2022, , 662, A123

    2022

  32. [32]

    2019, , 621, A40

    Eckert , D., Ghirardini , V., Ettori , S., et al. 2019, , 621, A40

    2019

  33. [33]

    2019, , 621, A39

    Ettori , S., Ghirardini , V., Eckert , D., et al. 2019, , 621, A39

    2019

  34. [34]

    2023, , 520, 4000

    Ferragamo , A., de Andres , D., Sbriglio , A., et al. 2023, , 520, 4000

    2023

  35. [35]

    2024, , 689, A298

    Ghirardini , V., Bulbul , E., Artis , E., et al. 2024, , 689, A298

    2024

  36. [36]

    2019, , 621, A41

    Ghirardini , V., Eckert , D., Ettori , S., et al. 2019, , 621, A41

    2019

  37. [37]

    M., & Trac , H

    Ho , M., Farahi , A., Rau , M. M., & Trac , H. 2021, , 908, 204

    2021

  38. [38]

    M., et al

    Ho , M., Ntampaka , M., Rau , M. M., et al. 2022, Nature Astronomy, 6, 936

    2022

  39. [39]

    M., Ntampaka , M., et al

    Ho , M., Rau , M. M., Ntampaka , M., et al. 2019, , 887, 25

    2019

  40. [40]

    & Schmidhuber, J

    Hochreiter, S. & Schmidhuber, J. 1997, Neural Computation, 9, 1735

    1997

  41. [41]

    2020, , 492, 3685

    Huang , S., Leauthaud , A., Hearin , A., et al. 2020, , 492, 3685

    2020

  42. [42]

    B., & Majumdar , S

    Iqbal , A., Kale , R., Nath , B. B., & Majumdar , S. 2018, , 480, L68

    2018

  43. [43]

    B., & Roychowdhury , S

    Iqbal , A., Majumdar , S., Nath , B. B., & Roychowdhury , S. 2023 a , , 518, 2735

    2023

  44. [44]

    2025, , 704, A334

    Iqbal , A., Majumdar , S., Rasia , E., et al. 2025, , 704, A334

    2025

  45. [45]

    W., Bobin , J., et al

    Iqbal , A., Pratt , G. W., Bobin , J., et al. 2023 b , , 679, A51

    2023

  46. [46]

    2007, , 659, 257

    Kawahara , H., Suto , Y., Kitayama , T., et al. 2007, , 659, 257

    2007

  47. [47]

    Kay , S. T. & Pratt , G. W. 2022, in Handbook of X-ray and Gamma-ray Astrophysics. Edited by Cosimo Bambi and Andrea Santangelo, 100

    2022

  48. [48]

    2024, , 682, A132

    Krippendorf , S., Baron Perez , N., Bulbul , E., et al. 2024, , 682, A132

    2024

  49. [49]

    2024, , 687, A129

    Lehle , K., Nelson , D., Pillepich , A., Truong , N., & Rohr , E. 2024, , 687, A129

    2024

  50. [50]

    2023, , 523, 1228

    Li , Q., Cui , W., Yang , X., et al. 2023, , 523, 1228

    2023

  51. [51]

    & Maughan , B

    Lovisari , L. & Maughan , B. J. 2022, in Handbook of X-ray and Gamma-ray Astrophysics, 65

    2022

  52. [52]

    Lundberg, S. M. & Lee, S.-I. 2017, in Proceedings of the 31st International Conference on Neural Information Processing Systems, NIPS'17 (Red Hook, NY, USA: Curran Associates Inc.), 4768–4777

    2017

  53. [53]

    & Mohr , J

    Majumdar , S. & Mohr , J. J. 2004, , 613, 41

    2004

  54. [54]

    W., Rapetti , D., & Ebeling , H

    Mantz , A., Allen , S. W., Rapetti , D., & Ebeling , H. 2010, , 406, 1759

    2010

  55. [55]

    F., Pratt , G

    Mu \ n oz-Echeverr \' a , M., Mac \' as-P \'e rez , J. F., Pratt , G. W., et al. 2024, , 682, A147

    2024

  56. [56]

    2024, , 686, A157

    Nelson , D., Pillepich , A., Ayromlou , M., et al. 2024, , 686, A157

    2024

  57. [57]

    J., et al

    Ntampaka , M., Trac , H., Sutherland , D. J., et al. 2015, , 803, 50

    2015

  58. [58]

    2019, , 876, 82

    Ntampaka , M., ZuHone , J., Eisenstein , D., et al. 2019, , 876, 82

    2019

  59. [59]

    2016, , 594, A13

    Planck Collaboration XIII . 2016, , 594, A13

    2016

  60. [60]

    W., Arnaud , M., Biviano , A., et al

    Pratt , G. W., Arnaud , M., Biviano , A., et al. 2019, , 215, 25

    2019

  61. [61]

    W., Arnaud , M., Piffaretti , R., et al

    Pratt , G. W., Arnaud , M., Piffaretti , R., et al. 2010, , 511, A85

    2010

  62. [62]

    W., Croston , J

    Pratt , G. W., Croston , J. H., Arnaud , M., & B \"o hringer , H. 2009, , 498, 361

    2009

  63. [63]

    D., Wojtak , R., Ansari , Z., Gall , C., & Hjorth , J

    Ramanah , K. D., Wojtak , R., Ansari , Z., Gall , C., & Hjorth , J. 2020, , 499, 1985

    2020

  64. [64]

    2015, , 813, L17

    Rasia , E., Borgani , S., Murante , G., et al. 2015, , 813, L17

    2015

  65. [65]

    2013, The Astronomical Review, 8, 40

    Rasia , E., Meneghetti , M., & Ettori , S. 2013, The Astronomical Review, 8, 40

    2013

  66. [66]

    2025, , 702, A182

    Rasia , E., Tripodi , R., Borgani , S., et al. 2025, , 702, A182

    2025

  67. [67]

    P., & McCarthy , I

    Semboloni , E., Hoekstra , H., Schaye , J., van Daalen , M. P., & McCarthy , I. G. 2011, , 417, 2020

    2011

  68. [68]

    M., Bellomi , E., Sayers , J., et al

    Silich , E. M., Bellomi , E., Sayers , J., et al. 2024, in European Physical Journal Web of Conferences, Vol. 293, mm Universe 2023 - Observing the Universe at mm Wavelengths, 00050

    2024

  69. [69]

    2024, , 682, A177

    Sun , M., Li , J., Cao , S., & Liu , X. 2024, , 682, A177

    2024

  70. [70]

    2026, arXiv e-prints, arXiv:2602.11989

    Tam , S.-I., Umetsu , K., Amara , A., et al. 2026, arXiv e-prints, arXiv:2602.11989

    arXiv 2026

  71. [71]

    2013, , 428, 2366

    Vazza , F., Br \"u ggen , M., & Gheller , C. 2013, , 428, 2366

    2013

  72. [72]

    A., Ebeling , H., et al

    Vikhlinin , A., Burenin , R. A., Ebeling , H., et al. 2009, , 692, 1033

    2009

  73. [73]

    2023, arXiv e-prints, arXiv:2304.11127

    Watanabe , S. 2023, arXiv e-prints, arXiv:2304.11127

    Pith/arXiv arXiv 2023

This paper was first reviewed by deepseek-v4-flash on August 1, 2026.