pith. sign in

arxiv: 2606.10038 · v1 · pith:USHS3NSNnew · submitted 2026-06-08 · 🌌 astro-ph.CO · astro-ph.GA

Learning the Universe with the 2nd Generation of CAMELS: Varying 35 parameters of the IllustrisTNG model in (50Mpc/h)³ boxes

Shy Genel , Yongseok Jo , Boon Kiat Oh , Megan Taylor Tillman , Max E. Lee , Jun-Young Lee , Elena Hern\'andez-Mart\'inez , Christopher C. Lovell
show 5 more authors
Xavier Sims Blakesley Burkhart Kentaro Nagamine Daniel Angl\'es-Alc\'azar Francisco Villaescusa-Navarro
This is my paper

Pith reviewed 2026-06-27 15:11 UTC · model grok-4.3

classification 🌌 astro-ph.CO astro-ph.GA
keywords cosmological simulationsparameter inferenceneural networksmatter power spectragalaxy distributionsintergalactic mediumvolume effectshalo properties
0
0 comments X

The pith

Larger (50 Mpc/h)^3 simulation volumes produce tighter marginal constraints on 35 parameters via neural networks than smaller boxes.

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

The paper generates 1,192 simulations that vary 35 cosmological, astrophysical, and numerical parameters in volumes eight times larger than prior runs. It trains multilayer perceptrons, convolutional networks, graph networks, and Gaussian processes on power spectra, projected maps, galaxy graphs, and halo thermodynamics to infer the parameters. Larger volumes reduce sample variance and include more massive halos, yielding tighter constraints whose improvement varies by input type. The gains fall short of scaling with the square root of the volume increase, which the authors link to mode coupling or parameter degeneracies. The work additionally tracks how four new parameters that set the amplitude and timing of the ionizing background alter intergalactic medium temperature statistics.

Core claim

Increasing simulation volume from (25 Mpc/h)^3 to (50 Mpc/h)^3 generally produces tighter marginal constraints on the 35 parameters, with the size of the improvement depending on the data representation fed to the neural networks.

What carries the argument

Neural-network inference of the 35 parameters from four distinct data representations extracted from the simulations, tested across two different volume sizes.

If this is right

  • Larger volumes deliver tighter constraints on the full set of 35 parameters.
  • The amount of improvement differs across power spectra, maps, graphs, and halo properties.
  • Constraint tightening scales more weakly than the square root of volume growth, consistent with mode coupling or degeneracies.
  • The four new ionizing-background parameters measurably alter intergalactic medium temperature statistics.

Where Pith is reading between the lines

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

  • Even larger volumes could be tested to determine whether further gains continue or saturate.
  • Joint use of multiple input representations might produce additional constraint improvements beyond any single channel.
  • The released simulation outputs allow other inference methods to be benchmarked against the neural-network results reported here.

Load-bearing premise

The observed tightening of constraints arises primarily from the increase in physical volume rather than from shifts in parameter sampling density or neural-network settings between the two simulation generations.

What would settle it

A side-by-side run that holds parameter sampling density and network hyperparameters fixed while changing only the box volume, then measures whether the constraint widths improve exactly in proportion to the volume ratio.

Figures

Figures reproduced from arXiv: 2606.10038 by Blakesley Burkhart, Boon Kiat Oh, Christopher C. Lovell, Daniel Angl\'es-Alc\'azar, Elena Hern\'andez-Mart\'inez, Francisco Villaescusa-Navarro, Jun-Young Lee, Kentaro Nagamine, Max E. Lee, Megan Taylor Tillman, Shy Genel, Xavier Sims, Yongseok Jo.

Figure 1
Figure 1. Figure 1: Density-temperature visualizations of two CAMELS simulations, from the first generation (a) and from the second generation (b), which is presented in this work. This paper is organized as follows. In Section 2 we describe the new CAMELS simulation suite in detail, including the cosmological and astrophysical parameter space, the IllustrisTNG subgrid model, and resolution and volume characteristics. Section… view at source ↗
Figure 2
Figure 2. Figure 2: Global statistical properties of the simulations across the SB35 (blue), SB28 (red), LH (green) and CV (black; solid for 50 h −1Mpc, dashed for 25 h −1Mpc) simulation sets, shown as medians (lines) and 16th-to-84th percentile ranges (shaded regions). The top row shows matter (left) and gas (middle) power spectra as well as the matter power spectrum ratio between corresponding hydrodynamical and N-body runs… view at source ↗
Figure 3
Figure 3. Figure 3: Scaling relations between galaxy and halo properties across the SB35 (blue), SB28 (red), LH (green) and CV (black) simulation sets. The figure includes relations such as stellar mass, baryonic mass, black hole mass, circular velocity, stellar half-mass radius, star formation rate, and total gas metallicity (ISM and CGM combined) versus halo mass or stellar mass. The SB35 simulations extend the dynamic rang… view at source ↗
Figure 4
Figure 4. Figure 4: The predicted T0 (top) and γ (bottom) from the 50 h −1Mpc SB35 set (blue; left column), which is compared to the 25 h −1Mpc SB28 set (red dashed; left column), as well as the 50 h −1Mpc 1P set (right four columns). The 50 h −1Mpc simulations vary the extragalactic ionizing background, which is not done in SB28. In the 1P panels, the black lines correspond to the fiducial simulation predictions, the blue an… view at source ↗
Figure 5
Figure 5. Figure 5: A comparison of the concentration–mass re￾lation for dark matter halos between our 50 h −1Mpc and 25 h −1Mpc boxes. We compare three pairs of simulation sets: hydrodynamical with the fiducial TNG model (CV sets; green), pure N-body with the fiducial cosmological parame￾ters (CV sets; orange), and pure N-body with variations in five cosmological parameters (SB28 and SB35 sets; blue). In all cases, the lack … view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of parameter inference from the z = 0 non-linear matter power spectrum in SB28 (left pan￾els) and SB35 (right panels). Each panel shows predicted versus true values for the tested parameters (Ωm at the top, σ8 at the bottom), with black solid curves indicating the run￾ning medians. The larger volume of SB35 leads to somewhat tighter predictions, especially for σ8, but the RMSE ratios are far fro… view at source ↗
Figure 8
Figure 8. Figure 8: A comparison of the cosmic variance between (25 h −1Mpc)3 and (50 h −1Mpc)3 volumes using our CV25 (left) and CV50 (right) sets run with the fiducial Illus￾trisTNG model. Four quantities are displayed, from top to bottom: the matter power spectrum ratio between the hydro￾dynamical and their corresponding dark matter-only simula￾tions, the stellar mass function, the stellar-to-halo mass ratio, and the histo… view at source ↗
Figure 7
Figure 7. Figure 7: Comparison between SB28 (left) and SB35 (right) for S8 parameter inference from the matter power spectrum, in the same format as [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Parameter inference for Ωm (top) and σ8 (bottom) using no-monopole 25 h −1Mpc maps from SB28 and SB35 (quadrants), including both self- and cross-validations. Models trained on SB28 maps (left) and on SB35 quadrants (right) are tested on SB28 maps (red) and on SB35 quadrants (blue). No clear out-of-distribution biases are observed. decay ∈ [10−8 , 10−4 ], number of layers nlayers ∈ [1, 5], and hidden chann… view at source ↗
Figure 10
Figure 10. Figure 10: Inference results on the cosmological param￾eters Ωm (top) and σ8 (bottom) from galaxy graphs using GNNs. We observe an increased constraining power with SB35 (right) compared to SB28 (left), but the scaling is sig￾nificantly weaker than the square root of the volume ratio, as discussed in Section 4.4. As seen in these previous sections, we find an improve￾ment in the constraining power between SB28 and S… view at source ↗
Figure 11
Figure 11. Figure 11: Top: A comparison of the distribution of the halo count per simulation box of halos with M200c > 1013 h −1 M⊙, in the 25 h −1Mpc SB28 simulations (red) and 50 h −1Mpc SB35 simulations (blue) at z = 0. The larger volumes yield a significantly higher count of massive halos, but when rescaled appropriately by the volume and the total number of simulations in each set (thin blue), the distribution of halo cou… view at source ↗
Figure 12
Figure 12. Figure 12: A comparison of thermal Sunyaev–Zeldovich Y –M relations at z = 0 for the fiducial IllustrisTNG model, as inferred from different simulations. Both panels include binned actual measurements from the TNG300 simulation (black) and from our respective CV sets (red and blue squares), which serve as ‘ground truth’, as well as the relation emulated with CARPoolGP based on the CAMELS-ZoomGZ zoom-in simulations (… view at source ↗
Figure 13
Figure 13. Figure 13: Correlation coefficients between predicted and true values of 35 parameters, based on inference networks that are trained on halo thermodynamical radial profiles. Three distinct networks use three radial ranges: the full range up to 2.5R200c (grey), the inner range up to 0.9R200c (green), and the outer range between 0.9R200c and 2.5R200c (purple). While adding the outer profiles to the inner ones adds lit… view at source ↗
Figure 14
Figure 14. Figure 14: A visual representation of the data in [PITH_FULL_IMAGE:figures/full_fig_p025_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Effect of single-parameter variations in the 50 h −1Mpc 1P simulation set on the matter power spectrum at z = 0. Each curve shows the ratio of the power spectra of a hydrodynamical simulation and its corresponding N-body simulation, where each parameter is varied separately in each panel. The simulation with the middle value for each parameter is shown in black, while blue and red curves denote the minimu… view at source ↗
Figure 16
Figure 16. Figure 16: Star formation rate density as a function of redshift in the 50 h −1Mpc 1P simulation set, showing the effect of varying individual parameters, similarly to [PITH_FULL_IMAGE:figures/full_fig_p028_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Stellar-to-halo mass relation at z = 0 for the 50 h −1Mpc 1P simulation set, illustrating how variations in cosmological and astrophysical parameters affect galaxy formation efficiency. Gnedin, N. Y., & Madau, P. 2022, Living Reviews in Computational Astrophysics, 8, 3, doi: 10.1007/s41115-022-00015-5 Guan, C., Wang, X., Chen, H., Zhang, Z., & Zhu, W. 2022, in Proceedings of Machine Learning Research, Vol… view at source ↗
read the original abstract

We present a new set of 1,192 cosmological simulations as part of the CAMELS project, in which a space of 35 cosmological, astrophysical, and numerical parameters is explored around the fiducial IllustrisTNG model. The volume of each of these simulations is (50Mpc/h)^3, eight times larger than that of previous CAMELS simulations. This provides lower sample variance as well as access to more massive halos and more diverse environments. We focus this work on exploring the advantages these differences provide for parameter inference powered by neural networks. We generate training sets based on the matter power spectra, projected maps of the volumes, graphs representing galaxy spatial distributions, and thermodynamical properties of massive halos. We employ multilayer perceptrons, convolutional neural networks, graph neural networks, and Gaussian processes, respectively, to extract information on the simulation parameters from these inputs while comparing systematically to analogous results from our previous generation of (25Mpc/h)^3 simulations. We generally find that the new, larger volumes produce tighter marginal constraints on the parameters, to degrees that vary between the different inputs. The improvements, however, scale more weakly than with the square root of the increase in the amount of data (i.e., physical volume). We interpret this as originating either from information loss due to mode coupling or from complex degeneracies in parameter space. We also discuss the effects on statistics of the intergalactic medium temperature from four new parameters that are varied in these simulations, which control the amplitude and timing of the ionizing background radiation. We publicly release the simulation outputs and ancillary data at https://camels.readthedocs.io.

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

2 major / 1 minor

Summary. The manuscript presents 1,192 new CAMELS simulations exploring a 35-parameter space (cosmological, astrophysical, and numerical) around the IllustrisTNG model in (50 Mpc/h)^3 volumes. Training sets are generated from matter power spectra, projected maps, galaxy graphs, and halo thermodynamical properties; multilayer perceptrons, CNNs, graph neural networks, and Gaussian processes are used to infer parameters. Results are compared systematically to the prior (25 Mpc/h)^3 generation, with the central finding that larger volumes yield tighter marginal constraints whose improvement scales more weakly than √8.

Significance. If the attribution to volume holds after controls, the work would help quantify how simulation volume affects ML-based cosmological parameter inference and inform the design of future large-volume campaigns. The public data release strengthens the contribution.

major comments (2)
  1. [Abstract] Abstract: the claim that larger volumes produce tighter marginal constraints is presented without any quantitative metrics, improvement factors, error bars, or tables summarizing the degree of tightening for each input type (power spectra, maps, graphs, halos).
  2. [Results and Methods (comparison)] Comparison to prior generation (throughout results sections): the manuscript states the comparison is 'systematic' but does not report whether the number of training realizations, exact parameter ranges and sampling density (Latin-hypercube/Sobol), neural-network hyperparameters/architectures, or preprocessing pipelines were held identical between the two generations; without this isolation, the weaker-than-√8 scaling cannot be unambiguously attributed to mode coupling or degeneracies rather than uncontrolled differences in sampling or training.
minor comments (1)
  1. [Abstract] Abstract: the data-release URL is provided and the description of the four new ionizing-background parameters is clear.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight opportunities to improve the clarity and rigor of our presentation. We address each major comment below and will incorporate revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that larger volumes produce tighter marginal constraints is presented without any quantitative metrics, improvement factors, error bars, or tables summarizing the degree of tightening for each input type (power spectra, maps, graphs, halos).

    Authors: We agree that the abstract would benefit from quantitative detail on the reported improvements. In the revised manuscript we will insert specific tightening factors (with uncertainties) for the marginal constraints obtained from each input type, drawn from the results sections. revision: yes

  2. Referee: [Results and Methods (comparison)] Comparison to prior generation (throughout results sections): the manuscript states the comparison is 'systematic' but does not report whether the number of training realizations, exact parameter ranges and sampling density (Latin-hypercube/Sobol), neural-network hyperparameters/architectures, or preprocessing pipelines were held identical between the two generations; without this isolation, the weaker-than-√8 scaling cannot be unambiguously attributed to mode coupling or degeneracies rather than uncontrolled differences in sampling or training.

    Authors: The comparisons were performed with matched training-set sizes, identical parameter ranges and Latin-hypercube sampling, the same network architectures and hyperparameters, and the same preprocessing pipelines as the prior generation. We acknowledge that these controls were not documented in a single dedicated location. We will add an explicit subsection (or table) in the Methods that tabulates the matching choices, thereby making the systematic nature of the comparison fully transparent and allowing readers to evaluate the attribution to volume. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical comparison of new independent simulations to prior generation

full rationale

The paper generates a fresh set of 1,192 simulations in (50 Mpc/h)^3 volumes, trains separate neural networks (MLPs, CNNs, GNNs, GPs) on power spectra, maps, graphs and halo properties, and reports observed tighter marginal constraints relative to the earlier (25 Mpc/h)^3 CAMELS generation. No equation, ansatz or self-citation reduces the reported constraints to quantities fitted from the same data; the comparison is presented as an external empirical result. The prior generation constitutes independent evidence, and the central claim does not rely on any load-bearing self-citation or definitional identity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the fidelity of the IllustrisTNG hydrodynamical model and the assumption that neural networks trained on simulated data can extract parameter information without dominant biases from simulation-specific artifacts.

axioms (1)
  • domain assumption Fiducial IllustrisTNG model provides a sufficiently accurate representation of galaxy formation physics for the purpose of parameter inference training
    All simulations are run around this fiducial model; the claim of improved constraints assumes this baseline is representative.

pith-pipeline@v0.9.1-grok · 5917 in / 1053 out tokens · 41503 ms · 2026-06-27T15:11:50.683559+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

97 extracted references · 93 canonical work pages · 15 internal anchors

  1. [1]

    Abbott , T. M. C., Aguena , M., Alarcon , A., et al. 2022, title Dark Energy Survey Year 3 results: Cosmological constraints from galaxy clustering and weak lensing , , 105, 023520, 10.1103/PhysRevD.105.023520

    work page doi:10.1103/physrevd.105.023520 2022

  2. [2]

    Akiba, T., Sano, S., Yanase, T., Ohta, T., & Koyama, M. 2019, in The 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (New York, NY, USA: Association for Computing Machinery), 2623--2631, 10.1145/3292500.3330701

    work page doi:10.1145/3292500.3330701 2019

  3. [3]

    Alam , S., Aubert , M., Avila , S., et al. 2021, title Completed SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Cosmological implications from two decades of spectroscopic surveys at the Apache Point Observatory , , 103, 083533, 10.1103/PhysRevD.103.083533

    work page doi:10.1103/physrevd.103.083533 2021

  4. [4]

    Anderson , L., Aubourg , \'E ., Bailey , S., et al. 2014, title The clustering of galaxies in the SDSS-III Baryon Oscillation Spectroscopic Survey: baryon acoustic oscillations in the Data Releases 10 and 11 Galaxy samples , , 441, 24, 10.1093/mnras/stu523

    work page doi:10.1093/mnras/stu523 2014

  5. [5]

    2026, title Claude (Opus-4.7), , https://www.anthropic.com

    Anthropic . 2026, title Claude (Opus-4.7), , https://www.anthropic.com

    2026

  6. [6]

    2025, title The BIG SOBOL SEQUENCE: How many simulations do we need for simulation-based inference in cosmology? , , 703, A301, 10.1051/0004-6361/202554602

    Bairagi , A., Wandelt , B., & Villaescusa-Navarro , F. 2025, title The BIG SOBOL SEQUENCE: How many simulations do we need for simulation-based inference in cosmology? , , 703, A301, 10.1051/0004-6361/202554602

    work page doi:10.1051/0004-6361/202554602 2025

  7. [7]

    2024, title A Cosmic-Scale Benchmark for Symmetry-Preserving Data Processing , arXiv e-prints, arXiv:2410.20516, 10.48550/arXiv.2410.20516

    Balla , J., Mishra-Sharma , S., Cuesta-Lazaro , C., Jaakkola , T., & Smidt , T. 2024, title A Cosmic-Scale Benchmark for Symmetry-Preserving Data Processing , arXiv e-prints, arXiv:2410.20516, 10.48550/arXiv.2410.20516

    work page doi:10.48550/arxiv.2410.20516 2024

  8. [8]

    S., Wechsler , R

    Behroozi , P. S., Wechsler , R. H., & Wu , H.-Y. 2013 a , title The ROCKSTAR Phase-space Temporal Halo Finder and the Velocity Offsets of Cluster Cores , , 762, 109, 10.1088/0004-637X/762/2/109

    work page internal anchor Pith review doi:10.1088/0004-637x/762/2/109 2013

  9. [9]

    S., Wechsler , R

    Behroozi , P. S., Wechsler , R. H., Wu , H.-Y., et al. 2013 b , title Gravitationally Consistent Halo Catalogs and Merger Trees for Precision Cosmology , , 763, 18, 10.1088/0004-637X/763/1/18

    work page internal anchor Pith review doi:10.1088/0004-637x/763/1/18 2013

  10. [10]

    2022, title The ASTRID simulation: galaxy formation and reionization , , 512, 3703, 10.1093/mnras/stac648

    Bird , S., Ni , Y., Di Matteo , T., et al. 2022, title The ASTRID simulation: galaxy formation and reionization , , 512, 3703, 10.1093/mnras/stac648

    work page doi:10.1093/mnras/stac648 2022

  11. [11]

    A., Schaye , J., Bower , R

    Crain , R. A., Schaye , J., Bower , R. G., et al. 2015, title The EAGLE simulations of galaxy formation: calibration of subgrid physics and model variations , , 450, 1937, 10.1093/mnras/stv725

    work page doi:10.1093/mnras/stv725 2015

  12. [12]

    , keywords =

    Crocce , M., Pueblas , S., & Scoccimarro , R. 2006, title Transients from initial conditions in cosmological simulations , , 373, 369, 10.1111/j.1365-2966.2006.11040.x

    work page doi:10.1111/j.1365-2966.2006.11040.x 2006

  13. [13]

    2019, title SIMBA: Cosmological simulations with black hole growth and feedback , , 486, 2827, 10.1093/mnras/stz937

    Dav \'e , R., Angl \'e s-Alc \'a zar , D., Narayanan , D., et al. 2019, title SIMBA: Cosmological simulations with black hole growth and feedback , , 486, 2827, 10.1093/mnras/stz937

    work page internal anchor Pith review doi:10.1093/mnras/stz937 2019

  14. [14]

    S., & White , S

    Davis , M., Efstathiou , G., Frenk , C. S., & White , S. D. M. 1985, title The evolution of large-scale structure in a universe dominated by cold dark matter , , 292, 371, 10.1086/163168

    work page doi:10.1086/163168 1985

  15. [15]

    de Santi , N. S. M., Shao , H., Villaescusa-Navarro , F., et al. 2023, title Robust Field-level Likelihood-free Inference with Galaxies , , 952, 69, 10.3847/1538-4357/acd1e2

    work page doi:10.3847/1538-4357/acd1e2 2023

  16. [16]

    M., Angl \'e s-Alc \'a zar , D., Thiele , L., et al

    Delgado , A. M., Angl \'e s-Alc \'a zar , D., Thiele , L., et al. 2023, title Predicting the impact of feedback on matter clustering with machine learning in CAMELS , , 526, 5306, 10.1093/mnras/stad2992

    work page doi:10.1093/mnras/stad2992 2023

  17. [17]

    2022, title A Comprehensive Study on Large-Scale Graph Training: Benchmarking and Rethinking , arXiv e-prints, arXiv:2210.07494, 10.48550/arXiv.2210.07494

    Duan , K., Liu , Z., Wang , P., et al. 2022, title A Comprehensive Study on Large-Scale Graph Training: Benchmarking and Rethinking , arXiv e-prints, arXiv:2210.07494, 10.48550/arXiv.2210.07494

    work page doi:10.48550/arxiv.2210.07494 2022

  18. [18]

    P., Ramp \'a s ek , L., Galkin , M., et al

    Dwivedi , V. P., Ramp \'a s ek , L., Galkin , M., et al. 2022, title Long Range Graph Benchmark , arXiv e-prints, arXiv:2206.08164, 10.48550/arXiv.2206.08164

    work page doi:10.48550/arxiv.2206.08164 2022

  19. [19]

    Overview of the Euclid mission

    Euclid Collaboration , Mellier , Y., Abdurro'uf , et al. 2025, title Euclid: I. Overview of the Euclid mission , , 697, A1, 10.1051/0004-6361/202450810

    work page doi:10.1051/0004-6361/202450810 2025

  20. [20]

    2009, title A New Calculation of the Ionizing Background Spectrum and the Effects of He II Reionization , , 703, 1416, 10.1088/0004-637X/703/2/1416

    Faucher-Gigu \`e re , C., Lidz , A., Zaldarriaga , M., & Hernquist , L. 2009, title A New Calculation of the Ionizing Background Spectrum and the Effects of He II Reionization , , 703, 1416, 10.1088/0004-637X/703/2/1416

    work page doi:10.1088/0004-637x/703/2/1416 2009

  21. [21]

    2024, title Cosmological baryon spread and impact on matter clustering in CAMELS , , 529, 4896, 10.1093/mnras/stae817

    Gebhardt , M., Angl \'e s-Alc \'a zar , D., Borrow , J., et al. 2024, title Cosmological baryon spread and impact on matter clustering in CAMELS , , 529, 4896, 10.1093/mnras/stae817

    work page doi:10.1093/mnras/stae817 2024

  22. [22]

    2026, title Cosmological back-reaction of baryons on dark matter in the CAMELS simulations , , 547, stag525, 10.1093/mnras/stag525

    Gebhardt , M., Angl \'e s-Alc \'a zar , D., Genel , S., et al. 2026, title Cosmological back-reaction of baryons on dark matter in the CAMELS simulations , , 547, stag525, 10.1093/mnras/stag525

    work page doi:10.1093/mnras/stag525 2026

  23. [23]

    2014, title Introducing the Illustris project: the evolution of galaxy populations across cosmic time , , 445, 175, 10.1093/mnras/stu1654

    Genel , S., Vogelsberger , M., Springel , V., et al. 2014, title Introducing the Illustris project: the evolution of galaxy populations across cosmic time , , 445, 175, 10.1093/mnras/stu1654

    work page doi:10.1093/mnras/stu1654 2014

  24. [24]

    Y., & Madau , P

    Gnedin , N. Y., & Madau , P. 2022, title Modeling cosmic reionization , Living Reviews in Computational Astrophysics, 8, 3, 10.1007/s41115-022-00015-5

    work page doi:10.1007/s41115-022-00015-5 2022

  25. [25]

    2022, in Proceedings of Machine Learning Research, Vol

    Guan, C., Wang, X., Chen, H., Zhang, Z., & Zhu, W. 2022, in Proceedings of Machine Learning Research, Vol. 162, Proceedings of the 39th International Conference on Machine Learning, ed. K. Chaudhuri, S. Jegelka, L. Song, C. Szepesvari, G. Niu, & S. Sabato (PMLR), 7968--7981. https://proceedings.mlr.press/v162/guan22d.html

    2022

  26. [26]

    2024, title Cosmological constraints from non-Gaussian and nonlinear galaxy clustering using the SIMBIG inference framework , Nature Astronomy, 8, 1457, 10.1038/s41550-024-02344-2

    Hahn , C., Lemos , P., Parker , L., et al. 2024, title Cosmological constraints from non-Gaussian and nonlinear galaxy clustering using the SIMBIG inference framework , Nature Astronomy, 8, 1457, 10.1038/s41550-024-02344-2

    work page doi:10.1038/s41550-024-02344-2 2024

  27. [27]

    2025, title Cosmological and Astrophysical Parameter Inference from Stacked Galaxy Cluster Profiles Using CAMELS-zoomGZ , , 981, 170, 10.3847/1538-4357/ada9e9

    Hern \'a ndez-Mart \' nez , E., Genel , S., Villaescusa-Navarro , F., et al. 2025, title Cosmological and Astrophysical Parameter Inference from Stacked Galaxy Cluster Profiles Using CAMELS-zoomGZ , , 981, 170, 10.3847/1538-4357/ada9e9

    work page doi:10.3847/1538-4357/ada9e9 2025

  28. [28]

    2013, title Nine-year Wilkinson Microwave Anisotropy Probe (WMAP) Observations: Cosmological Parameter Results , , 208, 19, 10.1088/0067-0049/208/2/19

    Hinshaw , G., Larson , D., Komatsu , E., et al. 2013, title Nine-year Wilkinson Microwave Anisotropy Probe (WMAP) Observations: Cosmological Parameter Results , , 208, 19, 10.1088/0067-0049/208/2/19

    work page internal anchor Pith review doi:10.1088/0067-0049/208/2/19 2013

  29. [29]

    J., Chartier , N., et al

    Ho , M., Bartlett , D. J., Chartier , N., et al. 2024, title LtU-ILI: An All-in-One Framework for Implicit Inference in Astrophysics and Cosmology , The Open Journal of Astrophysics, 7, 54, 10.33232/001c.120559

    work page doi:10.33232/001c.120559 2024

  30. [30]

    2025, title CosmoBench: A Multiscale, Multiview, Multitask Cosmology Benchmark for Geometric Deep Learning , arXiv e-prints, arXiv:2507.03707, 10.48550/arXiv.2507.03707

    Huang , N., Stiskalek , R., Lee , J.-Y., et al. 2025, title CosmoBench: A Multiscale, Multiview, Multitask Cosmology Benchmark for Geometric Deep Learning , arXiv e-prints, arXiv:2507.03707, 10.48550/arXiv.2507.03707

    work page doi:10.48550/arxiv.2507.03707 2025

  31. [31]

    A., et al

    Ishiyama , T., Prada , F., Klypin , A. A., et al. 2021, title The Uchuu simulations: Data Release 1 and dark matter halo concentrations , , 506, 4210, 10.1093/mnras/stab1755

    work page doi:10.1093/mnras/stab1755 2021

  32. [32]

    Jeffrey , N., & Wandelt , B. D. 2020, title Solving high-dimensional parameter inference: marginal posterior densities & Moment Networks , arXiv e-prints, arXiv:2011.05991, 10.48550/arXiv.2011.05991

    work page doi:10.48550/arxiv.2011.05991 2020

  33. [33]

    2025, title Towards Robustness Across Cosmological Simulation Models TNG, SIMBA, ASTRID, and EAGLE , arXiv e-prints, arXiv:2502.13239, 10.48550/arXiv.2502.13239

    Jo , Y., Genel , S., Sengupta , A., et al. 2025, title Towards Robustness Across Cosmological Simulation Models TNG, SIMBA, ASTRID, and EAGLE , arXiv e-prints, arXiv:2502.13239, 10.48550/arXiv.2502.13239

    work page doi:10.48550/arxiv.2502.13239 2025

  34. [34]

    Adam: A Method for Stochastic Optimization

    Kingma , D. P., & Ba , J. 2014, title Adam: A Method for Stochastic Optimization , arXiv e-prints, arXiv:1412.6980, 10.48550/arXiv.1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1412.6980 2014

  35. [35]

    A., Trujillo-Gomez , S., & Primack , J

    Klypin , A. A., Trujillo-Gomez , S., & Primack , J. 2011, title Dark Matter Halos in the Standard Cosmological Model: Results from the Bolshoi Simulation , , 740, 102, 10.1088/0004-637X/740/2/102

    work page doi:10.1088/0004-637x/740/2/102 2011

  36. [36]

    J., & McGibbon , R

    Kugel , R., Schaye , J., Schaller , M., Forouhar Moreno , V. J., & McGibbon , R. J. 2025, title The FLAMINGO Project: An assessment of the systematic errors in the predictions of models for galaxy cluster counts used to infer cosmological parameters , , 537, 2179, 10.1093/mnras/staf111

    work page doi:10.1093/mnras/staf111 2025

  37. [37]

    2023, title FLAMINGO: calibrating large cosmological hydrodynamical simulations with machine learning , , 526, 6103, 10.1093/mnras/stad2540

    Kugel , R., Schaye , J., Schaller , M., et al. 2023, title FLAMINGO: calibrating large cosmological hydrodynamical simulations with machine learning , , 526, 6103, 10.1093/mnras/stad2540

    work page doi:10.1093/mnras/stad2540 2023

  38. [38]

    2025, title ethlau/CAMELS\_emulator: v1.0.0, , v1.0.0 Zenodo, 10.5281/zenodo.14714183

    Lau, E. 2025, title ethlau/CAMELS\_emulator: v1.0.0, , v1.0.0 Zenodo, 10.5281/zenodo.14714183

    work page doi:10.5281/zenodo.14714183 2025

  39. [39]

    T., Nagai , D., Farahi , A., et al

    Lau , E. T., Nagai , D., Farahi , A., et al. 2025 a , title Baryon Pasting the Uchuu Light-cone Simulation , , 980, 122, 10.3847/1538-4357/ada940

    work page doi:10.3847/1538-4357/ada940 2025

  40. [40]

    T., Nagai , D., Bogd \'a n , \'A ., et al

    Lau , E. T., Nagai , D., Bogd \'a n , \'A ., et al. 2025 b , title X-Raying CAMELS: Constraining Baryonic Feedback in the Circumgalactic Medium with the CAMELS Simulations and eRASS X-Ray Observations , , 984, 190, 10.3847/1538-4357/adc450

    work page doi:10.3847/1538-4357/adc450 2025

  41. [41]

    2025, title Cosmology with Topological Deep Learning, The Astrophysical Journal, 989, 47, 10.3847/1538-4357/ade806

    Lee, J.-Y., & Villaescusa-Navarro, F. 2025, title Cosmology with Topological Deep Learning, The Astrophysical Journal, 989, 47, 10.3847/1538-4357/ade806

    work page doi:10.3847/1538-4357/ade806 2025

  42. [42]

    E., Genel , S., Wandelt , B

    Lee , M. E., Genel , S., Wandelt , B. D., et al. 2024, title Zooming by in the CARPoolGP Lane: New CAMELS-TNG Simulations of Zoomed-in Massive Halos , , 968, 11, 10.3847/1538-4357/ad3d4a

    work page doi:10.3847/1538-4357/ad3d4a 2024

  43. [43]

    Li , X., Zhou , Z., Yao , J., et al. 2023, title Neural Atoms: Propagating Long-range Interaction in Molecular Graphs through Efficient Communication Channel , arXiv e-prints, arXiv:2311.01276, 10.48550/arXiv.2311.01276

    work page doi:10.48550/arxiv.2311.01276 2023

  44. [44]

    2026, title One Latent to Fit Them All: A Unified Representation of Baryonic Feedback on Matter Distribution , , 996, L41, 10.3847/2041-8213/ae3084

    Lin , S., Li , Y., Genel , S., et al. 2026, title One Latent to Fit Them All: A Unified Representation of Baryonic Feedback on Matter Distribution , , 996, L41, 10.3847/2041-8213/ae3084

    work page doi:10.3847/2041-8213/ae3084 2026

  45. [45]

    LSST Science Book, Version 2.0

    LSST Science Collaboration , Abell , P. A., Allison , J., et al. 2009, title LSST Science Book, Version 2.0 , arXiv e-prints, arXiv:0912.0201, 10.48550/arXiv.0912.0201

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.0912.0201 2009

  46. [46]

    D., Navarro , J

    Ludlow , A. D., Navarro , J. F., Angulo , R. E., et al. 2014, title The mass-concentration-redshift relation of cold dark matter haloes , , 441, 378, 10.1093/mnras/stu483

    work page doi:10.1093/mnras/stu483 2014

  47. [47]

    First results from the IllustrisTNG simulations: radio haloes and magnetic fields

    Marinacci , F., Vogelsberger , M., Pakmor , R., et al. 2018, title First results from the IllustrisTNG simulations: radio haloes and magnetic fields , , 480, 5113, 10.1093/mnras/sty2206

    work page internal anchor Pith review doi:10.1093/mnras/sty2206 2018

  48. [48]

    P., Pillepich , A., Springel , V., et al

    Naiman , J. P., Pillepich , A., Springel , V., et al. 2018, title First results from the IllustrisTNG simulations: a tale of two elements - chemical evolution of magnesium and europium , , 477, 1206, 10.1093/mnras/sty618

    work page internal anchor Pith review doi:10.1093/mnras/sty618 2018

  49. [49]

    2018, title First results from the IllustrisTNG simulations: the galaxy colour bimodality , , 475, 624, 10.1093/mnras/stx3040

    Nelson , D., Pillepich , A., Springel , V., et al. 2018, title First results from the IllustrisTNG simulations: the galaxy colour bimodality , , 475, 624, 10.1093/mnras/stx3040

    work page internal anchor Pith review doi:10.1093/mnras/stx3040 2018

  50. [50]

    2019, title The IllustrisTNG simulations: public data release , Computational Astrophysics and Cosmology, 6, 2, 10.1186/s40668-019-0028-x

    Nelson , D., Springel , V., Pillepich , A., et al. 2019, title The IllustrisTNG simulations: public data release , Computational Astrophysics and Cosmology, 6, 2, 10.1186/s40668-019-0028-x

    work page doi:10.1186/s40668-019-0028-x 2019

  51. [51]

    2022, title The ASTRID simulation: the evolution of supermassive black holes , , 513, 670, 10.1093/mnras/stac351

    Ni , Y., Di Matteo , T., Bird , S., et al. 2022, title The ASTRID simulation: the evolution of supermassive black holes , , 513, 670, 10.1093/mnras/stac351

    work page doi:10.1093/mnras/stac351 2022

  52. [52]

    2023, title The CAMELS Project: Expanding the Galaxy Formation Model Space with New ASTRID and 28-parameter TNG and SIMBA Suites , , 959, 136, 10.3847/1538-4357/ad022a

    Ni , Y., Genel , S., Angl \'e s-Alc \'a zar , D., et al. 2023, title The CAMELS Project: Expanding the Galaxy Formation Model Space with New ASTRID and 28-parameter TNG and SIMBA Suites , , 959, 136, 10.3847/1538-4357/ad022a

    work page doi:10.3847/1538-4357/ad022a 2023

  53. [53]

    N., et al

    Nicola , A., Villaescusa-Navarro , F., Spergel , D. N., et al. 2022, title Breaking baryon-cosmology degeneracy with the electron density power spectrum , , 2022, 046, 10.1088/1475-7516/2022/04/046

    work page doi:10.1088/1475-7516/2022/04/046 2022

  54. [54]

    2025, title ChatGPT (GPT-4.5), , https://openai.com

    OpenAI . 2025, title ChatGPT (GPT-4.5), , https://openai.com

    2025

  55. [55]

    P., et al

    Pakmor , R., Springel , V., Coles , J. P., et al. 2023, title The MillenniumTNG Project: the hydrodynamical full physics simulation and a first look at its galaxy clusters , , 524, 2539, 10.1093/mnras/stac3620

    work page doi:10.1093/mnras/stac3620 2023

  56. [56]

    A., Genel , S., Villaescusa-Navarro , F., et al

    Perez , L. A., Genel , S., Villaescusa-Navarro , F., et al. 2023, title Constraining Cosmology with Machine Learning and Galaxy Clustering: The CAMELS-SAM Suite , , 954, 11, 10.3847/1538-4357/accd52

    work page doi:10.3847/1538-4357/accd52 2023

  57. [57]

    2018 a , title First results from the IllustrisTNG simulations: the stellar mass content of groups and clusters of galaxies , , 475, 648, 10.1093/mnras/stx3112

    Pillepich , A., Nelson , D., Hernquist , L., et al. 2018 a , title First results from the IllustrisTNG simulations: the stellar mass content of groups and clusters of galaxies , , 475, 648, 10.1093/mnras/stx3112

    work page internal anchor Pith review doi:10.1093/mnras/stx3112 2018

  58. [58]

    Simulating Galaxy Formation with the IllustrisTNG Model

    Pillepich , A., Springel , V., Nelson , D., et al. 2018 b , title Simulating galaxy formation with the IllustrisTNG model , , 473, 4077, 10.1093/mnras/stx2656

    work page internal anchor Pith review doi:10.1093/mnras/stx2656 2018

  59. [59]

    10.1051/0004-6361/201833910

    Planck Collaboration , Aghanim , N., Akrami , Y., et al. 2020, title Planck 2018 results. VI. Cosmological parameters , , 641, A6, 10.1051/0004-6361/201833910

    work page doi:10.1051/0004-6361/201833910 2020

  60. [60]

    , keywords =

    Power , C., & Knebe , A. 2006, title The impact of box size on the properties of dark matter haloes in cosmological simulations , , 370, 691, 10.1111/j.1365-2966.2006.10562.x

    work page doi:10.1111/j.1365-2966.2006.10562.x 2006

  61. [61]

    G., & Madau , P

    Puchwein , E., Haardt , F., Haehnelt , M. G., & Madau , P. 2019, title Consistent modelling of the meta-galactic UV background and the thermal/ionization history of the intergalactic medium , , 485, 47, 10.1093/mnras/stz222

    work page doi:10.1093/mnras/stz222 2019

  62. [62]

    H., Rai c evi \'c , M., & Schaye , J

    Rahmati , A., Pawlik , A. H., Rai c evi \'c , M., & Schaye , J. 2013, title On the evolution of the H I column density distribution in cosmological simulations , , 430, 2427, 10.1093/mnras/stt066

    work page doi:10.1093/mnras/stt066 2013

  63. [63]

    2015, title The merger rate of galaxies in the Illustris simulation: a comparison with observations and semi-empirical models , , 449, 49, 10.1093/mnras/stv264

    Rodriguez-Gomez , V., Genel , S., Vogelsberger , M., et al. 2015, title The merger rate of galaxies in the Illustris simulation: a comparison with observations and semi-empirical models , , 449, 49, 10.1093/mnras/stv264

    work page internal anchor Pith review doi:10.1093/mnras/stv264 2015

  64. [64]

    G., Kwan , J., Upadhye , A., & Font , A

    Salcido , J., McCarthy , I. G., Kwan , J., Upadhye , A., & Font , A. S. 2023, title SP(k) - a hydrodynamical simulation-based model for the impact of baryon physics on the non-linear matter power spectrum , , 523, 2247, 10.1093/mnras/stad1474

    work page doi:10.1093/mnras/stad1474 2023

  65. [65]

    , archivePrefix = "arXiv", eprint =

    Schaye , J., Dalla Vecchia , C., Booth , C. M., et al. 2010, title The physics driving the cosmic star formation history , , 402, 1536, 10.1111/j.1365-2966.2009.16029.x

    work page doi:10.1111/j.1365-2966.2009.16029.x 2010

  66. [66]

    , archivePrefix = "arXiv", eprint =

    Schaye , J., Crain , R. A., Bower , R. G., et al. 2015, title The EAGLE project: simulating the evolution and assembly of galaxies and their environments , , 446, 521, 10.1093/mnras/stu2058

    work page doi:10.1093/mnras/stu2058 2015

  67. [67]

    2023, title The FLAMINGO project: cosmological hydrodynamical simulations for large-scale structure and galaxy cluster surveys , , 526, 4978, 10.1093/mnras/stad2419

    Schaye , J., Kugel , R., Schaller , M., et al. 2023, title The FLAMINGO project: cosmological hydrodynamical simulations for large-scale structure and galaxy cluster surveys , , 526, 4978, 10.1093/mnras/stad2419

    work page doi:10.1093/mnras/stad2419 2023

  68. [68]

    2016, title Matter power spectrum and the challenge of percent accuracy , , 2016, 047, 10.1088/1475-7516/2016/04/047

    Schneider , A., Teyssier , R., Potter , D., et al. 2016, title Matter power spectrum and the challenge of percent accuracy , , 2016, 047, 10.1088/1475-7516/2016/04/047

    work page doi:10.1088/1475-7516/2016/04/047 2016

  69. [69]

    , year =

    Scoccimarro , R. 1998, title Transients from initial conditions: a perturbative analysis , , 299, 1097, 10.1046/j.1365-8711.1998.01845.x

    work page doi:10.1046/j.1365-8711.1998.01845.x 1998

  70. [70]

    2023, title Robust Field-level Inference of Cosmological Parameters with Dark Matter Halos , , 944, 27, 10.3847/1538-4357/acac7a

    Shao , H., Villaescusa-Navarro , F., Villanueva-Domingo , P., et al. 2023, title Robust Field-level Inference of Cosmological Parameters with Dark Matter Halos , , 944, 27, 10.3847/1538-4357/acac7a

    work page doi:10.3847/1538-4357/acac7a 2023

  71. [71]

    Sims , X., Angl \'e s-Alc \'a zar , D., Oh , B.-K., et al. 2026, title CAMELS Environments: The Impact of Local Neighbours on Galaxy Evolution across the SIMBA, IllustrisTNG, ASTRID, and Swift-EAGLE Simulations , arXiv e-prints, arXiv:2601.06290, 10.48550/arXiv.2601.06290

    work page doi:10.48550/arxiv.2601.06290 2026

  72. [72]

    Smith , L. N. 2015, title Cyclical Learning Rates for Training Neural Networks , arXiv e-prints, arXiv:1506.01186, 10.48550/arXiv.1506.01186

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1506.01186 2015

  73. [73]

    and Sarajedini, Ata and Geisler, Doug and Bica, Eduardo and Claria, Juan J

    Sokasian , A., Abel , T., & Hernquist , L. 2002, title The epoch of helium reionization , , 332, 601, 10.1046/j.1365-8711.2002.05291.x

    work page doi:10.1046/j.1365-8711.2002.05291.x 2002

  74. [74]

    Sorini , D., Bose , S., Pakmor , R., et al. 2025, title The impact of baryons on the internal structure of dark matter haloes from dwarf galaxies to superclusters in the redshift range 0 < z < 7 , , 536, 728, 10.1093/mnras/stae2613

    work page doi:10.1093/mnras/stae2613 2025

  75. [75]

    , archivePrefix = "arXiv", eprint =

    Springel , V. 2010, title E pur si muove: Galilean-invariant cosmological hydrodynamical simulations on a moving mesh , , 401, 791, 10.1111/j.1365-2966.2009.15715.x

    work page doi:10.1111/j.1365-2966.2009.15715.x 2010

  76. [76]

    S., & White , S

    Springel , V., Frenk , C. S., & White , S. D. M. 2006, title The large-scale structure of the Universe , , 440, 1137, 10.1038/nature04805

    work page doi:10.1038/nature04805 2006

  77. [77]

    The stellar population synthesis models

    Springel , V., & Hernquist , L. 2003, title Cosmological smoothed particle hydrodynamics simulations: a hybrid multiphase model for star formation , , 339, 289, 10.1046/j.1365-8711.2003.06206.x

    work page doi:10.1046/j.1365-8711.2003.06206.x 2003

  78. [78]

    Springel , V., White , S. D. M., Tormen , G., & Kauffmann , G. 2001, title Populating a cluster of galaxies - I. Results at [formmu2]z=0 , , 328, 726, 10.1046/j.1365-8711.2001.04912.x

    work page doi:10.1046/j.1365-8711.2001.04912.x 2001

  79. [79]

    Springel , V., White , S. D. M., Jenkins , A., et al. 2005, title Simulations of the formation, evolution and clustering of galaxies and quasars , , 435, 629, 10.1038/nature03597

    work page doi:10.1038/nature03597 2005

  80. [80]

    2018, title First results from the IllustrisTNG simulations: matter and galaxy clustering , , 475, 676, 10.1093/mnras/stx3304

    Springel , V., Pakmor , R., Pillepich , A., et al. 2018, title First results from the IllustrisTNG simulations: matter and galaxy clustering , , 475, 676, 10.1093/mnras/stx3304

    work page doi:10.1093/mnras/stx3304 2018

Showing first 80 references.