Pith. sign in

REVIEW 4 major objections 6 minor 3 cited by

A COMPASS to Model Comparison and Simulation-Based Inference in Galactic Chemical Evolution

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read COMPASS, a neural diffusion framework for simulation-based inference, ranks 40 nucleosynthetic yield sets against 69 stars and awards the NuGrid AGB plus TNG CC-SN yields combination almost all posterior probability.

desk verdict A useful SBI engineering contribution whose headline model ranking is not supported by the statistical calculation as written. read the letter →

arxiv 2507.05060 v2 pith:PR2RM7UI submitted 2025-07-07 astro-ph.GA astro-ph.IMcs.LGphysics.comp-phphysics.data-an

Berkay Gunes , Sven Buder , Tobias Buck This is my paper · ORCID
classification astro-ph.GAastro-ph.IMcs.LGphysics.comp-phphysics.data-an
keywords simulation-basedinferencegalacticchemicalevolutionscore-baseddiffusionmodelstransformersBayesianmodelcomparisonnucleosyntheticyieldsinitialmassfunctionTypeIasupernovae
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

COMPASS is a simulation-based inference framework that pairs score-based diffusion models with transformer networks to do Bayesian parameter estimation and model comparison for galactic chemical evolution directly from stellar abundance data. Applied to 69 solar-type stars with high-precision abundances, it ranks 40 combinations of nucleosynthetic yield tables and reports that one combination — NuGrid AGB yields paired with the core-collapse supernova yields used in the IllustrisTNG simulation — accumulates posterior probability near 100%. Under that preferred model the paper infers a high-mass IMF slope $\alpha_{\mathrm{IMF}} = -2.60 \pm 0.02$, steeper than the canonical Salpeter value of $-2.35$, and an elevated Type Ia supernova normalization $\log_{10} N_{\mathrm{Ia}} = -2.78 \pm 0.02$. The broader claim is that amortized neural inference — networks trained once and then applied instantly — can replace expensive MCMC in this setting and make principled model selection over competing simulators practical.

What carries the argument

The central machinery is the conditional score-based diffusion transformer: a DiT-style network with adaLN-Zero conditioning and a binary observation mask that lets latent tokens attend only to observed dimensions while preventing latent–latent leakage. The same network runs in both inference directions — with the mask set to observe abundances it estimates $p(\theta \mid x, M_i)$, and with the mask inverted it draws likelihood samples at the MAP point, which a Gaussian KDE turns into the maximized likelihood for model comparison. The comparison itself is an AIC-style softmax, and because every competing yield-set model has the same number of parameters the penalty cancels, leaving $P(M_i \mid x) = L(x \mid \hat{\theta}_i, M_i) \big/ \sum_j L(x \mid \hat{\theta}_j, M_j)$.

What would settle it

Recompute the comparison with a joint treatment — for instance importance-sampling the global parameters out of the full 69-star likelihood under each yield set — and check whether NuGrid AGB plus TNG CC-SN yields still dominate; if any competing yield set gains non-negligible posterior mass, the reported near-100% preference is an artifact of the per-star MAP approximation. A complementary check is to run the same per-star procedure on mock data generated from a known losing yield set and measure how often the cumulative probability spuriously converges to near unity at 69 stars.

Watch

Extended reading notes

Core claim

Each candidate GCE model is treated as a separate simulator, and COMPASS trains a conditional score-based diffusion model on joint parameter–abundance pairs, with a binary mask that switches the same network between neural posterior estimation and neural likelihood estimation. For every star the network produces a posterior over six parameters — the two shared global parameters $\alpha_{\mathrm{IMF}}$ and $\log_{10} N_{\mathrm{Ia}}$, three per-star local parameters (star-formation efficiency, SFR peak time, outflow fraction), and a birth time — and the likelihood at the per-star MAP estimate is converted by a Gaussian KDE into a maximized log-likelihood. Because all 40 yield-set models have equal parameter counts, the AIC complexity penalty cancels and the model posterior reduces to a softmax over the maximized likelihoods, with per-star contributions multiplied into the cumulative probability shown across the 69-star sample. The paper's two headline findings are the near-unity cumulative preference for NuGrid AGB plus TNG CC-SN yields and the sharply prior-separated constraints $\alpha_{\mathrm{IMF}} = -2.60 \pm 0.02$ and $\log_{10} N_{\mathrm{Ia}} = -2.78 \pm 0.02$.

Load-bearing premise

The ranking treats the 69 stars as independent and combines per-star likelihoods evaluated at each star's own best-fit values of the two shared global parameters, so the near-unity cumulative posterior probability is a composite-likelihood quantity rather than a rigorously derived joint Bayesian evidence.

Editorial extensions

If this is right

  • If the ranking is correct, nucleosynthetic yield prescriptions can be chosen by data: the NuGrid AGB plus TNG CC-SN combination becomes the empirically preferred calibration for one-zone galactic chemical evolution modeling.
  • A high-mass IMF slope of $\alpha_{\mathrm{IMF}} = -2.60 \pm 0.02$ means fewer very massive stars per unit stellar mass than a Salpeter slope, shifting predicted metal yields and feedback budgets in galaxy formation simulations.
  • An SN Ia normalization of $\log_{10} N_{\mathrm{Ia}} = -2.78 \pm 0.02$ raises the iron production budget of the preferred model, tying observed abundance ratios more tightly to SN Ia delay-time distribution choices.
  • Because inference is amortized after training, the same networks can be applied to new stars at negligible marginal cost, letting model comparison scale to much larger spectroscopic samples without rerunning MCMC.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The near-unity cumulative posterior is a composite-likelihood quantity: the paper multiplies per-star maximized likelihoods evaluated at star-specific MAP values of the shared global parameters, which is not the same as the joint marginal likelihood of the full dataset, so the ~100% figure is a statement of ranking strength rather than calibrated Bayesian evidence.
  • The paper itself states in Appendix D that the per-star Gaussian posterior approximation used to combine stars can underestimate uncertainties for large samples, which reinforces reading the cumulative probability comparatively rather than as an exact evidence value.
  • The conditional-mask, train-once-run-both-ways design is not tied to galactic chemistry; the same architecture could rank competing simulators in any field where observations arrive partial and variable-sized.
  • A testable extension would be to rerun the same 40-model comparison on a larger stellar sample; a genuine yield-set preference should sharpen as the sample grows, while a composite-likelihood artifact would be more likely to waver.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper introduces COMPASS, a simulation-based inference framework built on a score-based diffusion model and transformer architecture, and applies it to compare 40 nucleosynthetic yield-table combinations for Galactic Chemical Evolution using high-precision abundances from 69 stars in Nissen et al. (2020). The authors report a near-unity cumulative posterior probability for the combination of NuGrid AGB yields with IllustrisTNG core-collapse SN yields, and infer an IMF slope αIMF = -2.60 ± 0.02 and a SN Ia normalization log10 NIa = -2.78 ± 0.02 under that model. Model comparison is performed through AIC-derived weights, which are interpreted as Bayesian posterior probabilities; multi-star evidence is accumulated by multiplying per-star likelihoods evaluated at star-specific MAP estimates of the global parameters. The appendices describe the network architecture, calibration diagnostics, mock-data validation, and benchmarks against Hamiltonian Monte Carlo and a prior SBI pipeline.

Significance. The framework is technically ambitious and the validation effort is substantial: the authors provide TARP calibration, compare against HMC and an earlier SBI method, and test on both CHEMPY mock data and IllustrisTNG simulation particles. If the model ranking and parameter constraints survive a statistically correct analysis, the astrophysical conclusions would be of genuine interest to the GCE community, since yield-table selection and IMF/SN Ia constraints are long-standing problems. However, the headline quantitative claims rest on two linked statistical approximations that the paper explicitly acknowledges in Appendix D as potentially producing overconfident results. The qualitative ranking may well be correct, but the present derivation does not establish it as a Bayesian posterior model probability, and the reported ±0.02 uncertainties are not supported by the acknowledged approximations.

major comments (4)
  1. [Appendix A, Eq. (8)] The derivation equates Akaike weights with posterior model probabilities and obtains P(M_i|x) = L(x|θhat_i,M_i) / Σ_j L(x|θhat_j,M_j) by cancelling equal parameter counts. This algebra is valid only for AIC weights, which are not Bayesian posterior model probabilities: they do not integrate over parameter prior volume and only asymptotically estimate relative expected KL predictive accuracy. Interpreting Eq. (8) as a posterior model probability is therefore unjustified. The authors should either compute proper marginal likelihoods (e.g., by likelihood emulation plus prior integration) or explicitly relabel the quantities as Akaike weights and revise all Bayesian language in Section 5 and the abstract accordingly.
  2. [Section 5, bottom panel of Fig. 2] The cumulative model posterior P(M_k|x_0,...,x_i) is formed by combining per-star probabilities that are each evaluated at a star-specific MAP estimate θhat_i of the shared global parameters Λ=(αIMF, log10 NIa). Since Λ is common to all stars, the product of per-star maximized likelihoods is not the joint likelihood of the full dataset under any single parameter vector; it is a composite likelihood with Λ profiled out per star. The resulting near-unity probability is therefore not a valid joint Bayesian posterior and is likely overconfident. The paper itself states in Appendix D that the analogous Gaussian posterior combination 'can lead to underestimated uncertainties and over-confident constraints in the regime of large stellar samples,' yet Section 5 presents the near-unity result without this caveat. A joint analysis that respects the shared Λ, or at least a sensitivity test of the ranking under a valid composite-likelihood correction, is needed before the ranking can be taken as established.
  3. [Section 5, Table 4] The reported constraints αIMF = -2.60 ± 0.02 and log10 NIa = -2.78 ± 0.02 depend on the same per-star Gaussian combination that Appendix D identifies as potentially overconfident. These uncertainties reflect only a single yield-set choice and do not include systematic differences among the top six yield sets, whose αIMF values range from -2.71 to -2.52 in the same table, nor the yield-misspecification shifts shown in Appendix D, Fig. 9. The paper should provide a systematic error budget or otherwise temper the precision claim, since ±0.02 is far smaller than the spread among the competing yield sets.
  4. [Appendix B.6, Eq. (12)] The significance test defines K = L_Mj(x)/L_H0(x) and labels it the 'Bayes Factor'. This is a maximum-likelihood ratio, not a Bayes factor, which requires marginal likelihoods after integrating over parameter priors. This repeats the same maximization-versus-marginalization conflation seen in Eq. (8). The test should either be implemented with proper marginal likelihoods or removed, and the text should not refer to it as a Bayes factor.
minor comments (6)
  1. [Section 3] The phrase 'the chemical composition of stars is typical measured' should read 'typically measured'.
  2. [Figure 1] The notation 'SBIm' for the inference model is not defined in the text or caption; please spell out the abbreviation.
  3. [Section 4] The text says the network approximates p(θ|X,M) but also that it operates as both a Neural Likelihood Estimator and a Neural Posterior Estimator; the relationship between the trained score model and these two uses should be clarified in one or two sentences.
  4. [Section 4 and Appendix B] The prior is described as a uniform range of ±5σPrior centered on the prior means, but the actual numerical values of the prior means and standard deviations for the six parameters are not given; they are needed for reproducibility and to interpret the claimed separation from the prior.
  5. [Table 3] The column headers such as 'Chieffi Net', 'Nomoto Net', and 'Net' are not defined anywhere in the text; please define these yield-table abbreviations in the caption or table notes.
  6. [Appendix C, Fig. 7] The statement that the cumulative posterior probability 'reaches 100%' after ten observations is a sign of the overconfidence discussed in the major comments and should be reported with finite precision (for example, >0.999) together with a caveat about the approximate combination method.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the yield-set ranking and parameter posteriors are produced by applying separately trained simulation-based estimators to real data, not by re-using a fitted quantity as its own prediction.

full rationale

The central derivation chain is self-contained against external benchmarks. For each of the 40 yield combinations, COMPASS trains an independent score-based diffusion posterior/likelihood estimator on 10^6 CHEMPY mock pairs (Section 4, Training); the real Nissen et al. (2020) abundances are then evaluated only at inference time, yielding per-star maximized-likelihood estimates that enter Eq. (8) as normalized AIC weights. The preferred yield set (NuGrid AGB + TNG CC-SN) and the reported (alpha_IMF, log10 NIa) values are therefore outputs of a data-driven comparison, not identities to any fitted input. Benchmarking against HMC (Philcox & Rybizki 2019), Buck et al. (2025), and TNG mock data in Appendices C and D supplies independent validation; self-citations are contextual rather than load-bearing. The main statistical concern in the paper - that Eq. (8) relabels Akaike weights as 'posterior model probabilities' and that the cumulative curve multiplies per-star ratios as if conditionally independent, which Appendix D itself concedes can produce overconfident constraints for large samples - is a validity and calibration issue, not an input-output circularity. The qualitative ranking could in principle be affected by the composite-likelihood approximation, but no equation in the paper forces the result to equal its inputs by construction.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central claims depend on a chain of modeling choices: the CHEMPY simulator, the yield tables, the prior distributions, the noise model, and the AIC-based comparison via KDE likelihood estimates. The paper does not release code or data, and the KDE step is under-specified, so the reader would need to re-implement or trust the pipeline.

free parameters (5)
  • Likelihood noise level (5% Gaussian) = σ = 0.05
    Chosen to match typical Nissen et al. (2020) uncertainties; directly controls the scale of the likelihood and hence posterior widths and model probabilities, but is not derived from the data itself or the model.
  • Prior range (±5σPrior) = ±5σ around Rybizki et al. (2017) means
    The uniform prior support for the six parameters is fixed to ±5σ centered on prior means; this truncation can affect the posterior if the true parameter lies outside the support.
  • Diffusion network hyperparameters = sigma=2.5, depth=5, hidden=65, MLP ratio=3, 1 head
    Tuned via Optuna on validation criteria; the approximation error of the score model affects both posterior samples and likelihood estimates.
  • KDE bandwidth and sample count for likelihood estimation = Not specified
    The paper does not report how the KDE is constructed (bandwidth, number of samples) nor the resulting uncertainty in the estimated log-likelihoods used for model comparison; this is a hidden degree of freedom.
  • Diffusion steps for model comparison = 50 steps (vs 500 for parameter inference)
    The paper reduces to 50 steps for model comparison to save time, which may lower likelihood accuracy (Appendix C); the resulting log-likelihood estimates carry unknown error.
assumptions (6)
  • domain assumption CHEMPY is an adequate forward model for the solar-neighborhood abundances.
    The framework trains on CHEMPY-generated mock data; if the one-zone GCE model omits processes relevant to the Nissen stars, both model comparison and parameter inference are biased. The paper partially tests this with IllustrisTNG data (Appendix D) but does not validate on solar-neighborhood data with a different forward model.
  • domain assumption The candidate set of 40 yield combinations contains the true yield model (model completeness).
    Model comparison only selects among the given yield sets; if the true enrichment process is not in the set, the near-unity posterior is a relative selection, not absolute evidence. This is not discussed in the paper.
  • domain assumption The prior distributions (from Rybizki et al. 2017) are reasonable and the ±5σ range covers the plausible parameter space.
    Posteriors are computed under these priors; if the prior means are far from truth or the range is too narrow, the constraints are biased.
  • standard math Akaike weights provide valid posterior model probabilities when all models have the same parameter count.
    The paper simplifies AIC to a likelihood ratio (Eq. 8) and treats it as a Bayesian posterior; this is an information-theoretic heuristic, not a true marginal likelihood, and requires regularity conditions not verified here.
  • domain assumption Stellar abundances in the Nissen et al. (2020) sample are independent given the model.
    The cumulative model posterior multiplies per-star likelihoods; if stars share environmental correlations (e.g., birth in same clusters), the effective sample size is smaller and posterior overconfident.
  • domain assumption The diffusion model provides an unbiased estimate of the conditional likelihood p(x|θ,M).
    The model comparison relies on KDE-estimated likelihoods from the trained diffusion model; the paper provides posterior calibration for p(θ|x) but not direct validation of the likelihood estimates, whose bias could affect model rankings.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A COMPASS to Model Comparison and Simulation-Based Inference in Galactic Chemical Evolution." pith.science (2026). https://pith.science/paper/PR2RM7UI

@misc{pith2026250705060,
  author       = {Pith},
  title        = {Pith review of: A COMPASS to Model Comparison and Simulation-Based Inference in Galactic Chemical Evolution},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PR2RM7UI}},
  note         = {Machine review of arXiv:2507.05060}
}
read the original abstract

We present COMPASS, a novel simulation-based inference framework that combines score-based diffusion models with transformer architectures to jointly perform parameter estimation and Bayesian model comparison across competing Galactic Chemical Evolution (GCE) models. COMPASS handles high-dimensional, incomplete, and variable-size stellar abundance datasets. Applied to high-precision elemental abundance measurements, COMPASS evaluates 40 combinations of nucleosynthetic yield tables. The model strongly favours Asymptotic Giant Branch yields from NuGrid and core-collapse SN yields used in the IllustrisTNG simulation, achieving near-unity cumulative posterior probability. Using the preferred model, we infer a steep high-mass IMF slope and an elevated Supernova Ia normalization, consistent with prior solar neighbourhood studies but now derived from fully amortized Bayesian inference. Our results demonstrate that modern SBI methods can robustly constrain uncertain physics in astrophysical simulators and enable principled model selection when analysing complex, simulation-based data.

Figures

Figures reproduced from arXiv: 2507.05060 by the authors.

Figure 1
Figure 1. Flow chart of our model comparison workflow. Each of the 40 candidate models (Mi) is used to train a separate Score-Based Inference Model (SBIm). For a given observation xobs, the corresponding trained SBIm infers the posterior P(θ|xobs) to find the MAP parameters ˆθ. The same model is then used to estimate the maximized likelihood L(xobs| ˆθi,Mi). The set of likelihoods from all 40 models is then used to derive the… view at source ↗
Figure 3
Figure 3. Inferred Galactic Parameters from Observational Data Joint posterior contours (1σ, 2σ) for the IMF high-mass slope αIMF and SN Ia normalization log10 NIa, from Nissen et al. (2020) using COMPASS. Colors denote the top six yield set combinations identified in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Accuracy of SDE-Solvers Comparison of SDE solver performance in terms of predictive accuracy (− log P(θ|x)) versus inference time per sample. Euler-Maruyama and DPM-Solvers (1st, 2nd, and 3rd order) were tested with varying numbers of diffusion steps (indicated by colored points, ranging from 5 to 5000 steps). Accuracy is averaged over 1000 mock observations. The dashed horizontal line at − log P(θ|x) ≈ 0.693 repres… view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5: Posterior calibration diagnostics showing the TARP plots at the top and the true vs. predicted parameter plots for each of the six parameters at the bottom and the TARP over all parameters on the right To ensure reliable uncertainty quantification, we evaluate posterio…
Figure 6
Figure 6. Figure 6: Violin plots showing the distribution of single-observation posterior probabilities P(Mk|xi) for three competing CC-SN yield models (Chieffi, Limongi and West), when tested on mock data. Each panel corresponds to a different ground-truth model used to generate the mock…
Figure 7
Figure 7. Figure 7: Cumulative relative model posterior probability P(Mk|x0, . . . , xi) as a function of the number of combined mock observations i. This example illustrates the case where the mock data was generated using the Limongi & Chieffi (2018) CC-SN yields (corresponding to the m…
Figure 8
Figure 8. Figure 8: Parameter Inference with Matched TNG Yields. Comparison of inference results for the global galactic parameters Λ using COMPASS (green), SBI (blue) and HMC (red) on mock data generated with TNG yields. (Left): Joint posterior distribution P(Λ|x) inferred from 200 stars…
Figure 9
Figure 9. Figure 9: Parameter Inference Alternative Yields Comparison of inference results for the global galactic parameters Λ using COMPASS (green), SBI (blue) and HMC (red) on mock data generated with the alternative yield sets (Tab. 2) (Left): Joint posterior distribution P(Λ|x) infer…
Figure 10
Figure 10. Figure 10: Parameter Inference from IllustrisTNG Simulation Comparison of inference results for the global galactic parameters Λ using COMPASS (green), SBI (blue) and HMC (red) on stellar abundances from the IllustrisTNG simulated galaxy. (Left): Joint posterior distribution P(Λ…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Hierarchical Bayesian inference with compositional score modeling for stellar streams

    astro-ph.GA 2026-07 conditional novelty 6.0 of 10

    Combining three stellar streams and the rotation curve with compositional score modeling yields a posterior for the Milky Way potential—mildly oblate inner halo—but the underlying simulator fails a misspecification test.

  2. Amortized Simulation-Based Inference of Colliding-Wind Binaries from Short, Noisy Image Time Series

    astro-ph.SR 2026-06 unverdicted novelty 6.0 of 10

    A neural spline-flow posterior estimator with a factorized spatio-temporal encoder recovers mass-loss rates and orbital parameters of colliding-wind binaries from 10-frame synthetic Hα photon-count time series.

  3. The chemodynamical memory of a major merger in a NIHAO-UHD Milky Way analogue -- I. A golden thread through time and space

    astro-ph.GA 2025-10 conditional novelty 6.0 of 10

    In a Milky Way analogue simulation, accreted stars from the last major merger retain a measurable memory of their birth radius: core-born stars are more bound and metal-rich, outer-born stars less bound and metal-poor...

Reference graph

Works this paper leans on

76 extracted references · 23 canonical work pages · cited by 3 Pith papers

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    I., Ryde , N., Andersson , E

    Agertz , O., Renaud , F., Feltzing , S., Read , J. I., Ryde , N., Andersson , E. P., Rey , M. P., Bensby , T., and Feuillet , D. K. VINTERGATAN - I. The origins of chemically, kinematically, and structurally distinct discs in a simulated Milky Way-mass galaxy . , 503 0 (4): 0 5826--5845, June 2021. doi:10.1093/mnras/stab322

    doi:10.1093/mnras/stab322 2021

  3. [3]

    On newer statistical approaches to parameter estimation and structure determination

    Akaike, H. On newer statistical approaches to parameter estimation and structure determination. IFAC Proceedings Volumes, 11 0 (1): 0 1877--1884, 1978. ISSN 1474-6670. doi:https://doi.org/10.1016/S1474-6670(17)66162-7. URL https://www.sciencedirect.com/science/article/pii/S1474667017661627. 7th Triennial World Congress of the IFAC on A Link Between Scienc...

    work page doi:10.1016/s1474-6670(17)66162-7 1978

  4. [4]

    A bayesian extension of the minimum aic procedure of autoregressive model fitting

    AKAIKE, H. A bayesian extension of the minimum aic procedure of autoregressive model fitting. Biometrika, 66 0 (2): 0 237--242, 08 1979. ISSN 0006-3444. doi:10.1093/biomet/66.2.237. URL https://doi.org/10.1093/biomet/66.2.237

    work page doi:10.1093/biomet/66.2.237 1979

  5. [5]

    Akcay, S., Atapour-Abarghouei, A., and Breckon, T. P. Ganomaly: Semi-supervised anomaly detection via adversarial training. In Computer Vision--ACCV 2018: 14th Asian Conference on Computer Vision, Perth, Australia, December 2--6, 2018, Revised Selected Papers, Part III 14, pp.\ 622--637. Springer, 2019

    work page 2018

  6. [6]

    Optuna: A next-generation hyperparameter optimization framework, 2019

    Akiba, T., Sano, S., Yanase, T., Ohta, T., and Koyama, M. Optuna: A next-generation hyperparameter optimization framework, 2019. URL https://arxiv.org/abs/1907.10902

    arXiv 2019

  7. [7]

    H., Weinberg , D

    Andrews , B. H., Weinberg , D. H., Sch \"o nrich , R., and Johnson , J. A. Inflow, Outflow, Yields, and Stellar Population Mixing in Chemical Evolution Models . , 835 0 (2): 0 224, Feb 2017. doi:10.3847/1538-4357/835/2/224

    doi:10.3847/1538-4357/835/2/224 2017

  8. [8]

    J., and Scott , P

    Asplund , M., Grevesse , N., Sauval , A. J., and Scott , P. The Chemical Composition of the Sun . , 47 0 (1): 0 481--522, Sep 2009. doi:10.1146/annurev.astro.46.060407.145222

    arXiv 2009

Show all 76 references
  1. [9]

    P., Tollerud , E

    Astropy Collaboration , Robitaille , T. P., Tollerud , E. J., Greenfield , P., Droettboom , M., Bray , E., Aldcroft , T., Davis , M., Ginsburg , A., Price-Whelan , A. M., Kerzendorf , W. E., Conley , A., Crighton , N., Barbary , K., Muna , D., Ferguson , H., Grollier , F., Par...

    2013

  2. [10]

    M., West, C., and Heger, A

    Austin, S. M., West, C., and Heger, A. Reducing uncertainties in the production of the gamma-emitting nuclei 26al, 44ti, and 60fe in core-collapse supernovae by using effective helium burning rates. The Astrophysical Journal Letters, 839 0 (1): 0 L9, April 2017. ISSN 2041-8213...

    2017 doi

  3. [11]

    Model selection and akaike’s information criterion (aic): The general theory and its analytical extensions

    Bozdogan, H. Model selection and akaike’s information criterion (aic): The general theory and its analytical extensions. Psychometrika, 52 0 (3): 0 353–356, 1987. doi:10.1007/BF02294361

    1987 doi

  4. [12]

    On the origin of the chemical bimodality of disc stars: a tale of merger and migration

    Buck , T. On the origin of the chemical bimodality of disc stars: a tale of merger and migration . , 491 0 (4): 0 5435--5446, February 2020. doi:10.1093/mnras/stz3289

    2020 doi

  5. [13]

    V., Dutton, A

    Buck, T., Macci \`o , A. V., Dutton, A. A., Obreja, A., and Frings, J. Nihao xv: the environmental impact of the host galaxy on galactic satellite and field dwarf galaxies. Monthly Notices of the Royal Astronomical Society, 483 0 (1): 0 1314--1341, 2019

    2019

  6. [14]

    V., Minchev , I., Dutton , A

    Buck , T., Obreja , A., Macci \`o , A. V., Minchev , I., Dutton , A. A., and Ostriker , J. P. NIHAO-UHD: the properties of MW-like stellar discs in high-resolution cosmological simulations . , 491 0 (3): 0 3461--3478, January 2020. doi:10.1093/mnras/stz3241

    2020 doi

  7. [15]

    V., Pfrommer , C., Steinmetz , M., and Ness , M

    Buck , T., Rybizki , J., Buder , S., Obreja , A., Macci \`o , A. V., Pfrommer , C., Steinmetz , M., and Ness , M. The challenge of simultaneously matching the observed diversity of chemical abundance patterns in cosmological hydrodynamical simulations . , 508 0 (3): 0 3365--33...

    2021 doi

  8. [16]

    Buck , T., Obreja , A., Ratcliffe , B., Lu , Y., Minchev , I., and Macci \`o , A. V. The impact of early massive mergers on the chemical evolution of Milky Way-like galaxies: insights from NIHAO-UHD simulations . , 523 0 (1): 0 1565--1576, July 2023. doi:10.1093/mnras/stad1503

    2023 doi

  9. [17]

    H., and Buder, S

    Buck, T., Günes, B., Viterbo, G., Oliver, W. H., and Buder, S. Inferring galactic parameters from chemical abundances with simulation-based inference, 2025. URL https://arxiv.org/abs/2503.02456

    2025

  10. [18]

    and Anderson, D

    Burnham, K. and Anderson, D. Model selection and multimodel inference: a practical information-theoretic approach. Springer Verlag, 2002

    2002

  11. [19]

    Galactic Stellar and Substellar Initial Mass Function

    Chabrier , G. Galactic Stellar and Substellar Initial Mass Function . , 115: 0 763--795, July 2003. doi:10.1086/376392

    2003 doi

  12. [20]

    Variations of the Stellar Initial Mass Function in the Progenitors of Massive Early-type Galaxies and in Extreme Starburst Environments

    Chabrier , G., Hennebelle , P., and Charlot , S. Variations of the Stellar Initial Mass Function in the Progenitors of Massive Early-type Galaxies and in Extreme Starburst Environments . , 796 0 (2): 0 75, Dec 2014. doi:10.1088/0004-637X/796/2/75

    2014 doi

  13. [21]

    and Limongi , M

    Chieffi , A. and Limongi , M. Explosive Yields of Massive Stars from Z = 0 to Z = Z _ solar . , 608 0 (1): 0 405--410, June 2004. doi:10.1086/392523

    2004 doi

  14. [22]

    J., Debattista , V

    Clarke , A. J., Debattista , V. P., Nidever , D. L., Loebman , S. R., Simons , R. C., Kassin , S., Du , M., Ness , M., Fisher , D. B., Quinn , T. R., Wadsley , J., Freeman , K. C., and Popescu , C. C. The imprint of clump formation at high redshift - I. A disc -abundance dicho...

    2019 doi

  15. [23]

    W., Herwig , F., Pignatari , M., Jones , S., and Fryer , C

    C \^o t \'e , B., Ritter , C., O'Shea , B. W., Herwig , F., Pignatari , M., Jones , S., and Fryer , C. L. Uncertainties in Galactic Chemical Evolution Models . , 824 0 (2): 0 82, Jun 2016. doi:10.3847/0004-637X/824/2/82

    2016 doi

  16. [24]

    The frontier of simulation-based inference

    Cranmer, K., Brehmer, J., and Louppe, G. The frontier of simulation-based inference. Proceedings of the National Academy of Sciences, 117 0 (48): 0 30055--30062, 2020. doi:10.1073/pnas.1912789117. URL https://www.pnas.org/doi/abs/10.1073/pnas.1912789117

    2020 doi

  17. [25]

    L., Gil-Pons , P., Lau , H

    Doherty , C. L., Gil-Pons , P., Lau , H. H. B., Lattanzio , J. C., and Siess , L. Super and massive AGB stars - II. Nucleosynthesis and yields - Z = 0.02, 0.008 and 0.004 . , 437 0 (1): 0 195--214, Jan 2014. doi:10.1093/mnras/stt1877

    2014 doi

  18. [26]

    Elsemüller, L., Schnuerch, M., Bürkner, P.-C., and Radev, S. T. A deep learning method for comparing bayesian hierarchical models, 2023

    2023

  19. [27]

    K., Karakas , A

    Fishlock , C. K., Karakas , A. I., Lugaro , M., and Yong , D. Evolution and Nucleosynthesis of Asymptotic Giant Branch Stellar Models of Low Metallicity . , 797 0 (1): 0 44, Dec 2014. doi:10.1088/0004-637X/797/1/44

    2014 doi

  20. [28]

    S., McCarthy , I

    Font , A. S., McCarthy , I. G., Poole-Mckenzie , R., Stafford , S. G., Brown , S. T., Schaye , J., Crain , R. A., Theuns , T., and Schaller , M. The ARTEMIS simulations: stellar haloes of Milky Way-mass galaxies . , 498 0 (2): 0 1765--1785, October 2020. doi:10.1093/mnras/staa2463

    2020 doi

  21. [29]

    Gloeckler, M., Deistler, M., Weilbach, C., Wood, F., and Macke, J. H. All-in-one simulation-based inference, 2024. URL https://arxiv.org/abs/2404.09636

    2024 arXiv

  22. [30]

    and Dixon, P

    Glover, S. and Dixon, P. Likelihood ratios: A simple and flexible statistic for empirical psychologists. Psychonomic Bulletin & Review, 11 0 (5): 0 791--806, 2004. doi:10.3758/BF03196706. URL https://doi.org/10.3758/BF03196706

    2004 doi

  23. [31]

    Generative adversarial networks

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y. Generative adversarial networks. Communications of the ACM, 63 0 (11): 0 139--144, 2020

    2020

  24. [32]

    Grand , R. J. J., Bustamante , S., G \'o mez , F. A., Kawata , D., Marinacci , F., Pakmor , R., Rix , H.-W., Simpson , C. M., Sparre , M., and Springel , V. Origin of chemically distinct discs in the Auriga cosmological simulations . , 474: 0 3629--3639, March 2018. doi:10.109...

    2018 doi

  25. [33]

    F., Wetzel , A., Kere s , D., Faucher-Gigu \`e re , C.-A., Quataert , E., Boylan-Kolchin , M., Murray , N., Hayward , C

    Hopkins , P. F., Wetzel , A., Kere s , D., Faucher-Gigu \`e re , C.-A., Quataert , E., Boylan-Kolchin , M., Murray , N., Hayward , C. C., Garrison-Kimmel , S., Hummels , C., Feldmann , R., Torrey , P., Ma , X., Angl \'e s-Alc \'a zar , D., Su , K.-Y., Orr , M., Schmitz , D., E...

    2018

  26. [34]

    B., and Matteucci , F

    Jim \'e nez , N., Tissera , P. B., and Matteucci , F. Type Ia Supernova Progenitors and Chemical Enrichment in Hydrodynamical Simulations. I. The Single-degenerate Scenario . , 810 0 (2): 0 137, Sep 2015. doi:10.1088/0004-637X/810/2/137

    2015 doi

  27. [35]

    V., Faucher, N., Pasquato, M., Buck, T., Dixon, K

    Jin, Z., Macci \`o , A. V., Faucher, N., Pasquato, M., Buck, T., Dixon, K. L., Arora, N., Blank, M., and Vulanovic, P. Quantitatively rating galaxy simulations against real observations with anomaly detection. Monthly Notices of the Royal Astronomical Society, 529 0 (4): 0 353...

    2024

  28. [38]

    Karakas , A. I. and Lugaro , M. Stellar Yields from Metal-rich Asymptotic Giant Branch Models . , 825 0 (1): 0 26, July 2016 b . doi:10.3847/0004-637X/825/1/26

    2016 doi

  29. [39]

    Simsims: Simulation-based supernova ia model selection with thousands of latent variables

    Karchev, K., Trotta, R., and Weniger, C. Simsims: Simulation-based supernova ia model selection with thousands of latent variables. arXiv preprint arXiv:2311.15650, 2023

    2023 arXiv

  30. [40]

    Galactic chemical evolution: Carbon through zinc

    Kobayashi, C., Umeda, H., Nomoto, K., Tominaga, N., and Ohkubo, T. Galactic chemical evolution: Carbon through zinc. The Astrophysical Journal, 653 0 (2): 0 1145–1171, December 2006. ISSN 1538-4357. doi:10.1086/508914. URL http://dx.doi.org/10.1086/508914

    2006 doi

  31. [41]

    Sdss-v pioneering panoptic spectroscopy

    Kollmeier, J., Anderson, S., Blanc, G., Blanton, M., Covey, K., Crane, J., Drory, N., Frinchaboy, P., Froning, C., Johnson, J., et al. Sdss-v pioneering panoptic spectroscopy. Bulletin of the American Astronomical Society, 2019

    2019

  32. [42]

    and Chieffi, A

    Limongi, M. and Chieffi, A. Presupernova evolution and explosive nucleosynthesis of rotating massive stars in the metallicity range -3 [fe/h] 0 . The Astrophysical Journal Supplement Series, 237 0 (1): 0 13, July 2018. ISSN 1538-4365. doi:10.3847/1538-4365/aacb24. URL http://d...

    2018 doi

  33. [43]

    Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps, 2022

    Lu, C., Zhou, Y., Bao, F., Chen, J., Li, C., and Zhu, J. Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps, 2022. URL https://arxiv.org/abs/2206.00927

    2022 arXiv

  34. [44]

    T., Crain , R

    Mackereth , J. T., Crain , R. A., Schiavon , R. P., Schaye , J., Theuns , T., and Schaller , M. The origin of diverse -element abundances in galaxy discs . , 477 0 (4): 0 5072--5089, Jul 2018. doi:10.1093/mnras/sty972

    2018 doi

  35. [45]

    The Supernova Delay Time Distribution in Galaxy Clusters and Implications for Type-Ia Progenitors and Metal Enrichment

    Maoz , D., Sharon , K., and Gal-Yam , A. The Supernova Delay Time Distribution in Galaxy Clusters and Implications for Type-Ia Progenitors and Metal Enrichment . , 722 0 (2): 0 1879--1894, Oct 2010. doi:10.1088/0004-637X/722/2/1879

    2010 doi

  36. [46]

    Maoz , D., Mannucci , F., and Brandt , T. D. The delay-time distribution of Type Ia supernovae from Sloan II . , 426 0 (4): 0 3282--3294, Nov 2012. doi:10.1111/j.1365-2966.2012.21871.x

    2012

  37. [47]

    Chemodynamical evolution of the Milky Way disk

    Minchev , I., Chiappini , C., and Martig , M. Chemodynamical evolution of the Milky Way disk. I. The solar vicinity . , 558: 0 A9, October 2013. doi:10.1051/0004-6361/201220189

    2013 doi

  38. [48]

    Moll \'a , M., Cavichia , O., Gavil \'a n , M., and Gibson , B. K. Galactic chemical evolution: stellar yields and the initial mass function . , 451 0 (4): 0 3693--3708, Aug 2015. doi:10.1093/mnras/stv1102

    2015 doi

  39. [49]

    D., Romeijn, J.-W., and Rouder, J

    Morey, R. D., Romeijn, J.-W., and Rouder, J. N. The philosophy of bayes factors and the quantification of statistical evidence. Journal of Mathematical Psychology, 72: 0 6--18, 2016. ISSN 0022-2496. doi:https://doi.org/10.1016/j.jmp.2015.11.001. URL https://www.sciencedirect.c...

    2016 doi

  40. [50]

    E., Christensen-Dalsgaard, J., Mosumgaard, J

    Nissen, P. E., Christensen-Dalsgaard, J., Mosumgaard, J. R., Silva Aguirre, V., Spitoni, E., and Verma, K. High-precision abundances of elements in solar-type stars: Evidence of two distinct sequences in abundance-age relations. Astronomy & Astrophysics, 640: 0 A81, August 202...

    2020 doi

  41. [51]

    K., Brachwitz , F., Tsujimoto , T., Kubo , Y., and Kishimoto , N

    Nomoto , K., Iwamoto , K., Nakasato , N., Thielemann , F. K., Brachwitz , F., Tsujimoto , T., Kubo , Y., and Kishimoto , N. Nucleosynthesis in type Ia supernovae . , 621: 0 467--476, Feb 1997. doi:10.1016/S0375-9474(97)00291-1

    1997 doi

  42. [52]

    Nucleosynthesis in Stars and the Chemical Enrichment of Galaxies

    Nomoto , K., Kobayashi , C., and Tominaga , N. Nucleosynthesis in Stars and the Chemical Enrichment of Galaxies . , 51 0 (1): 0 457--509, Aug 2013. doi:10.1146/annurev-astro-082812-140956

    2013 doi

  43. [53]

    J., Mohamed, S., and Lakshminarayanan, B

    Papamakarios, G., Nalisnick, E., Rezende, D. J., Mohamed, S., and Lakshminarayanan, B. Normalizing flows for probabilistic modeling and inference, 2021

    2021

  44. [54]

    and Xie, S

    Peebles, W. and Xie, S. Scalable diffusion models with transformers, 2023. URL https://arxiv.org/abs/2212.09748

    2023 arXiv

  45. [55]

    Philcox, O., Rybizki, J., and Gutcke, T. A. On the optimal choice of nucleosynthetic yields, initial mass function, and number of sne ia for chemical evolution modeling. The Astrophysical Journal, 861 0 (1): 0 40, June 2018. ISSN 1538-4357. doi:10.3847/1538-4357/aac6e4. URL ht...

    2018 doi

  46. [56]

    Philcox, O. H. E. and Rybizki, J. Inferring galactic parameters from chemical abundances: A multi-star approach. The Astrophysical Journal, 887 0 (1): 0 9, December 2019. ISSN 1538-4357. doi:10.3847/1538-4357/ab5186. URL http://dx.doi.org/10.3847/1538-4357/ab5186

    2019 doi

  47. [57]

    Simulating galaxy formation with the IllustrisTNG model

    Pillepich , A., Springel , V., Nelson , D., Genel , S., Naiman , J., Pakmor , R., Hernquist , L., Torrey , P., Vogelsberger , M., Weinberger , R., and Marinacci , F. Simulating galaxy formation with the IllustrisTNG model . , 473: 0 4077--4106, January 2018. doi:10.1093/mnras/stx2656

    2018 doi

  48. [58]

    a nen , E., Keskitalo , R., Kisner , T. S., Kneissl , R., Knoche , J., Knox , L., Kunz , M., Kurki-Suonio , H., Lagache , G., L \

    Planck Collaboration , Ade , P. A. R., Aghanim , N., Arnaud , M., Ashdown , M., Aumont , J., Baccigalupi , C., Banday , A. J., Barreiro , R. B., Bartlett , J. G., Bartolo , N., Battaner , E., Battye , R., Benabed , K., Beno \^ t , A., Benoit-L \'e vy , A., Bernard , J. P., Ber...

    2015

  49. [59]

    Galactic chemical enrichment with new metallicity dependent yields, 1997

    Portinari, L., Chiosi, C., and Bressan, A. Galactic chemical enrichment with new metallicity dependent yields, 1997. URL https://arxiv.org/abs/astro-ph/9711337

    1997 arXiv

  50. [60]

    Galactic chemical enrichment with new metallicity dependent stellar yields

    Portinari , L., Chiosi , C., and Bressan , A. Galactic chemical enrichment with new metallicity dependent stellar yields . , 334: 0 505--539, Jun 1998

    1998

  51. [61]

    M., Sip ' o cz , B

    Price-Whelan , A. M., Sip ' o cz , B. M., G "u nther , H. M., Lim , P. L., Crawford , S. M., Conseil , S., Shupe , D. L., Craig , M. W., Dencheva , N., Ginsburg , A., VanderPlas , J. T., Bradley , L. D., P 'e rez-Su 'a rez , D., de Val-Borro , M., Paper Contributors , P., Aldc...

    2018

  52. [62]

    Nugrid stellar data set – ii

    Ritter, C., Herwig, F., Jones, S., Pignatari, M., Fryer, C., and Hirschi, R. Nugrid stellar data set – ii. stellar yields from h to bi for stellar models with mzams = 1–25 \ M _ and z = 0.0001–0.02. Monthly Notices of the Royal Astronomical Society, 480 0 (1): 0 538–571, June ...

    2018 doi

  53. [63]

    Quantifying the uncertainties of chemical evolution studies

    Romano , D., Chiappini , C., Matteucci , F., and Tosi , M. Quantifying the uncertainties of chemical evolution studies. I. Stellar lifetimes and initial mass function . , 430: 0 491--505, Feb 2005. doi:10.1051/0004-6361:20048222

    2005 doi

  54. [64]

    and Just , A

    Rybizki , J. and Just , A. Towards a fully consistent Milky Way disc model - III. Constraining the initial mass function . , 447 0 (4): 0 3880--3891, March 2015. doi:10.1093/mnras/stu2734

    2015 doi

  55. [65]

    Rybizki, J., Just, A., and Rix, H.-W. Chempy: A flexible chemical evolution model for abundance fitting: Do the sun’s abundances alone constrain chemical evolution models? Astronomy & Astrophysics, 605: 0 A59, September 2017. ISSN 1432-0746. doi:10.1051/0004-6361/201730522. UR...

    2017 doi

  56. [66]

    S., Fattahi , A., Navarro , J

    Sawala , T., Frenk , C. S., Fattahi , A., Navarro , J. F., Bower , R. G., Crain , R. A., Dalla Vecchia , C., Furlong , M., Helly , J. C., Jenkins , A., Oman , K. A., Schaller , M., Schaye , J., Theuns , T., Trayford , J., and White , S. D. M. The APOSTLE simulations: solutions...

    1931 doi

  57. [67]

    and Binney , J

    Sch \"o nrich , R. and Binney , J. Chemical evolution with radial mixing . , 396: 0 203--222, June 2009. doi:10.1111/j.1365-2966.2009.14750.x

    2009

  58. [68]

    R., Ciaraldi-Schoolmann, F., Röpke, F

    Seitenzahl, I. R., Ciaraldi-Schoolmann, F., Röpke, F. K., Fink, M., Hillebrandt, W., Kromer, M., Pakmor, R., Ruiter, A. J., Sim, S. A., and Taubenberger, S. Three-dimensional delayed-detonation models with nucleosynthesis for type ia supernovae. Monthly Notices of the Royal As...

    2012 doi

  59. [69]

    o flich , P., Liebend \

    Thielemann , F. K., Argast , D., Brachwitz , F., Hix , W. R., H \"o flich , P., Liebend \"o rfer , M., Martinez-Pinedo , G., Mezzacappa , A., Panov , I., and Rauscher , T. Nuclear cross sections, nuclear structure and stellar nucleosynthesis . , 718: 0 139--146, May 2003. doi:...

    2003 doi

  60. [70]

    Pixel recurrent neural networks

    Van Den Oord, A., Kalchbrenner, N., and Kavukcuoglu, K. Pixel recurrent neural networks. In International conference on machine learning, pp.\ 1747--1756. PMLR, 2016

    2016

  61. [71]

    N., Kaiser, L., and Polosukhin, I

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., and Polosukhin, I. Attention is all you need, 2023. URL https://arxiv.org/abs/1706.03762

    2023 arXiv

  62. [72]

    Vincenzo , F., Matteucci , F., Recchi , S., Calura , F., McWilliam , A., and Lanfranchi , G. A. The IGIMF and other IMFs in dSphs: the case of Sagittarius . , 449 0 (2): 0 1327--1339, May 2015. doi:10.1093/mnras/stv357

    2015 doi

  63. [73]

    Introducing the illustris project: simulating the coevolution of dark and visible matter in the universe

    Vogelsberger, M., Genel, S., Springel, V., Torrey, P., Sijacki, D., Xu, D., Snyder, G., Nelson, D., and Hernquist, L. Introducing the illustris project: simulating the coevolution of dark and visible matter in the universe. Monthly Notices of the Royal Astronomical Society, 44...

    2014

  64. [74]

    and Farrell, S

    Wagenmakers, E.-J. and Farrell, S. Aic model selection using akaike weights. Psychonomic Bulletin & Review, 11 0 (1): 0 192--196, 2004. doi:10.3758/BF03206482. URL https://doi.org/10.3758/BF03206482

    2004 doi

  65. [75]

    A., Stinson, G

    Wang, L., Dutton, A. A., Stinson, G. S., Macci \`o , A. V., Penzo, C., Kang, X., Keller, B. W., and Wadsley, J. Nihao project--i. reproducing the inefficiency of galaxy formation across cosmic time with a large sample of cosmological hydrodynamical simulations. Monthly Notices...

    2015

  66. [76]

    R., Johnson , L

    Weisz , D. R., Johnson , L. C., Foreman-Mackey , D., Dolphin , A. E., Beerman , L. C., Williams , B. F., Dalcanton , J. J., Rix , H.-W., Hogg , D. W., Fouesneau , M., Johnson , B. D., Bell , E. F., Boyer , M. L., Gouliermis , D., Guhathakurta , P., Kalirai , J. S., Lewis , A. ...

    2015 doi

  67. [77]

    A deep learning approach to test the small-scale galaxy morphology and its relationship with star formation activity in hydrodynamical simulations

    Zanisi, L., Huertas-Company, M., Lanusse, F., Bottrell, C., Pillepich, A., Nelson, D., Rodriguez-Gomez, V., Shankar, F., Hernquist, L., Dekel, A., et al. A deep learning approach to test the small-scale galaxy morphology and its relationship with star formation activity in hyd...

    2021

  68. [78]

    T., Oliver , W

    Zhou , L., Radev , S. T., Oliver , W. H., Obreja , A., Jin , Z., and Buck , T. Evaluating Sparse Galaxy Simulations via Out-of-Distribution Detection and Amortized Bayesian Model Comparison . arXiv e-prints, art. arXiv:2410.10606, October 2024. doi:10.48550/arXiv.2410.10606

    doi arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.