Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI

Adrian E. Bayer; Leander Thiele; Naoya Takeishi

REVIEW 3 major objections 4 minor 32 references

Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI

T0 review · 3 major / 4 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 The paper proposes a multi-fidelity training loss for simulation-based inference that combines feature matching and response distillation, claiming it outperforms weight-initialization transfer and improves posterior quality when…

desk verdict A real methodological addition to multi-fidelity SBI with a mostly sound derivation, but the headline claim of consistent improvement rests on single runs with no error bars. read the letter →

arxiv 2507.00514 v1 pith:YIL7MOBR submitted 2025-07-01 astro-ph.CO cs.LG

Leander Thiele , Adrian E. Bayer , Naoya Takeishi This is my paper · ORCID

classification astro-ph.COcs.LG

keywords simulation-basedinferencemulti-fidelitylearningneuralposteriorestimationknowledgedistillationfeaturematchingcosmologicalmatterpowerspectrumnormalizingflows

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

The paper claims that expensive high-fidelity cosmological simulations can be supplemented with many cheap low-fidelity simulations, not merely by pretraining on the cheap ones but by training on both together. Its multi-fidelity neural posterior estimation loss adds two terms to the standard NPE loss: feature matching, in which a transfer network maps low-fidelity summaries to high-fidelity summaries and those mapped summaries condition the posterior, and response distillation, in which a teacher posterior trained on the large low-fidelity set guides the student posterior trained on the small high-fidelity set. In a 5-parameter $\Lambda$CDM power-spectrum task and on Gaussian and SLCP synthetic benchmarks, the method is claimed to beat weight-initialization transfer learning, with the largest gains at the smallest high-fidelity budgets and for multimodal posteriors. If correct, this would lower the simulation cost of trustworthy cosmological inference.

What carries the argument

The load-bearing object is the stochastic transfer network $r_\psi(x \mid y)$, which maps a low-fidelity summary $y$ to a distribution over high-fidelity summaries $x$, together with a three-term training loss $L_1 + \alpha L_2 + \beta L_3$ (Eq. 5). $L_1$ is the ordinary NPE loss on high-fidelity pairs; $L_2$ is feature matching, which draws $x \sim r_\psi(\cdot \mid y)$, evaluates the NPE loss on the drawn summaries, and adds a surrogate forward-KL term aligning $r_\psi$ with the true $p(x \mid y)$; $L_3$ is response distillation, which samples parameters from the low-fidelity teacher $\tilde q_\xi(\theta \mid y)$ and trains the student $q_\phi(\theta \mid x)$ to place mass on those parameters. Replacing the intractable reverse KL with the forward-KL surrogate (Eq. 3') and then with a squared-error heuristic (Section 3.2.3) is what makes the training practical.

What would settle it

Compare the proposed method against weight initialization on a synthetic pair of simulators where the low-fidelity summary is deliberately made uninformative about the high-fidelity summary, for example by adding independent noise in a different coordinate system. If the method still matches or beats weight initialization, the gain is not coming from the feature-matching or distillation mechanism; if it degrades sharply, that confirms the mechanism relies on predictive fidelity. Alternatively, compute the target posterior KL directly on a large held-out set for the Gaussian task at $n_1 = 100$ and test whether the improvement over weight initialization exceeds run-to-run noise.

Watch

Extended reading notes

Core claim

The central claim is that a training objective combining feature matching and response distillation lets a small high-fidelity simulation set borrow information from a larger low-fidelity set, yielding better posterior approximations than training on the high-fidelity set alone or transferring via weight initialization, particularly for small simulation budgets and difficult inference problems. The method is a superset of weight initialization: it can initialize from the low-fidelity teacher and then continue using low-fidelity information throughout fine-tuning. The quantitative support comes from NLTP, C2ST, and MMD metrics on the cosmology task and from MMD and C2ST on the two synthetic benchmarks, where the proposed method consistently improves on the baselines.

Load-bearing premise

The method's extra losses help only if the low-fidelity summaries carry information about what the high-fidelity summaries would have been; if the two fidelities diverge unpredictably, the transfer network has nothing to transfer and the added terms can distort the posterior.

Editorial extensions

If this is right

With only 100 high-fidelity simulations, the proposed training produces posteriors close to a reference trained on 28k high-fidelity samples, whereas high-fidelity-only training is far off.
The method handles any number of fidelity levels and allows summaries or embeddings of different shapes across fidelities.
Because it is a superset of weight initialization, it can replace transfer-learning baselines without sacrificing their benefits.
The gains are largest for small high-fidelity budgets and for problems with complicated, multimodal posteriors such as SLCP.
Training converges faster than weight initialization on the tested tasks.

Reading between the lines

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

The same two-term recipe of feature matching plus response distillation should transfer directly to neural ratio estimation, which the paper flags as deferred; a ratio-estimation variant might be more robust for field-level summaries.
The seed-matched low- and high-fidelity simulation pairs used in the cosmology experiment are a favorable setting because the summaries are tightly coupled; applying the method to unpaired or biased fidelities, such as different resolutions or code physics, is the natural stress test.
The squared-error surrogate for the feature-matching KL is the weakest link: a stronger bound or a proper likelihood for $r_\psi$ could turn the heuristic into a variational objective with a guaranteed upper bound on the target NPE loss.
If the cosmology results hold at the 100-simulation budget, the method could make pipeline design for upcoming surveys cheaper by allowing exploratory analyses on approximate simulators before committing to the final high-fidelity campaign.

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.

Desk Editor's Note

read the letter

Short version: this is a genuine methodological contribution to multi-fidelity SBI, but the central empirical claim is not yet established because every comparison in the paper is a single run with no error bars. What is new: prior multi-fidelity SBI used calibration (Jia 2024a) or weight initialization (Krouglova et al. 2025; Saoulis et al. 2025). This paper adds a stochastic feature-matching map between low- and high-fidelity embeddings plus response distillation, and shows that weight initialization is a special case of the proposed training loss. The Jensen derivation in Section 3.2.1 is correct as far as it goes: it produces an upper bound with a KL term that motivates the feature-matching surrogate, and the distillation term follows the same pattern. The synthetic results on Gaussian and SLCP show the method beating weight initialization across the tested budgets, and the cosmology example shows moderate gains at small high-fidelity sizes. That is useful for the subfield, and the authors are honest about the scope. Soft spots, in proportion: the feature-matching loss replaces the intractable KL (3) with the reverse KL (3') and then a squared-error heuristic in Section 3.2.3. The paper flags these as heuristics, so there is no misrepresentation, but it means the method is a recipe without a theoretical guarantee that the surrogate minimizes the posterior KL. More important, Figures 2 and 4 report point estimates only. There is no statement of the number of seeds or standard deviations, and the abstract claims consistent improvement despite that. With n1 = 100 for cosmology, training stochasticity and finite-test variance are large enough that one favorable seed could drive the whole conclusion. The absence of code and hyperparameter details compounds the problem, and the seed-matched FastPM/Quijote pairs are an optimistic setting, though the authors acknowledge that matching seeds is not required in principle. Net: the method deserves serious attention, and the paper is written clearly and candidly. But the evidence as reported is suggestive, not conclusive. A serious referee should ask for repeated seeds, error bars, and full implementation details before acceptance.

Referee Report

3 major / 4 minor

Summary. The paper proposes a multi-fidelity training objective for neural posterior estimation (NPE) that combines feature matching with response distillation. The high-fidelity NPE is trained with the standard loss on a small high-fidelity dataset, augmented by a term that maps low-fidelity summaries into the high-fidelity embedding space and a distillation term from a low-fidelity teacher posterior. The method generalizes weight-initialization transfer learning and is evaluated on a cosmological power-spectrum task using Quijote/FastPM simulations and on the Gaussian and SLCP synthetic benchmarks.

Significance. If validated, the method would be practically useful for simulation-based inference with expensive simulators, since it exploits cheaper low-fidelity simulations and can accommodate summaries of different shapes. The Jensen-inequality derivation in Section 3.2.1 is a principled starting point, and the use of a real cosmological simulation pair (Quijote/FastPM) is a strength, as is the inclusion of the SLCP benchmark. However, the empirical evidence is currently too thin to support the claimed consistent improvements: the quantitative results are single point estimates without seed statistics, and the auxiliary losses involve heuristics whose effect on the target posterior is not characterized. The method is a reasonable contribution but needs stronger empirical support and clearer framing of its theoretical status.

major comments (3)

[Section 3.2.1, Eq. (3')] The Jensen upper bound above Eq. (3) contains KL(rψ(x|y) || p(x|y)), but Eq. (3') replaces this with the reversed KL, KL(p(x|y) || rψ(x|y)). Reversing the KL direction does not preserve an upper bound, and the paper provides no argument that this surrogate, together with the squared-error replacement in Section 3.2.3, minimizes the target posterior KL. Since Eq. (5) is the central training loss, this gap should either be resolved theoretically or explicitly framed as an empirical proposal with robustness checks.
[Sections 2 and 4, Figures 2 and 4] All quantitative results are single point estimates of NLTP, C2ST, and MMD; the paper does not report the number of random seeds or error bars. For the headline regime of small high-fidelity budgets (n1 = 100 for cosmology, n1 = 1k for SLCP), stochastic training and finite-test variance are substantial, so the statement that the proposed method 'consistently outperforms' weight initialization is not supported by the evidence as presented. Please provide repeated-seed results with error bars or a statistical test.
[Section 2, footnote 1] The cosmology evaluation pairs FastPM and Quijote simulations with matched parameters and seeds, and the footnote states that performance benefits from this matching. This is an optimistic setting for the feature-matching component, because with unmatched simulations rψ must absorb additional stochasticity. The transferability of the claimed improvement to realistic independent low- and high-fidelity runs is therefore not demonstrated; an experiment with unmatched seed pairs, or a careful discussion of why the seed matching is not responsible for the observed gains, is needed.

minor comments (4)

[Section 2] The test-loss metric NLTP is never defined; please specify its formula.
[Section 3.2.3] The squared-error replacement for -log rψ is described in a single sentence; please clarify how samples v ∼ rψ(·|Sσ(y)) are generated and whether the reparameterization trick is used for gradient flow.
[Section 2] Hyperparameters are said to be optimized with optuna, but the chosen hyperparameters are not reported; providing them, or a code release, would be necessary for reproducibility.
[Section 5] The introduction claims the method can accommodate any number of fidelity levels, but this extension is not demonstrated or detailed; please spell out the recursion or cite a specific construction.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the multi-fidelity objective is derived from a Jensen upper bound and tested on held-out high-fidelity data, with no fitted input renamed as a prediction.

full rationale

The derivation chain in Section 3 is self-contained. The feature-matching term starts from the y-based posterior KL, applies Jensen's inequality to obtain the bound with rpsi, and then explicitly replaces the intractable KL (3) with the reverse-direction KL (3') labeled 'As a surrogate of (3)...'. The final implementation further replaces the log-density of rpsi with a squared-error term and explicitly flags it as a heuristic ('We empirically found... more stable learning. We note that further analysis of this heuristic may lead to an interesting discussion.'). These are acknowledged approximations, not definitions that equate the target with the loss. The response-distillation term (4) uses a teacher q_tilde_xi trained independently on D2 with the standard NPE loss; the teacher is fixed before student training, so the student is not being graded by its own output. The claim that the method is a superset of weight initialization is structural (taking alpha=beta=0 and initializing phi with xi reduces to finetuning), not a prediction. All quantitative claims in Sections 2 and 4 are empirical comparisons against reference posteriors built from 28k (cosmology) or 100k (SLCP) high-fidelity samples, using held-out test sets, and the baselines (high-fidelity-only and weight-initialization) are trained on the same data splits. No parameter is fitted to the evaluation metric and then reported as a prediction. The seed-matched FastPM/Quijote pairs are a design choice explicitly stated in a footnote, with the remark that seed matching is not required for the method to function; this affects the difficulty of the benchmark, not the circularity of the derivation. Self-citations (Thiele et al. 2024; Novaes et al. 2025) appear only as examples of SBI applications in the introduction and do not carry the argument. The absence of repeated-seed error bars is a legitimate statistical-robustness concern, but it is not circularity under the criteria used here.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities. Its assumptions are about the informativeness of low-fidelity simulations and the adequacy of surrogate losses.

free parameters (2)

α (feature matching loss weight) = not reported
Weight on L2 in Eq. (5); chosen by hyperparameter optimization but value not given.
β (distillation loss weight) = not reported
Weight on L3 in Eq. (5); not specified.

assumptions (5)

standard math Jensen's inequality provides a valid upper bound on -log qφ(θ|y).
Used in Section 3.2.1 to derive the feature-matching loss.
standard math Monte Carlo samples approximate the expectations in the loss.
Used throughout Section 3.2; standard in SBI.
domain assumption The stochastic map rψ(x|y) can approximate p(x|y) accurately enough.
Required for the feature-matching term to transfer information; no guarantee given.
domain assumption Low-fidelity summaries y are informative about high-fidelity summaries x.
The method's effectiveness relies on this; the cosmology test uses seed-matched FastPM/Quijote pairs (footnote 1).
domain assumption Teacher posterior q̃ξ(θ|y) is a good estimate of p(θ|y).
Response distillation in Section 3.2.2 assumes the teacher trained on D2 is reliable.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI." pith.science (2026). https://pith.science/paper/YIL7MOBR

@misc{pith2026250700514,
  author       = {Pith},
  title        = {Pith review of: Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YIL7MOBR}},
  note         = {Machine review of arXiv:2507.00514}
}

read the original abstract

The simulation cost for cosmological simulation-based inference can be decreased by combining simulation sets of varying fidelity. We propose an approach to such multi-fidelity inference based on feature matching and knowledge distillation. Our method results in improved posterior quality, particularly for small simulation budgets and difficult inference problems.

Figures

Figures reproduced from arXiv: 2507.00514 by the authors.

**Figure 1.** Graphical summary of the proposed multi-fidelity architecture and training loss. Filled circles indicate the components of our proposed training loss. The notation is given in Section 3. Hatched components are weight-initialized (dotted arrows) from pretraining on D2, the cross-hatched ones being frozen. The final model only evaluates the path indicated in orange. The right side shows the weight-initialization sche… view at source ↗

**Figure 2.** Performance evaluation according to various metrics for the cosmological inference problem. Lower is better for all. and transfer learning via weight initialization (Krouglova et al., 2025; Saoulis et al., 2025). These existing approaches highlight the potential of multi-fidelity SBI to reduce the required simulation budget. Our contribution employs a tailored training setup which enables training on multiple fideli… view at source ↗

**Figure 3.** Posteriors on an example realization from the test set. We compare models trained on 100 high-fidelity simulations to the reference posterior (grey) trained on 28k high-fidelity simulations. We take MLPs for the embedding and transfer networks, and a spline flow (Durkan et al., 2019) from sbi (TejeroCantero et al., 2020; Durkan et al., 2020) for the density estimator. Hyperparameters are optimized with optuna (Ak… view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: (Left) MMD between the true and learned posteriors for GAUSSIAN. (Right) C2ST metrics for SLCP. The reference performance is by NPE on a large high-fid. set with n1 = 100k. plus Gaussian noise. y is supposed to have “low-fidelity” by setting dim y < dim x and being add…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 22 canonical work pages

[1]

Optuna: A next-generation hyperparameter optimization framework

Akiba, T., Sano, S., Yanase, T., Ohta, T., and Koyama, M. Optuna: A next-generation hyperparameter optimization framework. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp.\ 2623–2631, 2019

work page 2019
[2]

How many simulations do we need for simulation-based inference in cosmology?, 2025

Bairagi, A., Wandelt, B., and Villaescusa-Navarro, F. How many simulations do we need for simulation-based inference in cosmology?, 2025. arXiv:2503.13755

arXiv 2025
[3]

K., Witte, S

Cole, A., Miller, B. K., Witte, S. J., Cai, M. X., Grootes, M. W., Nattino, F., and Weniger, C. Fast and credible likelihood-free cosmology with truncated marginal neural ratio estimation. Journal of Cosmology and Astroparticle Physics, 2022 0 (09): 0 004, 2022

work page 2022
[4]

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

work page 2020
[5]

Neural spline flows

Durkan, C., Bekasov, A., Murray, I., and Papamakarios, G. Neural spline flows. In Advances in Neural Information Processing Systems, volume 32, pp.\ 7511--7522, 2019

work page 2019
[6]

nflows : normalizing flows in PyTorch , 2020

Durkan, C., Bekasov, A., Murray, I., and Papamakarios, G. nflows : normalizing flows in PyTorch , 2020. doi: 10.5281/zenodo.4296287

doi:10.5281/zenodo.4296287 2020
[7]

Fastpm: a new scheme for fast simulations of dark matter and haloes

Feng, Y., Chu, M.-Y., Seljak, U., and McDonald, P. Fastpm: a new scheme for fast simulations of dark matter and haloes. Monthly Notices of the Royal Astronomical Society, 463 0 (3): 0 2273--2286, 2016

work page 2016
[8]

T., Dodelson , S., Doux , C., Eckert , K., Elvin-Poole , J., Everett , S., Ferte , A., Gruen , D., Gruendl , R., Harrison , I., Hartley , W

Gatti , M., Jeffrey , N., Whiteway , L., Williamson , J., Jain , B., Ajani , V., Anbajagane , D., Giannini , G., Zhou , C., Porredon , A., Prat , J., Yamamoto , M., Blazek , J., Kacprzak , T., Samuroff , S., Alarcon , A., Amon , A., Bechtol , K., Becker , M., Bernstein , G., Campos , A., Chang , C., Chen , R., Choi , A., Davis , C., Derose , J., Diehl , H...

work page 2024

Show all 32 references

[9]

Automatic posterior transformation for likelihood-free inference

Greenberg, D., Nonnenmacher, M., and Macke, J. Automatic posterior transformation for likelihood-free inference. In Proceedings of the 36th International Conference on Machine Learning, pp.\ 2404--2414, 2019

2019
[10]

M., Rasch, M

Gretton, A., Borgwardt, K. M., Rasch, M. J., Sch \"o lkopf, B., and Smola, A. A kernel two-sample test. Journal of Machine Learning Research, 13 0 (25): 0 723--773, 2012

2012
[11]

Cosmological constraints from non-Gaussian and nonlinear galaxy clustering using the SIMBIG inference framework

Hahn , C., Lemos , P., Parker , L., R \'e galdo-Saint Blancard , B., Eickenberg , M., Ho , S., Hou , J., Massara , E., Modi , C., Moradinezhad Dizgah , A., and Spergel , D. Cosmological constraints from non-Gaussian and nonlinear galaxy clustering using the SIMBIG inference fr...

2024
[12]

Likelihood-free MCMC with amortized approximate ratio estimators

Hermans, J., Begy, V., and Louppe, G. Likelihood-free MCMC with amortized approximate ratio estimators. In Proceedings of the 37th International Conference on Machine Learning, pp.\ 4239--4248, 2020

2020
[13]

A trust crisis in simulation-based inference? Beware , your posterior approximations can be unfaithful

Hermans, J., Delaunoy, A., Rozet, F., Wehenkel, A., Begy, V., and Louppe, G. A trust crisis in simulation-based inference? Beware , your posterior approximations can be unfaithful. Transactions on Machine Learning Research, 2022

2022
[14]

Distilling the knowledge in a neural network, 2015

Hinton, G., Vinyals, O., and Dean, J. Distilling the knowledge in a neural network, 2015. arXiv:1503.02531

2015 arXiv
[15]

Simulation-based inference has its own Dodelson-Schneider effect (but it knows that it does), 2024

Homer, J., Friedrich, O., and Gruen, D. Simulation-based inference has its own Dodelson-Schneider effect (but it knows that it does), 2024. arXiv:2412.02311

2024 arXiv
[16]

Cosmological analysis with calibrated neural quantile estimation and approximate simulators, 2024 a

Jia, H. Cosmological analysis with calibrated neural quantile estimation and approximate simulators, 2024 a . arXiv:2411.14748

2024 arXiv
[17]

Simulation-based inference with quantile regression

Jia, H. Simulation-based inference with quantile regression. In Proceedings of the 41st International Conference on Machine Learning, pp.\ 21731--21752, 2024 b

2024
[18]

N., Johnson, H

Krouglova, A. N., Johnson, H. R., Confavreux, B., Deistler, M., and Gonçalves, P. J. Multifidelity simulation-based inference for computationally expensive simulators, 2025. arXiv:2502.08416

2025
[19]

M., Blancard, B

Lemos, P., Parker, L., Hahn, C., Ho, S., Eickenberg, M., Hou, J., Massara, E., Modi, C., Dizgah, A. M., Blancard, B. R.-S., and Spergel, D. Field-level simulation-based inference of galaxy clustering with convolutional neural networks. Phys. Rev. D, 109: 0 083536, 2024

2024
[20]

and Oquab, M

Lopez-Paz, D. and Oquab, M. Revisiting classifier two-sample tests, 2018. arXiv:1610.06545

2018 arXiv
[21]

J., Bassetto, G., \" O cal, K., Nonnenmacher, M., and Macke, J

Lueckmann, J.-M., Goncalves, P. J., Bassetto, G., \" O cal, K., Nonnenmacher, M., and Macke, J. H. Flexible statistical inference for mechanistic models of neural dynamics. In Advances in Neural Information Processing Systems, volume 30, pp.\ 1289--1299, 2017

2017
[22]

Lueckmann, J.-M., Bassetto, G., Karaletsos, T., and Macke, J. H. Likelihood-free inference with emulator networks. In Proceedings of the 1st Symposium on Advances in Approximate Bayesian Inference, pp.\ 32--53, 2019

2019
[23]

Benchmarking simulation-based inference

Lueckmann, J.-M., Boelts, J., Greenberg, D., Goncalves, P., and Macke, J. Benchmarking simulation-based inference. In Proceedings of the 24th International Conference on Artificial Intelligence and Statistics, pp.\ 343--351, 2021

2021
[24]

and Philcox, O

Modi, C. and Philcox, O. H. E. Hybrid SBI or how I learned to stop worrying and learn the likelihood, 2023. arXiv:2309.10270

2023 arXiv
[25]

P., Thiele , L., Armijo , J., Cheng , S., Cowell , J

Novaes , C. P., Thiele , L., Armijo , J., Cheng , S., Cowell , J. A., Marques , G. A., Ferreira , E. G. M., Shirasaki , M., Osato , K., and Liu , J. Cosmology from HSC Y1 weak lensing data with combined higher-order statistics and simulation-based inference . , 111 0 (8): 0 08...

2025
[26]

and Murray, I

Papamakarios, G. and Murray, I. Fast -free inference of simulation models with Bayesian conditional density estimation. In Advances in Neural Information Processing Systems, volume 29, pp.\ 1036--1044, 2016

2016
[27]

Sequential neural likelihood: Fast likelihood-free inference with autoregressive flows

Papamakarios, G., Sterratt, D., and Murray, I. Sequential neural likelihood: Fast likelihood-free inference with autoregressive flows. In Proceedings of the 22nd International Conference on Artificial Intelligence and Statistics, pp.\ 837--848, 2019

2019
[28]

A., Piras, D., Jeffrey, N., Mancini, A

Saoulis, A. A., Piras, D., Jeffrey, N., Mancini, A. S., Ferreira, A. M. G., and Joachimi, B. Transfer learning for multifidelity simulation-based inference in cosmology, 2025. arXiv:2505.21215

2025
[29]

J., Greenberg, D

Tejero-Cantero, A., Boelts, J., Deistler, M., Lueckmann, J.-M., Durkan, C., Gonçalves, P. J., Greenberg, D. S., and Macke, J. H. sbi: A toolkit for simulation-based inference. Journal of Open Source Software, 5 0 (52): 0 2505, 2020

2020
[30]

N., Ho , S., and Wandelt , B

Thiele , L., Massara , E., Pisani , A., Hahn , C., Spergel , D. N., Ho , S., and Wandelt , B. Neutrino Mass Constraint from an Implicit Likelihood Analysis of BOSS Voids . , 969 0 (2): 0 89, 2024

2024
[31]

M., Ramanah, D

Villaescusa-Navarro, F., Hahn, C., Massara, E., Banerjee, A., Delgado, A. M., Ramanah, D. K., Charnock, T., Giusarma, E., Li, Y., Allys, E., Brochard, A., Uhlemann, C., Chiang, C.-T., He, S., Pisani, A., Obuljen, A., Feng, Y., Castorina, E., Contardo, G., Kreisch, C. D., Nicol...

2020
[32]

Zhang , G., Modi , C., and Philcox , O. H. E. Modeling galaxy surveys with hybrid SBI , 2025. arXiv:2505.13591

2025 arXiv

Pith tools

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

[1] [1]

Optuna: A next-generation hyperparameter optimization framework

Akiba, T., Sano, S., Yanase, T., Ohta, T., and Koyama, M. Optuna: A next-generation hyperparameter optimization framework. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp.\ 2623–2631, 2019

work page 2019

[2] [2]

How many simulations do we need for simulation-based inference in cosmology?, 2025

Bairagi, A., Wandelt, B., and Villaescusa-Navarro, F. How many simulations do we need for simulation-based inference in cosmology?, 2025. arXiv:2503.13755

arXiv 2025

[3] [3]

K., Witte, S

Cole, A., Miller, B. K., Witte, S. J., Cai, M. X., Grootes, M. W., Nattino, F., and Weniger, C. Fast and credible likelihood-free cosmology with truncated marginal neural ratio estimation. Journal of Cosmology and Astroparticle Physics, 2022 0 (09): 0 004, 2022

work page 2022

[4] [4]

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

work page 2020

[5] [5]

Neural spline flows

Durkan, C., Bekasov, A., Murray, I., and Papamakarios, G. Neural spline flows. In Advances in Neural Information Processing Systems, volume 32, pp.\ 7511--7522, 2019

work page 2019

[6] [6]

nflows : normalizing flows in PyTorch , 2020

Durkan, C., Bekasov, A., Murray, I., and Papamakarios, G. nflows : normalizing flows in PyTorch , 2020. doi: 10.5281/zenodo.4296287

doi:10.5281/zenodo.4296287 2020

[7] [7]

Fastpm: a new scheme for fast simulations of dark matter and haloes

Feng, Y., Chu, M.-Y., Seljak, U., and McDonald, P. Fastpm: a new scheme for fast simulations of dark matter and haloes. Monthly Notices of the Royal Astronomical Society, 463 0 (3): 0 2273--2286, 2016

work page 2016

[8] [8]

T., Dodelson , S., Doux , C., Eckert , K., Elvin-Poole , J., Everett , S., Ferte , A., Gruen , D., Gruendl , R., Harrison , I., Hartley , W

Gatti , M., Jeffrey , N., Whiteway , L., Williamson , J., Jain , B., Ajani , V., Anbajagane , D., Giannini , G., Zhou , C., Porredon , A., Prat , J., Yamamoto , M., Blazek , J., Kacprzak , T., Samuroff , S., Alarcon , A., Amon , A., Bechtol , K., Becker , M., Bernstein , G., Campos , A., Chang , C., Chen , R., Choi , A., Davis , C., Derose , J., Diehl , H...

work page 2024