Leveraging generative models to assist Monte Carlo sampling

Marylou Gabri\'e

REVIEW 2 major objections 6 minor 201 references

Leveraging generative models to assist Monte Carlo sampling

T0 review · 2 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A generative model trained to approximate a target distribution can be turned into an exact sampler by embedding it in importance sampling, Metropolis-Hastings, or transport-based schemes.

desk verdict Solid tutorial review of generative-model-assisted sampling; needs a sign fix in Eq. (29) and a small caveat about the author's own benchmarks, but deserves a real referee. read the letter →

arxiv 2608.07648 v1 pith:DCH3WKND submitted 2026-08-07 stat.ML cond-mat.dis-nncond-mat.stat-mechcs.LGphysics.comp-ph

Marylou Gabrié This is my paper · ORCID

classification stat.MLcond-mat.dis-nncond-mat.stat-mechcs.LGphysics.comp-ph MSC 65C0562F1568T07

keywords generativemodelsMonteCarlosamplingnormalizingflowsdiffusionmodecollapseannealedimportancemetastability

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

This tutorial review argues that generative models — normalizing flows, autoregressive models, and diffusion models — can serve as engines for Monte Carlo sampling of distributions known only up to a normalization constant. The key statement is that the resulting samplers remain exact even when the learned model only approximates the target: embedding the model in importance sampling, independent Metropolis-Hastings, or annealed reweighting corrects the approximation error in the limit of many samples. The review also surveys how to train such models without any data, using reverse-KL variational inference, annealing, and stochastic optimal control, and maps the known failure modes: mode collapse, the curse of dimensionality, and the cost of likelihood evaluation. A sympathetic reader would care because this is a coherent toolbox that may complement tempering and collective-variable methods for metastable, high-dimensional targets.

What carries the argument

The central objects are exact-likelihood generative models — normalizing flows, whose change-of-variables formula $\rho_\theta(x)=\rho_0(T_\theta^{-1}(x))|\det\nabla T_\theta^{-1}(x)|$ makes proposal densities tractable — and autoregressive models with factorized conditionals. These densities feed three carrying mechanisms: importance weights $w=\pi/\rho_\theta$, Metropolis-Hastings acceptance ratios, and transport maps that pull the target back to a Gaussian latent space. For continuous-time models, the carrying mechanism is the stochastic-flow reweighting identity: joint densities over a discretized noising-denoising path are ratios of Gaussian kernels, so no divergence evaluation is needed. The reverse-KL divergence $D_{\mathrm{KL}}(\rho_\theta\,\|\,\pi)=\mathbb{E}_{\rho_\theta}[\log\rho_\theta+U]+\mathrm{const}$ is the training objective that makes data-free learning possible.

What would settle it

Take a single multimodal target with known normalization, for instance a 64-dimensional Gaussian mixture with known mode weights, and run each reviewed exact sampler on the same computational budget, measuring effective sample size and mode coverage; if no flow- or diffusion-based sampler matches or exceeds a well-tuned adaptive sequential Monte Carlo with tempered transitions, the review's claim that these mechanisms offer a complementary route would be undermined.

Watch

Extended reading notes

Core claim

The paper's central claim is that a generative model approximating a target distribution $\pi$ can be converted into an exact sampler through three families of constructions: reweighting (importance sampling and free-energy estimators), independent-proposal Markov chains (Metropolis-Hastings and iterated sampling-importance-resampling), and transport-map reparametrization (neutra-MCMC). In every case the Monte Carlo correction — weights, acceptance ratios, or super-detailed-balance steps — ensures asymptotic unbiasedness regardless of how imperfect the model is. For continuous-time models, the review identifies a second mechanism: rewriting the diffusion or interpolant generative process as a discrete stochastic flow and debiasing it with annealed importance weights computed from the forward and backward Gaussian kernels, which avoids divergence computations. The review further claims that training without data is feasible via reverse-KL variational inference, and that mode collapse can be mitigated by annealing schedules, learned reference distributions, or maximum-likelihood retraining on adaptively collected samples. On impact it is cautious: these samplers have not yet produced scientific results unattainable by existing methods, but they offer a qualitatively different exploration mechanism — jumping across low-probability regions.

Load-bearing premise

The tutorial's comparative guidance rests on the assumption that the specific studies it uses to compare methods — particularly the flow benchmark of [Gre+23], the mode-collapse analyses of [GNG25] and [FLG26], and the diffusion comparison of [GN26] — are representative of the broader literature, even though the paper itself concedes that systematic comparisons are lacking.

Editorial extensions

If this is right

Because the samplers remain exact for imperfect models, deployment in high-stakes scientific computing is not blocked by model error; the practical limits become sample size, variance, and compute.
The exponential decay of effective sample size with dimension for independent-proposal methods implies that practical gains will come from hybrid designs that combine flows with tempering, sequential Monte Carlo, or coarse-graining.
Diffusion-based annealing paths may offer a new way to bridge simple and target distributions that avoids phase transitions that break traditional tempering schedules.
Free energy differences can be estimated with the reviewed methods, from targeted free energy perturbation and Bennett acceptance ratio to the adaptive-transport estimator, all without a training dataset.
The taxonomy gives practitioners a structured map for choosing among exact samplers based on whether the dominant difficulty is ill-conditioning, multimodality, or high dimensionality.

Reading between the lines

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

If exactness is truly robust to model quality, the field's core open problem is not bias but variance: the effective sample size and acceptance rate, not the asymptotic guarantee, will decide which method is viable for a given target.
The 'duality of concavity' requirement for score-estimation samplers suggests a testable extension: replacing the hand-picked initialization time $t_0$ with a learned Gaussian-mixture reference may broaden the class of multimodal targets these samplers can handle.
The review itself notes that systematic benchmarks are lacking; a standardized comparison of all reviewed exact samplers on the same high-dimensional multimodal targets would likely reveal which of the reviewed mechanisms scales best, and would decide how much of the field's promise is real.

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

Best to send this tutorial review to a competent referee. It's a useful map of a young field, and the author is unusually candid about its limits: mode collapse, dimension scaling, absent benchmarks, and the fact that these samplers haven't produced unattainable science yet. The distinction between generative modeling and sampling is clear, and the sections on exact samplers, training without data, and continuous-time models are well organized. The math is standard and mostly consistent.

The paper does not claim new results, so novelty is correctly low. That is not a problem for a review.

The soft spots are real but manageable. First, Eq. (29) has a sign error. The stress-test is right: as printed, ESS ≈ N exp(−d log(E[w]^2/E[w]^2)), and since E[w]^2 ≤ E[w]^2 by Jensen, the log is non-positive, so the formula predicts ESS growing with dimension. The intended expression is ESS ≈ N exp(−d log(E[w]^2/E[w]^2)) with the ratio inverted; then the exponential decay the text claims follows. This is a one-character fix but it sits in a key pedagogical calculation, so it should be corrected before publication.

Second, the selection of comparisons draws heavily on the author's own papers — [Gre+23], [GNG25], [FLG26], [GN26] — and the review itself acknowledges that no systematic benchmark exists. That's a fair concern to raise, but not a reason to reject: the review is explicitly a selection, and the author flags the absence of independent comparisons. Still, it would be helpful to mark which claims rely on the author's own studies, or to include at least one outside benchmark.

The citation pattern is otherwise broad and fair; the review points to prior surveys and original works.

Who benefits: newcomers to generative-model-assisted sampling, and practitioners wanting a structured overview of methods and open problems. It's not a research paper, but it does a real service. A serious referee can add value by checking the formulas and benchmark claims.

Recommendation: send to peer review. With the sign fix and a short note on benchmark provenance, this would be a solid published tutorial.

Referee Report

2 major / 6 minor

Summary. This tutorial review surveys the use of generative models—autoregressive models, normalizing flows, continuous normalizing flows, diffusion models, and stochastic interpolants—as proposals or transport maps in Monte Carlo sampling of target densities known only up to a normalization constant. It is organized around three stages: exact samplers built from exact-likelihood models (Section 3), training strategies that do not require data from the target (Section 4), and verifiable sampling with continuous-time generative models (Section 5). The paper's central claims are that imperfectly trained generative models can nonetheless be embedded in exact sampling schemes, and that the main open problems are computational cost, dimension scaling, mode collapse, and the absence of systematic benchmarks.

Significance. As a tutorial, the manuscript is well structured and fills a genuine gap: it provides unified notation, algorithm boxes, and candid discussion of limitations, including mode collapse, missing benchmarks, and the fact that generative-model-assisted samplers have not yet produced scientific discoveries unattainable with classical methods. The author repeatedly acknowledges where systematic comparison is lacking (Sections 3.3 and 5), which strengthens the review's credibility. However, the review's comparative guidance is only as reliable as its small evidence base, and Section 3.2 contains an internally inconsistent formula that underpins a central limitation claim; the tutorial therefore needs correction before it can be recommended as a dependable entry point for newcomers.

major comments (2)

[3.2, Eq. (29)] The effective sample size is displayed as ESS≈N exp(−d log(E[w]^2/E[w^2])). Because E[w]^2<E[w^2] for any imperfect proposal, the logarithm is negative, so the displayed formula predicts that ESS grows exponentially with dimension. This directly contradicts the sentence that follows ('the ESS decays exponentially fast with the dimension') and undermines the intended conclusion that independent-proposal methods are cursed by dimensionality. The formula should read ESS≈N exp(−d log(E[w^2]/E[w]^2))=N(E[w]^2/E[w^2])^d. Since this calculation is the review's stated analytical justification for a central limitation of neural-IS and flow-IMH, it must be corrected.
[3.2, 3.3, 5] The comparative assertions in Section 3.2—for example that neutra-MCMCs are the most robust to ill-conditioning and that independent-proposal methods are the most affected by high dimensionality—are based on a single benchmark study ([Gre+23]) together with [DRW21], while the paper itself notes in Sections 3.3 and 5 that systematic comparison is still lacking. The tutorial should state this evidence-base limitation at the point of the claims, not only in later intermediate summaries, so that readers do not mistake provisional findings for established field-wide conclusions.

minor comments (6)

[Various] Typos: Section 2.1 'autoregressivee' should be 'autoregressive'; Section 3.1.1 'protein-lingand' should be 'protein–ligand'; Section 5.1 'succesfully' should be 'successfully'; Section 6 'armorphous' should be 'amorphous'; Figure 4 caption 'pannel' should be 'panel'.
[4.3.2, Algorithm 15] The loop 'for τ=0...T do' references \tilde{x}^{(i)}_{τ−1} at τ=0; the loop should start at τ=1 or \tilde{x}^{(i)}_0 should be initialized explicitly before the loop.
[3.2] The statement that the highest dimension admissible for independent proposals lies 'somewhere between a hundred to a few thousands' is presented as an order-of-magnitude heuristic without a citation; it should be labeled as a rough extrapolation from the cited experiments.
[5, final paragraph] The claim that the time discretization must become finer as the dimension increases is stated without derivation or reference; please add support or soften the claim to an open concern.
[References] The citation key [Zha+25] appears to be used for two different works (Section 3.3.4 on parallel tempering and Section 5.2 on annealed importance sampling with a variance-exploding diffusion); please disambiguate the references.
[Figure 5] The caption contains garbled notation ("YÆ t /Æ (t)", "qÆ t (·|yÆ t )") that should be typeset properly.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the review's equations are self-contained, and its self-citations are transparent, externally falsifiable, and explicitly disclaimed as incomplete benchmarks.

full rationale

This tutorial review contains no derivation that reduces to its own input. The samplers claimed to remain exact despite an imperfect learned model (Neural-IS, flow-IMH, flow-iSIR, neutra-MCMC, CV-FlowMC, and the diffusion stochastic-flow debiasers of Section 5.2) are exact by standard importance-sampling and Metropolis-Hastings theory, with acceptance probabilities and weight formulas written out in the text (Algorithms 1-4 and 9; Eqs. 23-25, 37, 67, 75, and 77), so the central claim in Section 3 ('even if the agreement between ρθ and π is imperfect ... the samplers described below remain exact') does not depend on any citation, including the author's own prior work. The review does lean on author-co-authored studies for its comparative judgments: Section 3.2 rediscusses the [Gre+23] benchmark to conclude that neutra-MCMCs are most robust to ill-conditioning and that independent-proposal methods are most affected by high dimensionality, and Section 4.1.1 draws its mode-collapse analysis from [GNG25] and [FLG26]. These self-citations are transparent attributions of externally reproducible numerical experiments and of a theoretical study with stated assumptions, and the high-dimensionality conclusion is additionally corroborated by an independent study ([DRW21]) cited in the same passage. The paper also explicitly disclaims the representative-sample premise on which its comparisons rest: 'a systematic comparison of these algorithms is still lacking' (Section 3.3) and 'a complete benchmark of the different approaches is still lacking' (Section 5), which precludes treating the self-cited benchmarks as forced or definitional. A separate correctness defect, not a circularity, is the ESS formula in Eq. (29): as printed, ESS ≈ N e^{−d log(E[w1d]^2/E[w1d^2])} has a positive exponent because log(E[w]^2/E[w^2]) is negative, so the displayed formula implies ESS growing with dimension and contradicts the surrounding conclusion that 'the ESS decays exponentially fast with the dimension'; the intended decaying form is e^{−d log(E[w^2]/E[w]^2)}. This internal inconsistency should be flagged for a correctness pass, but it does not make any prediction circular. Accordingly the verdict is the low end of the non-circular band: self-citation is present but not load-bearing in the forbidden sense.

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

This is a review paper, so the ledger contains no free parameters fitted to data and no newly invented entities. The mathematical content is inherited from the cited literature, and the review explicitly states its main domain assumption: the target density is known pointwise up to a constant and is hard to sample but inexpensive to evaluate. The other axioms are standard background results in optimal transport, diffusion theory, and stochastic analysis that the review invokes without proof, which is appropriate for a tutorial.

assumptions (5)

domain assumption The target distribution π admits a continuous density known pointwise up to a normalization constant, and its unnormalized density is cheap to evaluate.
Stated in Section 1.1 and used by every algorithm in the review; for example, Algorithm 1 computes weights from π(x)/ρθ(x).
standard math Brenier's theorem: for absolutely continuous measures there exists an optimal transport map pushing the base measure to the target.
Invoked in Section 2.2.1 to justify normalizing flows as expressive transport models.
standard math Anderson's reverse-time diffusion theorem gives the denoising SDE with marginals equal to the time-reversal of the forward noising process.
Used in Section 2.3, eq. (13), to define diffusion-model sampling, and relied on again in Section 4.3.
standard math Girsanov theorem and change-of-measure for SDEs make the path-measure KL objective in Section 4.3.3 well defined.
The review sketches eq. (64) as a combination of change-of-measure and Girsanov; no proof is given in the body.
domain assumption Sufficient regularity of velocity fields, potentials, and interpolation paths is assumed so that continuity, Fokker-Planck, and instantaneous change-of-variables equations hold.
Used in Sections 2.2.2, 2.4, and 4.2 when deriving ODE and SDE transport equations; the regularity conditions are not stated in the review.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Leveraging generative models to assist Monte Carlo sampling." pith.science (2026). https://pith.science/paper/DCH3WKND

@misc{pith2026260807648,
  author       = {Pith},
  title        = {Pith review of: Leveraging generative models to assist Monte Carlo sampling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DCH3WKND}},
  note         = {Machine review of arXiv:2608.07648}
}

read the original abstract

Sampling high-dimensional probability distributions is a central task in scientific computing, with applications ranging from Bayesian inference to statistical physics and molecular simulation. Despite decades of methodological developments, two major challenges remain: scaling to high dimensions and efficiently exploring multimodal distributions characterized by metastable states. Classical approaches such as Markov chain Monte Carlo, tempering methods, or enhanced sampling based on collective variables have achieved major successes, but they also face intrinsic limitations. This tutorial review explores a new paradigm that has recently emerged at the interface of machine learning and computational statistical physics: the use of generative models as tools for sampling. In this context, models such as normalizing flows and diffusion models are not used in their traditional data-driven setting, but rather as flexible probabilistic models that can assist the sampling of distributions known only up to a normalization constant. This manuscript reviews the early development of this rapidly evolving field and discusses several methodological directions, including exact samplers based on generative models and strategies to train such models in the absence of data. While an exhaustive survey of the literature is not attempted, we present a selection of key ideas and methods, along with a discussion of their strengths and limitations. The review is intended to be an accessible tutorial for both physics and machine learning audiences, and it aims to provide a starting point for researchers interested in exploring this exciting area of research.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

201 extracted references · 39 canonical work pages

[1]

Communications in Mathematical Sciences , volume =

Density Estimation by Dual Ascent of the Log-Likelihood , author =. Communications in Mathematical Sciences , volume =
[2]

Normalizing

Papamakarios, George and Nalisnick, Eric and Rezende, Danilo Jimenez and Mohamed, Shakir and Lakshminarayanan, Balaji , year = 2021, journal =. Normalizing

2021
[3]

and Brubaker, Marcus A

Kobyzev, Ivan and Prince, Simon J.D. and Brubaker, Marcus A. , year = 2021, month = nov, journal =. Normalizing

2021
[4]

Density Estimation Using

Dinh, Laurent and. Density Estimation Using. International
[5]

Durkan, Conor and Bekasov, Artur and Murray, Iain and Papamakarios, George , year = 2019, volume =. Neural. Advances in

2019
[6]

Communications on Pure and Applied Mathematics , volume =

Polar Factorization and Monotone Rearrangement of Vector-Valued Functions , author =. Communications on Pure and Applied Mathematics , volume =. doi:10.1002/cpa.3160440402 , copyright =

doi:10.1002/cpa.3160440402
[7]

Chen, Ricky T. Q. and Rubanova, Yulia and Bettencourt, Jesse and Duvenaud, David K , year = 2018, volume =. Neural. Advances in

2018
[8]

Equivariant

K. Equivariant. Proceedings of the 37th

Show all 201 references

[9]

Satorras, Victor and Hoogeboom, Emiel and Fuchs, Fabian and Posner, Ingmar and Welling, Max , year = 2021, volume =. E(n). Advances in

2021
[10]

Advances in Neural Information Processing Systems , volume =

Midgley, Laurence and Stimper, Vincent and Antor. Advances in Neural Information Processing Systems , volume =
[11]

Geiger, Mario and Smidt, Tess , year = 2022, month = jul, number =. E3nn:. doi:10.48550/arXiv.2207.09453 , archiveprefix =. 2207.09453 , primaryclass =

2022 doi arXiv
[12]

Satorras, V. E(n). Proceedings of the 38th
[13]

Proceedings of the 32nd

Deep. Proceedings of the 32nd
[14]

Song, Yang and. Score-. International
[15]

Denoising

Ho, Jonathan and Jain, Ajay and Abbeel, Pieter , year = 2020, month = dec, eprint =. Denoising. Neural

2020
[16]

Stochastic Processes and their Applications , volume =

Reverse-Time Diffusion Equation Models , author =. Stochastic Processes and their Applications , volume =
[17]

Building

Albergo, Michael Samuel and. Building. The
[18]

Lipman, Yaron and Chen, Ricky T. Q. and. Flow
[19]

and Boffi, Nicholas M

Albergo, Michael S. and Boffi, Nicholas M. and. Stochastic. arXiv , keywords =:2303.08797 , primaryclass =

arXiv
[20]

M. Neural. ACM Transactions on Graphics , volume =
[21]

Boltzmann Generators:

No. Boltzmann Generators:. Science , volume =
[22]

Albergo, M. S. and Kanwar, G. and Shanahan, P. E. , year = 2019, month = aug, journal =. Flow-Based Generative Models for

2019
[23]

Physical Review E , volume =

Asymptotically Unbiased Estimation of Physical Observables with Neural Samplers , author =. Physical Review E , volume =. doi:10.1103/PhysRevE.101.023304 , archiveprefix =. 1910.13496 , primaryclass =

1910 arXiv
[24]

and Anders, Christopher J

Nicoli, Kim A. and Anders, Christopher J. and Funcke, Lena and Hartung, Tobias and Jansen, Karl and Kessel, Pan and Nakajima, Shinichi and Stornati, Paolo , year = 2021, month = jan, journal =. Estimation of

2021
[25]

Physical Review E , volume =

Targeted Free Energy Perturbation , author =. Physical Review E , volume =
[26]

Efficient Estimation of Free Energy Differences from

Bennett, Charles H , year = 1976, journal =. Efficient Estimation of Free Energy Differences from

1976
[27]

Using bijective maps to improve free-energy estimates , author =. Phys. Rev. E , volume =. 2009 , month =. doi:10.1103/PhysRevE.79.011113 , url =

2009 doi
[28]

Normalizing

Jia, He and Seljak, Uros , year = 2020, month = feb, pages =. Normalizing. Proceedings of

2020
[29]

The Journal of Chemical Physics , volume =

Targeted Free Energy Estimation via Learned Mappings , author =. The Journal of Chemical Physics , volume =
[30]

Machine Learning: Science and Technology , volume =

Normalizing Flows for Atomic Solids , author =. Machine Learning: Science and Technology , volume =. doi:10.1088/2632-2153/ac6b16 , archiveprefix =. 2111.08696 , primaryclass =

arXiv
[31]

Estimating

Wirnsberger, Peter and Ibarz, Borja and Papamakarios, George , year = 2023, month = sep, journal =. Estimating

2023
[32]

Ding, Xinqiang and Zhang, Bin , year = 2021, month = mar, journal =

2021
[33]

Scalable

Schebek, Maximilian and No. Scalable. Nature Communications , volume =. doi:10.1038/s41467-026-73900-9 , copyright =

doi
[34]

doi:10.48550/arXiv.2512.23930 , archiveprefix =

Assessing Generative Modeling Approaches for Free Energy Estimates in Condensed Matter , author =. doi:10.48550/arXiv.2512.23930 , archiveprefix =. 2512.23930 , primaryclass =

doi
[35]

Estimating

Schebek, Maximilian and Frob. Estimating. doi:10.48550/arXiv.2512.18147 , archiveprefix =. 2512.18147 , primaryclass =

doi
[36]

and Shirts, Michael R

Olehnovics, Edgar and Liu, Yifei Michelle and Mehio, Nada and Sheikh, Ahmad Y. and Shirts, Michael R. and Salvalaglio, Matteo , year = 2024, month = jul, journal =. Assessing the

2024
[37]

Equilibrium Free Energies from Nonequilibrium Measurements Using Maximum-Likelihood Methods , author =. Phys. Rev. Lett. , volume =. 2003 , month =. doi:10.1103/PhysRevLett.91.140601 , url =

2003 doi
[38]

Particle

Andrieu, Christophe and Doucet, Arnaud and Holenstein, Roman , year = 2010, month = jun, journal =. Particle

2010
[39]

Performance of Machine-Learning-Assisted

Del Bono, Luca Maria and. Performance of Machine-Learning-Assisted. Physical Review E , volume =
[40]

and Langmore, Ian and Tran, Dustin and Vasudevan, Srinivas , year = 2019, month = mar, number =

Hoffman, Matthew and Sountsov, Pavel and Dillon, Joshua V. and Langmore, Ian and Tran, Dustin and Vasudevan, Srinivas , year = 2019, month = mar, number =. arXiv , keywords =:1903.03704 , primaryclass =

2019 arXiv
[41]

Transport

Cabezas, Alberto and Nemeth, Christopher , year = 2023, month = apr, pages =. Transport. Proceedings of

2023
[42]

and Marzouk, Youssef M

Parno, Matthew D. and Marzouk, Youssef M. , year = 2018, month = jan, journal =. Transport

2018
[43]

and Papaspiliopoulos, O

Agapiou, S. and Papaspiliopoulos, O. and. Importance. Statistical Science , volume =. 26408299 , eprinttype =
[44]

Efficient

Del Debbio, Luigi and Rossney, Joe Marsh and Wilson, Michael , year = 2021, month = nov, journal =. Efficient. doi:10.1103/PhysRevD.104.094507 , archiveprefix =. 2105.12481 , primaryclass =

2021 arXiv
[45]

Statistics and Computing , volume =

Annealed Importance Sampling , author =. Statistics and Computing , volume =. doi:10.1023/A:1008923215028 , keywords =

doi
[46]

and Wang, Jian-Sheng , year = 1986, month = nov, journal =

Swendsen, Robert H. and Wang, Jian-Sheng , year = 1986, month = nov, journal =. Replica

1986
[47]

Geyer, Charles J , year = 1991, pages =. Markov. Computing

1991
[48]

Exchange

Hukushima, Koji and Nemoto, Koji , year = 1996, month = jun, journal =. Exchange. doi:10.1143/JPSJ.65.1604 , keywords =

1996 doi
[49]

Sequential

Del Moral, Pierre and Doucet, Arnaud and Jasra, Ajay , year = 2006, month = jun, journal =. Sequential

2006
[50]

and Iba, Y

Hukushima, K. and Iba, Y. , year = 2003, month = nov, journal =. Population

2003
[51]

The European Physical Journal C , volume =

Learning Trivializing Flows , author =. The European Physical Journal C , volume =
[52]

Efficient Mapping of Phase Diagrams with Conditional

Schebek, Maximilian and Invernizzi, Michele and No. Efficient Mapping of Phase Diagrams with Conditional. Machine Learning: Science and Technology , volume =
[53]

Machine Learning: Science and Technology , volume =

Normalizing Flows as an Enhanced Sampling Method for Atomistic Supercooled Liquids , author =. Machine Learning: Science and Technology , volume =
[54]

Midgley, Laurence Illing and Stimper, Vincent and Simm, Gregor N. C. and Sch. Flow. The
[55]

and Bose, Joey and Lin, Chen and Klein, Leon and Bronstein, Michael M

Tan, Charlie B. and Bose, Joey and Lin, Chen and Klein, Leon and Bronstein, Michael M. and Tong, Alexander , year = 2025, month = oct, pages =. Scalable. Proceedings of the 42nd

2025
[56]

Stochastic

Wu, Hao and K. Stochastic. Advances in
[57]

Annealed

Arbel, Michael and Matthews, Alex and Doucet, Arnaud , year = 2021, month = jul, pages =. Annealed. Proceedings of the 38th

2021
[58]

Continual

Matthews, Alex and Arbel, Michael and Rezende, Danilo Jimenez and Doucet, Arnaud , year = 2022, month = jun, pages =. Continual. Proceedings of the 39th

2022
[59]

Journal of High Energy Physics , volume =

Stochastic Normalizing Flows as Non-Equilibrium Transformations , author =. Journal of High Energy Physics , volume =. doi:10.1007/JHEP07(2022)015 , keywords =

2022 doi
[60]

, year = 1997, month = apr, journal =

Jarzynski, C. , year = 1997, month = apr, journal =. Nonequilibrium

1997
[61]

, year = 1998, month = mar, journal =

Crooks, Gavin E. , year = 1998, month = mar, journal =. Nonequilibrium. doi:10.1023/A:1023208217925 , keywords =

1998 doi
[62]

and Crooks, Gavin E

Nilmeier, Jerome P. and Crooks, Gavin E. and Minh, David D. L. and Chodera, John D. , year = 2011, month = nov, journal =. Nonequilibrium Candidate

2011
[63]

Sampling the Lattice

Caselle, Michele and Cellini, Elia and Nada, Alessandro , year = 2024, month = feb, journal =. Sampling the Lattice. doi:10.1007/JHEP02(2024)048 , keywords =

2024 doi
[64]

Numerical Determination of the Width and Shape of the Effective String Using

Caselle, Michele and Cellini, Elia and Nada, Alessandro , year = 2025, month = feb, journal =. Numerical Determination of the Width and Shape of the Effective String Using. doi:10.1007/JHEP02(2025)090 , keywords =

2025 doi
[65]

and Nabergoj, David and Seljak, Uros , year = 2022, month = sep, journal =

Karamanis, Minas and Beutler, Florian and Peacock, John A. and Nabergoj, David and Seljak, Uros , year = 2022, month = sep, journal =. Accelerating Astronomical and Cosmological Inference with. doi:10.1093/mnras/stac2272 , archiveprefix =. 2207.05652 , primaryclass =

2022 arXiv
[66]

Accelerated

Zhang, Leo and Potaptchik, Peter and He, Jiajun and Du, Yuanqi and Doucet, Arnaud and Vargas, Francisco and Dau, Hai-Dang and Syed, Saifuddin , year = 2025, month = sep, number =. Accelerated. doi:10.48550/arXiv.2502.10328 , archiveprefix =. 2502.10328 , primaryclass =

2025 doi
[67]

Skipping the

Invernizzi, Michele and Kr. Skipping the. The Journal of Physical Chemistry Letters , volume =
[68]

Sampling

Kish, Leslie , year = 1965, journal =. Sampling. 2091346 , eprinttype =

1965
[69]

Flow Perturbation to Accelerate

Peng, Xin and Gao, Ang , year = 2025, month = jul, journal =. Flow Perturbation to Accelerate. doi:10.1038/s41467-025-62039-8 , copyright =

2025 doi
[70]

Unbiased

Wu, Dian and Rossi, Riccardo and Carleo, Giuseppe , year = 2021, month = nov, journal =. Unbiased. doi:10.1103/PhysRevResearch.3.L042024 , archiveprefix =. 2105.05650 , primaryclass =

2021 arXiv
[71]

Enhanced

H. Enhanced. Living Journal of Computational Molecular Science , volume =. doi:10.33011/livecoms.4.1.1583 , copyright =

doi
[72]

Speed-up of

Frenkel, Daan , year = 2004, month = dec, journal =. Speed-up of

2004
[73]

Generalized

Chen, Yunjie and Roux, Beno. Generalized. The Journal of Chemical Physics , volume =
[74]

Statistics and Computing , volume =

Sampling from Multimodal Distributions Using Tempered Transitions , author =. Statistics and Computing , volume =. doi:10.1007/BF00143556 , keywords =

doi
[75]

Physical Review E , volume =

Computation of a Chemical Potential Using a Residence Weight Algorithm , author =. Physical Review E , volume =
[76]

and Kucukelbir, Alp and McAuliffe, Jon D

Blei, David M. and Kucukelbir, Alp and McAuliffe, Jon D. , year = 2017, month = apr, journal =. Variational. doi:10.1080/01621459.2017.1285773 , archiveprefix =. 1601.00670 , primaryclass =

2017
[77]

and Ghahramani, Zoubin and Jaakkola, Tommi S

Jordan, Michael I. and Ghahramani, Zoubin and Jaakkola, Tommi S. and Saul, Lawrence K. , year = 1999, month = nov, journal =. An. doi:10.1023/A:1007665907178 , keywords =

1999 doi
[78]

Variational

Rezende, Danilo and Mohamed, Shakir , year = 2015, month = jun, pages =. Variational. Proceedings of the 32nd

2015
[79]

Wu, Dian and Wang, Lei and Zhang, Pan , year = 2019, month = feb, journal =. Solving. doi:10.1103/PhysRevLett.122.080602 , archiveprefix =. 1809.10606 , primaryclass =

2019 arXiv
[80]

Divergence Measures and Message Passing , author =
[81]

Blessing, Denis and Jia, Xiaogang and Esslinger, Johannes and Vargas, Francisco and Neumann, Gerhard , year = 2024, month = jul, pages =. Beyond. Proceedings of the 41st

2024
[82]

Physical Review D , volume =

Detecting and Mitigating Mode-Collapse for Flow-Based Sampling of Lattice Field Theories , author =. Physical Review D , volume =
[83]

Forty-Second

Temperature-. Forty-Second
[84]

Improving

Huang, Chin-Wei and Tan, Shawn and Lacoste, Alexandre and Courville, Aaron C , year = 2018, volume =. Improving. Advances in

2018
[85]

, year = 2025, month = may, number =

Wang, Yihang and Chi, Chris and Dinner, Aaron R. , year = 2025, month = may, number =. Mitigating Mode Collapse in Normalizing Flows by Annealing with an Adaptive Schedule:. doi:10.48550/arXiv.2505.03652 , archiveprefix =. 2505.03652 , primaryclass =

2025 doi arXiv
[86]

LeMinh, T. Natural. doi:10.48550/arXiv.2501.04667 , archiveprefix =. 2501.04667 , primaryclass =

doi
[87]

Nature Machine Intelligence , volume =

Variational Neural Annealing , author =. Nature Machine Intelligence , volume =. doi:10.1038/s42256-021-00401-3 , copyright =

doi
[88]

Machine Learning: Science and Technology , volume =

Lattice Protein Folding with Variational Annealing , author =. Machine Learning: Science and Technology , volume =
[89]

Markovian

Naesseth, Christian and Lindsten, Fredrik and Blei, David , year = 2020, volume =. Markovian. Advances in

2020
[90]

and Lederman, Roy R

Brofos, James and Gabrie, Marylou and Brubaker, Marcus A. and Lederman, Roy R. , year = 2022, month = may, pages =. Adaptation of the. Proceedings of

2022
[91]

arXiv , keywords =:2107.00734 , primaryclass =

Flow-Based Sampling for Multimodal Distributions in Lattice Field Theory , author =. arXiv , keywords =:2107.00734 , primaryclass =

arXiv
[92]

and Milo

McNaughton, B. and Milo. Boosting. Physical Review E , volume =. doi:10.1103/PhysRevE.101.053312 , archiveprefix =. 2002.04292 , primaryclass =

2002 arXiv
[93]

Machine-Learning-Assisted

Ciarella, Simone and Trinquier, Jeanne and Weigt, Martin and Zamponi, Francesco , year = 2023, month = mar, journal =. Machine-Learning-Assisted

2023
[94]

Demonstrating

Bono, Luca Maria Del and. Demonstrating. doi:10.48550/arXiv.2510.19544 , archiveprefix =. 2510.19544 , primaryclass =

doi arXiv
[95]

Choi, Kristy and Meng, Chenlin and Song, Yang and Ermon, Stefano , year = 2022, month = may, pages =. Density. Proceedings of

2022
[96]

Forty-First

Iterated. Forty-First
[97]

Salmona, Antoine and De Bortoli, Valentin and Delon, Julie and Desolneux, Agnes , editor =. Can. Advances in
[98]

Relaxing

Cornish, Rob and Caterini, Anthony and Deligiannidis, George and Doucet, Arnaud , year = 2020, journal =. Relaxing

2020
[99]

Theoretical Guarantees for Sampling and Inference in Generative Models with Latent Diffusions , booktitle =

Tzen, Belinda and Raginsky, Maxim , year = 2019, month = jun, pages =. Theoretical Guarantees for Sampling and Inference in Generative Models with Latent Diffusions , booktitle =

2019
[100]

Improved Sampling via Learned Diffusions , booktitle =
[101]

and Miller, Benjamin Kurt and Yan, Bing and

Havens, Aaron J. and Miller, Benjamin Kurt and Yan, Bing and. Adjoint. Forty-Second
[102]

Transactions on Machine Learning Research , issn =

An Optimal Control Perspective on Diffusion-Based Generative Modeling , author =. Transactions on Machine Learning Research , issn =
[103]

Efficient and

Zhang, Fengzhe and Midgley, Laurence Illing and. Efficient and. Transactions on Machine Learning Research , issn =
[104]

Particle

Phillips, Angus and Dau, Hai-Dang and Hutchinson, Michael John and Bortoli, Valentin De and Deligiannidis, George and Doucet, Arnaud , year = 2024, month = jul, pages =. Particle. Proceedings of the 41st

2024
[105]

He, Jiajun and Du, Yuanqi and Vargas, Francisco and Zhang, Dinghuai and Padhy, Shreyas and OuYang, RuiKang and Gomes, Carla P. and. No. Frontiers in
[106]

Learning

M. Learning. Transactions on Machine Learning Research , issn =
[107]

Liouville

Tian, Yifeng and Panda, Nishant and Lin, Yen Ting , year = 2024, month = jul, pages =. Liouville. Proceedings of the 41st

2024
[108]

Fan, Mingzhou and Zhou, Ruida and Tian, Chao and Qian, Xiaoning , year = 2024, month = jul, pages =. Path-. Proceedings of the 41st

2024
[109]

Transport Meets

Vargas, Francisco and Padhy, Shreyas and Blessing, Denis and N. Transport Meets. The
[110]

Forty-Second

Albergo, Michael Samuel and. Forty-Second
[111]

Escorted Free Energy Simulations: Improving Convergence by Reducing Dissipation , shorttitle =

Vaikuntanathan, Suriyanarayanan and Jarzynski, Christopher , year = 2008, month = may, journal =. Escorted Free Energy Simulations: Improving Convergence by Reducing Dissipation , shorttitle =. doi:10.1103/PhysRevLett.100.190601 , pmid =

2008 doi
[112]

, year = 1949, month = sep, journal =

Metropolis, Nicholas and Ulam, S. , year = 1949, month = sep, journal =. The. doi:10.1080/01621459.1949.10483310 , pmid =

1949
[113]

Journal of Research of the National Bureau of Standards , volume =

Various. Journal of Research of the National Bureau of Standards , volume =
[114]

and Rosenbluth, Marshall N

Metropolis, Nicholas and Rosenbluth, Arianna W. and Rosenbluth, Marshall N. and Teller, Augusta H. and Teller, Edward , year = 1953, month = jun, journal =. Equation of

1953
[115]

, year = 2004, series =

Liu, Jun S. , year = 2004, series =. Monte. doi:10.1007/978-0-387-76371-2 , isbn =

2004 doi
[116]

Tjelmeland, Hakon and Hegstad, Bjorn Kare , year = 2001, journal =. Mode. doi:10.1111/1467-9469.00232 , copyright =

2001
[117]

A Framework for Adaptive

Pompe, Emilia and Holmes, Chris and. A Framework for Adaptive. The Annals of Statistics , volume =. doi:10.1214/19-AOS1916 , keywords =

doi
[118]

An Introduction to Sequential

Chopin, Nicolas and Papaspiliopoulos, Omiros , year = 2020, series =. An Introduction to Sequential. doi:10.1007/978-3-030-47845-2 , isbn =

2020 doi
[119]

Highly Accurate Protein Structure Prediction with

Jumper, John and Evans, Richard and Pritzel, Alexander and Green, Tim and Figurnov, Michael and Ronneberger, Olaf and Tunyasuvunakool, Kathryn and Bates, Russ and. Highly Accurate Protein Structure Prediction with. Nature , volume =
[120]

Corso, Gabriele and St. The
[121]

Li, Shuo-Hui and Wang, Lei , year = 2018, month = dec, journal =. Neural. doi:10.1103/PhysRevLett.121.260601 , keywords =

2018 doi
[122]

and Kanwar, Gurtej and Racani

Abbott, Ryan and Botev, Aleksandar and Boyda, Denis and Hackett, Daniel C. and Kanwar, Gurtej and Racani. Applications of Flow Models to the Generation of Correlated Lattice. Physical Review D , volume =
[123]

Nature Reviews Physics , volume =

Advances in Machine-Learning-Based Sampling Motivated by Lattice Quantum Chromodynamics , author =. Nature Reviews Physics , volume =. doi:10.1038/s42254-023-00616-w , copyright =

doi
[124]

Anatole , year = 2024, month = apr, journal =

Coretti, Alessandro and Falkner, Sebastian and Weinreich, Jan and Dellago, Christoph and von Lilienfeld, O. Anatole , year = 2024, month = apr, journal =. Boltzmann

2024
[125]

Journal of Statistical Mechanics: Theory and Experiment , volume =

Learning to Sample Better , author =. Journal of Statistical Mechanics: Theory and Experiment , volume =
[126]

The Journal of Chemical Physics , volume =

Learning Mappings between Equilibrium States of Liquid Systems Using Normalizing Flows , author =. The Journal of Chemical Physics , volume =
[127]

, year = 1989, month = jan, journal =

Hutchinson, M.F. , year = 1989, month = jan, journal =. A. doi:10.1080/03610918908812806 , keywords =

1989 doi
[128]

Elucidating the

Karras, Tero and Aittala, Miika and Aila, Timo and Laine, Samuli , year = 2022, month = dec, journal =. Elucidating the

2022
[129]

Sampling,

Montanari, Andrea , year = 2023, month = may, number =. Sampling,. arXiv , keywords =:2305.10690 , primaryclass =

2023 arXiv
[130]

Diffusion

De Bortoli, Valentin and Thornton, James and Heng, Jeremy and Doucet, Arnaud , year = 2021, volume =. Diffusion. Advances in

2021
[131]

Uria, Benigno and C. Neural. Journal of Machine Learning Research , volume =
[132]

Proceedings of the 32nd

Germain, Mathieu and Gregor, Karol and Murray, Iain and Larochelle, Hugo , year = 2015, month = jun, pages =. Proceedings of the 32nd

2015
[133]

Sampling from the

Alaoui, Ahmed El and Montanari, Andrea and Sellke, Mark , year = 2022, month = oct, pages =. Sampling from the. 2022. doi:10.1109/FOCS54457.2022.00038 , keywords =

2022
[134]

Posterior

Montanari, Andrea and Wu, Yuchen , year = 2024, month = aug, number =. Posterior. doi:10.48550/arXiv.2304.11449 , archiveprefix =. 2304.11449 , primaryclass =

2024 doi arXiv
[135]

Proceedings of the National Academy of Sciences , volume =

Sampling with Flows, Diffusion, and Autoregressive Neural Networks from a Spin-Glass Perspective , author =. Proceedings of the National Academy of Sciences , volume =
[136]

Current Opinion in Structural Biology , volume =

Generative Molecular Dynamics , author =. Current Opinion in Structural Biology , volume =
[137]

Nature Machine Intelligence , volume =

Efficient Rare Event Sampling with Unsupervised Normalizing Flows , author =. Nature Machine Intelligence , volume =. doi:10.1038/s42256-024-00918-3 , copyright =

doi
[138]

and Dellago, Christoph , year = 2023, month = sep, journal =

Falkner, Sebastian and Coretti, Alessandro and Romano, Salvatore and Geissler, Phillip L. and Dellago, Christoph , year = 2023, month = sep, journal =. Conditioning

2023
[139]

Advances in Neural Information Processing Systems , author =

Timewarp:. Advances in Neural Information Processing Systems , author =
[140]

Implicit

Schreiner, Mathias and Winther, Ole and Olsson, Simon , year = 2023, month = dec, journal =. Implicit

2023
[141]

Transferable

Diez, Juan Viguera and Schreiner, Mathias and Olsson, Simon , year = 2025, month = oct, number =. Transferable. doi:10.48550/arXiv.2510.07589 , archiveprefix =. 2510.07589 , primaryclass =

2025 doi
[142]

Science , volume =

Scalable Emulation of Protein Equilibrium Ensembles with Generative Deep Learning , author =. Science , volume =
[143]

Collective

Wolff, Ulli , year = 1989, month = jan, journal =. Collective

1989
[144]

and Wang, Jian-Sheng , year = 1987, month = jan, journal =

Swendsen, Robert H. and Wang, Jian-Sheng , year = 1987, month = jan, journal =. Nonuniversal Critical Dynamics in

1987
[145]

Journal of Statistical Mechanics: Theory and Experiment , volume =

Efficient Swap Algorithms for Molecular Dynamics Simulations of Equilibrium Supercooled Liquids , author =. Journal of Statistical Mechanics: Theory and Experiment , volume =
[146]

Grigera, Tom. Fast. Physical Review E , volume =
[147]

and Krauth, Werner and Wilson, David B

Bernard, Etienne P. and Krauth, Werner and Wilson, David B. , year = 2009, month = nov, journal =. Event-Chain

2009
[148]

and Bair, Eric and Hooker, Giles and Pande, Vijay S

Shirts, Michael R. and Bair, Eric and Hooker, Giles and Pande, Vijay S. , year = 2003, month = oct, journal =. Equilibrium

2003
[149]

Geometric and Functional Analysis , publisher =

Thin shell implies spectral gap up to polylog via a stochastic localization scheme , author =. Geometric and Functional Analysis , publisher =
[150]

Probability Theory and Related Fields , publisher =

Taming correlations through entropy-efficient measure decompositions with applications to mean-field approximation , author =. Probability Theory and Related Fields , publisher =
[151]

Analysis of high-dimensional distributions using pathwise methods , author =
[152]

Localization schemes: A framework for proving mixing bounds for

Chen, Yuansi and Eldan, Ronen , year = 2022, booktitle =. Localization schemes: A framework for proving mixing bounds for

2022
[153]

, year = 2024, month = jun, journal =

Rotskoff, Grant M. , year = 2024, month = jun, journal =. Sampling Thermodynamic Ensembles of Molecular Systems with Generative Neural Networks:

2024
[154]

and Perdikaris, P

Raissi, M. and Perdikaris, P. and Karniadakis, G.E. , year = 2019, month = feb, journal =. Physics-Informed Neural Networks:

2019
[155]

Du, Yuanqi and He, Jiajun and Vargas, Francisco and Wang, Yuanqing and Gomes, Carla P. and. The
[156]

Chen, Han and Liu, Sifan and Yang, Jun , year = 2026, month = jul, number =. Markov. doi:10.48550/arXiv.2607.11631 , archiveprefix =. 2607.11631 , primaryclass =

2026 doi arXiv
[157]

M. Neural. The Journal of Physical Chemistry Letters , volume =
[158]

The Journal of Chemical Physics , volume =

Solvation Free Energies from Neural Thermodynamic Integration , author =. The Journal of Chemical Physics , volume =
[159]

, year = 1935, month = may, journal =

Kirkwood, John G. , year = 1935, month = may, journal =. Statistical

1935
[160]

Diffusion-Based

Grenioux, Louis and Noble, Maxence , year = 2026, month = feb, journal =. Diffusion-Based

2026
[161]

Wong, Kaze W. K. and Isi, Maximiliano and Edwards, Thomas D. P. , year = 2023, month = nov, journal =. Fast

2023
[162]

Physical Review D , volume =

Robust Parameter Estimation within Minutes on Gravitational Wave Signals from Binary Neutron Star Inspirals , author =. Physical Review D , volume =
[163]

and Tweedie, Richard L

Roberts, Gareth O. and Tweedie, Richard L. , year =. Exponential Convergence of. Bernoulli , volume =
[164]

Rossky, P. J. and Doll, J. D. and Friedman, H. L. , year =. Brownian Dynamics as Smart. The Journal of Chemical Physics , volume =
[165]

Duane, Simon and Kennedy, A. D. and Pendleton, Brian J. and Roweth, Duncan , year =. Hybrid. Physics Letters B , volume =
[166]

, year =

Neal, Radford M. , year =. Handbook of
[167]

and Gelman, Andrew , year =

Hoffman, Matthew D. and Gelman, Andrew , year =. The. Journal of Machine Learning Research , volume =
[168]

Leli. Free. 2010 , publisher =

2010
[169]

Annealing in Variational Inference Mitigates Mode Collapse:

Fogliani, Luigi and Loureiro, Bruno and Gabrié, Marylou , year = 2026, month = feb, number =. Annealing in Variational Inference Mitigates Mode Collapse:. doi:10.48550/arXiv.2602.12923 , archiveprefix =. 2602.12923 , primaryclass =

2026 doi
[170]

Riemannian

Grenioux, Louis and Galliano, Leonardo and Berthier, Ludovic and Biroli, Giulio and Gabrié, Marylou , year = 2025, month = dec, number =. Riemannian. doi:10.48550/arXiv.2512.16607 , archiveprefix =. 2512.16607 , primaryclass =

2025 doi arXiv
[171]

Efficient

Sch. Efficient. doi:10.48550/arXiv.2512.16812 , archiveprefix =. 2512.16812 , primaryclass =

doi arXiv
[172]

Machine Learning: Science and Technology , author =

A theoretical perspective on mode collapse in variational inference , volume =. Machine Learning: Science and Technology , author =. 2025 , note =. doi:10.1088/2632-2153/adde2a , abstract =

2025 doi
[173]

Improving the evaluation of samplers on multi-modal targets , url =

Grenioux, Louis and Noble, Maxence and Gabrié, Marylou , month = apr, year =. Improving the evaluation of samplers on multi-modal targets , url =
[174]

The Thirteenth International Conference on Learning Representations , author =

Learned. The Thirteenth International Conference on Learning Representations , author =. 2025 , keywords =

2025
[175]

Journal of Computational Physics , author =

Sampling metastable systems using collective variables and. Journal of Computational Physics , author =. 2025 , pages =. doi:10.1016/j.jcp.2025.113806 , abstract =

2025
[176]

Tamagnone, Samuel and Laio, Alessandro and Gabrié, Marylou , month = sep, year =. Coarse-. doi:10.1021/acs.jctc.4c00700 , journal =

doi
[177]

Stochastic

Grenioux, Louis and Noble, Maxence and Gabrié, Marylou and Durmus, Alain Oliviero , month = jul, year =. Stochastic. Proceedings of the 41st
[178]

Molina-Taborda, Ana and Cossio, Pilar and Lopez-Acevedo, Olga and Gabrié, Marylou , month = oct, year =. Active. doi:10.1021/acs.jctc.4c00506 , journal =

doi
[179]

Kaze W. k. Wong and Gabrié, Marylou and Daniel Foreman-Mackey , title =. Journal of Open Source Software , keywords =. 2023 , publisher =. doi:10.21105/joss.05021 , url =

2023 doi
[180]

Balanced

Grenioux, Louis and Moulines, Eric and Gabrié, Marylou , year =. Balanced
[181]

Grenioux, Louis and Durmus, Alain Oliviero and Moulines, Eric and Gabrié, Marylou , year =. On. Proceedings of the 40th
[182]

Optimizing

Schönle, Christoph and Gabrié, Marylou , month = oct, year =. Optimizing
[183]

Gabrié, Marylou and Ganguli, Surya and Lucibello, Carlo and Zecchina, Riccardo , year =. Neural. Spin. doi:10.1142/9789811273926_0024 , urldate =

doi
[184]

Local-Global

Sergey Samsonov and Evgeny Lagutin and Gabrié, Marylou and Alain Durmus and Alexey Naumov and Eric Moulines , booktitle=. Local-Global. 2022 , url=

2022
[185]

and Vanden-Eijnden, Eric , doi =

Gabrié, Marylou and Rotskoff, Grant M. and Vanden-Eijnden, Eric , doi =. Proceedings of the National Academy of Sciences , month =. 2022 , keywords =. arXiv , arxivId =:2105.12603 , file =

2022 arXiv
[186]

and Lederman, Roy R

Brofos, James and Gabrié, Marylou and Brubaker, Marcus A. and Lederman, Roy R. , year =. Adaptation of the. Proceedings of
[187]

Dawid, Anna and Arnold, Julian and Requena, Borja and Gresch, Alexander and P. Machine. doi:10.1017/9781009504942 , isbn =

doi
[188]

Lawrence, Hannah and Barmherzig, David and Li, Henry and Eickenberg, Michael and Gabrié, Marylou , year =. Phase. Proceedings of the 2nd
[189]

2021 , month = feb, number =

Dual. 2021 , month = feb, number =. arxiv , langid =:2107.05134 , primaryclass =

2021 arXiv
[190]

and Vanden-Eijnden, Eric , booktitle =

Gabrié, Marylou and Rotskoff, Grant M. and Vanden-Eijnden, Eric , booktitle =. 2021 , keywords =

2021
[191]

Barmherzig and Michael Eickenberg and Gabrié, Marylou , booktitle =

Hannah Lawrence and David A. Barmherzig and Michael Eickenberg and Gabrié, Marylou , booktitle =. Low-Photon Holographic Phase Retrieval via a Deep Decoder Neural Network , url =. 2021 , bdsk-url-1 =. doi:10.1364/AIS.2021.JTu5A.19 , journal =

2021 doi
[192]

2021 , volume =

On the Interplay between Data Structure and Loss Function in Classification Problems , booktitle =. 2021 , volume =

2021
[193]

Blind Calibration for Compressed Sensing: State Evolution and an Online Algorithm , shorttitle =

Gabrié, Marylou and Barbier, Jean and Krzakala, Florent and Zdeborov. Blind Calibration for Compressed Sensing: State Evolution and an Online Algorithm , shorttitle =. 2020 , month = jul, journal =. doi:10.1088/1751-8121/ab8416 , urldate =

2020 doi
[194]

2020 , month = jun, journal =

Mean-Field Inference Methods for Neural Networks , author =. 2020 , month = jun, journal =. doi:10.1088/1751-8121/ab7f65 , urldate =. arxiv , langid =:1911.00890 , primaryclass =

2020 arXiv
[195]

Towards an Understanding of Neural Networks : Mean-Field Incursions , shorttitle =

Gabrié, Marylou , year =. Towards an Understanding of Neural Networks : Mean-Field Incursions , shorttitle =
[196]

Blind Calibration for Sparse Regression: A State Evolution Analysis , year=

Gabrié, Marylou and Barbier, Jean and Krzakala, Florent and Zdeborová, Lenka , booktitle=. Blind Calibration for Sparse Regression: A State Evolution Analysis , year=. doi:10.1109/CAMSAP45676.2019.9022479 , keywords =

2019
[197]

Entropy and Mutual Information in Models of Deep Neural Networks , booktitle =

Gabrié, Marylou and Manoel, Andre and Luneau, Cl. Entropy and Mutual Information in Models of Deep Neural Networks , booktitle =. 2018 , volume =

2018
[198]

2017 , month = dec, journal =

Phase Transitions in the \ q\ -Coloring of Random Hypergraphs , author =. 2017 , month = dec, journal =. doi:10.1088/1751-8121/aa9529 , urldate =. arxiv , langid =:1707.01983 , primaryclass =

2017 arXiv
[199]

and Gabrié, Marylou and Manoel, Andre and Caltagirone, Francesco and Krzakala, Florent , year =

Tramel, Eric W. and Gabrié, Marylou and Manoel, Andre and Caltagirone, Francesco and Krzakala, Florent , year =. Deterministic and. Physical Review X , volume =. doi:10.1103/PhysRevX.8.041006 , urldate =

doi
[200]

and Manoel, Andre and Caltagirone, Francesco and Gabrié, Marylou and Krzakala, Florent , year =

Tramel, Eric W. and Manoel, Andre and Caltagirone, Francesco and Gabrié, Marylou and Krzakala, Florent , year =. Inferring Sparsity:. 2016. doi:10.1109/ITW.2016.7606837 , abstract =. arxiv , isbn =:1606.03956 , pages =

2016
[201]

and Krzakala, Florent , year =

Gabrié, Marylou and Tramel, Eric W. and Krzakala, Florent , year =. Training. Advances in Neural Information Processing Systems 28 , eprint =. doi:10.1109/CVPRW.2009.5206577 , abstract =

2009

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

Communications in Mathematical Sciences , volume =

Density Estimation by Dual Ascent of the Log-Likelihood , author =. Communications in Mathematical Sciences , volume =

[2] [2]

Normalizing

Papamakarios, George and Nalisnick, Eric and Rezende, Danilo Jimenez and Mohamed, Shakir and Lakshminarayanan, Balaji , year = 2021, journal =. Normalizing

2021

[3] [3]

and Brubaker, Marcus A

Kobyzev, Ivan and Prince, Simon J.D. and Brubaker, Marcus A. , year = 2021, month = nov, journal =. Normalizing

2021

[4] [4]

Density Estimation Using

Dinh, Laurent and. Density Estimation Using. International

[5] [5]

Durkan, Conor and Bekasov, Artur and Murray, Iain and Papamakarios, George , year = 2019, volume =. Neural. Advances in

2019

[6] [6]

Communications on Pure and Applied Mathematics , volume =

Polar Factorization and Monotone Rearrangement of Vector-Valued Functions , author =. Communications on Pure and Applied Mathematics , volume =. doi:10.1002/cpa.3160440402 , copyright =

doi:10.1002/cpa.3160440402

[7] [7]

Chen, Ricky T. Q. and Rubanova, Yulia and Bettencourt, Jesse and Duvenaud, David K , year = 2018, volume =. Neural. Advances in

2018

[8] [8]

Equivariant

K. Equivariant. Proceedings of the 37th