pith. sign in

arxiv: 2606.02455 · v1 · pith:CI3ZE2XJnew · submitted 2026-06-01 · 💻 cs.LG · cond-mat.mtrl-sci· physics.chem-ph· physics.comp-ph· stat.CO

Speculative Sampling For Faster Molecular Dynamics

Arthur Kosmala , Stephan G\"unnemann , Meng Gao , Brandon Wood This is my paper

Pith reviewed 2026-06-28 15:27 UTC · model grok-4.3

classification 💻 cs.LG cond-mat.mtrl-sciphysics.chem-phphysics.comp-phstat.CO
keywords molecular dynamicsspeculative samplingLangevin dynamicssimulation accelerationdraft modeltarget modeltransport map
0
0 comments X

The pith

Langevin Speculative Dynamics accelerates molecular dynamics 3-9 times by verifying draft steps with a target model without adding relative error.

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

The paper introduces Langevin Speculative Dynamics to overcome the serial bottleneck in molecular dynamics simulations. A fast draft model proposes candidate steps that a slower target model then verifies in parallel through a transport map that maps the draft distribution to the target. The method extends speculative sampling to second-order Langevin dynamics and derives the resulting speedup in terms of physical parameters. It reports that the approach works across multiple systems and model pairs while producing trajectories whose distribution matches the target exactly. A reader would care because this raises single-system throughput on existing hardware without changing the statistical properties of the output.

Core claim

LSD is a distributed and model-agnostic speculative sampler for molecular dynamics that uses a draft model to propose fast simulation steps and verifies them in parallel with a slower target model by applying a transport map from the draft to the target distribution. The work extends speculative sampling to second-order Langevin dynamics, derives the achievable speedup as a function of physical parameters, shows that the method generalizes across different systems and draft-target combinations with a 3-9x speedup, and confirms theoretically and empirically that LSD samples trajectories from its target model distribution.

What carries the argument

The transport map from draft to target distribution that preserves the exact target distribution when applied to second-order Langevin dynamics.

If this is right

  • LSD generalizes across different atomic systems and draft-target model pairs.
  • It delivers a 3-9x speedup expressed as a function of physical parameters.
  • Trajectories are drawn exactly from the target model distribution with no added relative error.
  • The parallel verification step can be performed on distributed hardware.

Where Pith is reading between the lines

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

  • The derived speedup formula could guide selection of draft models for new physical systems before running full simulations.
  • The same transport-map construction might be tried on other second-order integrators if an analogous distribution-preserving map can be found.
  • Because verification is parallelizable, LSD could combine with existing distributed MD frameworks to increase effective throughput on clusters.

Load-bearing premise

The transport map from the draft to the target distribution preserves the exact target distribution when extended to second-order Langevin dynamics.

What would settle it

Generate long trajectories with LSD and with the target model alone on the same system and compare any moment or correlation function of the resulting distributions; a statistically significant difference falsifies the no-relative-error claim.

read the original abstract

Molecular dynamics (MD) is a key tool for simulating the dynamical behavior of atomic systems. However, MD is inherently serial, which makes it difficult to increase single-system throughput with concurrent compute. To address this, we introduce Langevin Speculative Dynamics (LSD), a distributed and model-agnostic speculative sampler for accelerating MD without adding relative error. Inspired by speculative methods in language and diffusion modeling, LSD uses a draft model to propose fast simulation steps and verifies them in parallel with a slower target model, applying a transport map from the draft to the target distribution. We extend speculative sampling to second-order Langevin dynamics, derive the achievable speedup as a function of physical parameters, show that LSD generalizes across different systems and draft-target combinations with a 3-9x speedup, and confirm theoretically and empirically that LSD samples trajectories from its target model distribution.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces Langevin Speculative Dynamics (LSD), a model-agnostic speculative sampler for molecular dynamics. It uses a draft model to propose fast steps that are verified in parallel by a target model via a transport map, extends the approach to second-order (underdamped) Langevin dynamics, derives the achievable speedup from physical parameters, reports 3-9x empirical speedups across systems and draft-target pairs, and claims to sample trajectories exactly from the target distribution with no relative error.

Significance. If the no-bias claim holds, the work would offer a practical route to accelerate inherently serial MD simulations without introducing sampling error, which is valuable for computational chemistry and biophysics. The parameter-based speedup derivation and cross-system generalization would be additional strengths if shown to be rigorous and reproducible.

major comments (2)
  1. [Abstract] Abstract: the central no-relative-error claim requires that the transport map, when applied inside the speculative step for second-order Langevin dynamics, leaves the target model's stationary measure (and full trajectory measure) exactly invariant. Standard speculative proofs apply to discrete tokens or first-order overdamped Langevin; the extension to the coupled (x,v) process must correctly handle velocity marginals, friction, and symplectic structure, but the abstract provides no equations or proof outline for this step.
  2. [Abstract] Abstract (theoretical confirmation paragraph): the derivation of speedup 'as a function of physical parameters' is presented as load-bearing for the practical contribution, yet the abstract gives no indication of how the acceptance probability or correction term is defined for underdamped dynamics to ensure the speedup expression remains free of fitted quantities or hidden dependencies on the draft model.
minor comments (1)
  1. The abstract states that LSD 'generalizes across different systems and draft-target combinations,' but does not name the systems, metrics, or controls used to support this claim; these details belong in the main text or a dedicated results section.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive feedback. The concerns focus on the level of theoretical detail provided in the abstract regarding the invariance proof for second-order dynamics and the explicit form of the speedup derivation. The full manuscript contains these elements in Sections 3 and 4; we are prepared to strengthen the abstract accordingly while preserving its conciseness.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central no-relative-error claim requires that the transport map, when applied inside the speculative step for second-order Langevin dynamics, leaves the target model's stationary measure (and full trajectory measure) exactly invariant. Standard speculative proofs apply to discrete tokens or first-order overdamped Langevin; the extension to the coupled (x,v) process must correctly handle velocity marginals, friction, and symplectic structure, but the abstract provides no equations or proof outline for this step.

    Authors: Section 3 of the manuscript provides the required extension. The transport map is constructed to act jointly on (x, v) such that both the position and velocity marginals are preserved under the target underdamped dynamics; the friction term enters the acceptance ratio explicitly, and the map is designed to respect the underlying symplectic structure via a velocity-dependent correction that leaves the full trajectory measure invariant. A concise outline of this argument can be added to the abstract in revision. revision: partial

  2. Referee: [Abstract] Abstract (theoretical confirmation paragraph): the derivation of speedup 'as a function of physical parameters' is presented as load-bearing for the practical contribution, yet the abstract gives no indication of how the acceptance probability or correction term is defined for underdamped dynamics to ensure the speedup expression remains free of fitted quantities or hidden dependencies on the draft model.

    Authors: Section 4 derives the expected speedup directly from the physical parameters (friction coefficient, timestep, and draft-model fidelity) by expressing the acceptance probability as the Metropolis-Hastings ratio between the target and draft transition kernels for the underdamped process. The resulting expression contains no fitted constants or draft-specific parameters beyond the measurable density ratio; we will insert a parenthetical reference to this derivation in the abstract. revision: partial

Circularity Check

0 steps flagged

No circularity; derivation of speedup and distribution preservation is self-contained

full rationale

The paper derives the achievable speedup explicitly as a function of physical parameters and presents a theoretical confirmation that the transport map extension to second-order Langevin dynamics preserves the target model's stationary measure and trajectory distribution. No steps in the abstract reduce these claims to fitted inputs by construction, self-citations, or ansatzes smuggled from prior work; the speedup formula and invariance proof are positioned as independent derivations, with empirical 3-9x results treated as separate validation. The method is described as a direct extension of existing speculative sampling without load-bearing circular reductions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are identifiable from the provided text.

pith-pipeline@v0.9.1-grok · 5690 in / 1006 out tokens · 20215 ms · 2026-06-28T15:27:33.006854+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

58 extracted references · 13 canonical work pages

  1. [1]

    Proceedings of the 42nd International Conference on Machine Learning , pages =

    Accelerated Diffusion Models via Speculative Sampling , author =. Proceedings of the 42nd International Conference on Machine Learning , pages =. 2025 , editor =

    2025

  2. [3]

    , author=

    Robust and efficient configurational molecular sampling via Langevin dynamics. , author=. The Journal of chemical physics , year=

  3. [5]

    Proceedings of the 40th International Conference on Machine Learning , articleno =

    Leviathan, Yaniv and Kalman, Matan and Matias, Yossi , title =. Proceedings of the 40th International Conference on Machine Learning , articleno =. 2023 , publisher =

    2023

  4. [8]

    Applied Stochastic Differential Equations , publisher=

    Särkkä, Simo and Solin, Arno , year=. Applied Stochastic Differential Equations , publisher=

  5. [9]

    International Conference on Learning Representations , year=

    Score-Based Generative Modeling through Stochastic Differential Equations , author=. International Conference on Learning Representations , year=

  6. [10]

    Borgwardt and Malte J

    Arthur Gretton and Karsten M. Borgwardt and Malte J. Rasch and Bernhard Sch. A Kernel Two-Sample Test , journal =. 2012 , volume =

    2012

  7. [11]

    NeurIPS 2023 AI for Science Workshop , year=

    Learning Interatomic Potentials at Multiple Scales , author=. NeurIPS 2023 AI for Science Workshop , year=

    2023

  8. [12]

    The Thirteenth International Conference on Learning Representations , year=

    Towards Fast, Specialized Machine Learning Force Fields: Distilling Foundation Models via Energy Hessians , author=. The Thirteenth International Conference on Learning Representations , year=

  9. [13]

    Filippo Bigi and Sanggyu Chong and Agustinus Kristiadi and Michele Ceriotti , booktitle=. Flash. 2025 , url=

    2025

  10. [14]

    Lars Leon Schaaf and Ilyes Batatia and Jules Tilly and Thomas D Barrett , booktitle=. Boost. 2024 , url=

    2024

  11. [17]

    Accelerating Large Language Model Decoding with Speculative Sampling

    Chen, Charlie and Borgeaud, Sebastian and Irving, Geoffrey and Lespiau, Jean-Baptiste and Sifre, Laurent and Jumper, John , biburl =. Accelerating Large Language Model Decoding with Speculative Sampling. , url =. CoRR , keywords =

  12. [18]

    AMUSD: Asynchronous Multi-Device Speculative Decoding for LLM Acceleration , year=

    McDanel, Bradley , booktitle=. AMUSD: Asynchronous Multi-Device Speculative Decoding for LLM Acceleration , year=

  13. [19]

    Kitchin and Daniel S

    Brandon M Wood and Misko Dzamba and Xiang Fu and Meng Gao and Muhammed Shuaibi and Luis Barroso-Luque and Kareem Abdelmaqsoud and Vahe Gharakhanyan and John R. Kitchin and Daniel S. Levine and Kyle Michel and Anuroop Sriram and Taco Cohen and Abhishek Das and Sushree Jagriti Sahoo and Ammar Rizvi and Zachary Ward Ulissi and C. Lawrence Zitnick , booktitle...

    2025

  14. [20]

    Proceedings of the 42nd International Conference on Machine Learning , pages =

    The dark side of the forces: assessing non-conservative force models for atomistic machine learning , author =. Proceedings of the 42nd International Conference on Machine Learning , pages =. 2025 , editor =

    2025

  15. [21]

    International Conference on Learning Representations , year=

    Towards Training Billion Parameter Graph Neural Networks for Atomic Simulations , author=. International Conference on Learning Representations , year=

  16. [25]

    Huang, Kerson , biburl =

  17. [26]

    Proceedings of the 42nd International Conference on Machine Learning , pages =

    Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction , author =. Proceedings of the 42nd International Conference on Machine Learning , pages =. 2025 , editor =

    2025

  18. [27]

    , year =

    Lindvall, T. , year =. Lectures on the

  19. [28]

    2025 , eprint=

    Orb-v3: atomistic simulation at scale , author=. 2025 , eprint=

    2025

  20. [30]

    Timewarp: Transferable Acceleration of Molecular Dynamics by Learning Time-Coarsened Dynamics , volume =

    Klein, Leon and Foong, Andrew and Fjelde, Tor and Mlodozeniec, Bruno and Brockschmidt, Marc and Nowozin, Sebastian and Noe, Frank and Tomioka, Ryota , booktitle =. Timewarp: Transferable Acceleration of Molecular Dynamics by Learning Time-Coarsened Dynamics , volume =

  21. [31]

    Boltzmann priors for Implicit Transfer Operators , volume =

    Viguera Diez, Juan and Schreiner, Mathias and Engkvist, Ola and Olsson, Simon , booktitle =. Boltzmann priors for Implicit Transfer Operators , volume =

  22. [32]

    Ilyes Batatia and David Peter Kovacs and Gregor N. C. Simm and Christoph Ortner and Gabor Csanyi , booktitle=. 2022 , url=

    2022

  23. [34]

    2023 , publisher=

    Understanding molecular simulation: from algorithms to applications , author=. 2023 , publisher=

    2023

  24. [36]

    Problems and Projects , url =

    Seven strictures on similarity , author =. Problems and Projects , url =

  25. [37]

    Procrustes

    Michael P. Allen and Dominic J. Tildesley. Computer Simulation of Liquids. Oxford University Press, 06 2017. ISBN 9780198803195. doi:10.1093/oso/9780198803195.001.0001. https://doi.org/10.1093/oso/9780198803195.001.0001

    work page doi:10.1093/oso/9780198803195.001.0001 2017

  26. [38]

    Krishnapriyan

    Ishan Amin, Sanjeev Raja, and Aditi S. Krishnapriyan. Towards fast, specialized machine learning force fields: Distilling foundation models via energy hessians. In The Thirteenth International Conference on Learning Representations, 2025. https://openreview.net/forum?id=1durmugh3I

    2025

  27. [39]

    Ilyes Batatia, David Peter Kovacs, Gregor N. C. Simm, Christoph Ortner, and Gabor Csanyi. MACE : Higher order equivariant message passing neural networks for fast and accurate force fields. In Alice H. Oh, Alekh Agarwal, Danielle Belgrave, and Kyunghyun Cho, editors, Advances in Neural Information Processing Systems, 2022. https://openreview.net/forum?id=...

    2022

  28. [40]

    Mailoa, Mordechai Kornbluth, Nicola Molinari, Tess E

    Simon Batzner, Albert Musaelian, Lixin Sun, Mario Geiger, Jonathan P. Mailoa, Mordechai Kornbluth, Nicola Molinari, Tess E. Smidt, and Boris Kozinsky. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. Nature Communications, 13 0 (1), May 2022. ISSN 2041-1723. doi:10.1038/s41467-022-29939-5. http://dx.doi.org/10...

    work page doi:10.1038/s41467-022-29939-5 2022

  29. [41]

    Flash MD : long-stride, universal prediction of molecular dynamics

    Filippo Bigi, Sanggyu Chong, Agustinus Kristiadi, and Michele Ceriotti. Flash MD : long-stride, universal prediction of molecular dynamics. In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025 a . https://openreview.net/forum?id=ogZu06NgQs

    2025

  30. [42]

    Langer, and Michele Ceriotti

    Filippo Bigi, Marcel F. Langer, and Michele Ceriotti. The dark side of the forces: assessing non-conservative force models for atomistic machine learning. In Aarti Singh, Maryam Fazel, Daniel Hsu, Simon Lacoste-Julien, Felix Berkenkamp, Tegan Maharaj, Kiri Wagstaff, and Jerry Zhu, editors, Proceedings of the 42nd International Conference on Machine Learni...

    2025

  31. [43]

    Coupling and convergence for hamiltonian monte carlo

    Nawaf Bou-Rabee, Andreas Eberle, and Raphael Zimmer. Coupling and convergence for hamiltonian monte carlo. The Annals of Applied Probability, 30 0 (3), June 2020. ISSN 1050-5164. doi:10.1214/19-aap1528. http://dx.doi.org/10.1214/19-AAP1528

    work page doi:10.1214/19-aap1528 2020

  32. [44]

    Accelerating Molecular Dynamics Simulations with Foundation Neural Network Models Using Multiple Time Steps and Distillation

    Côme Cattin, Thomas Plé, Olivier Adjoua, Nicolaï Gouraud, Louis Lagardère, and Jean-Philip Piquemal. Accelerating Molecular Dynamics Simulations with Foundation Neural Network Models Using Multiple Time Steps and Distillation . The Journal of Physical Chemistry Letters, January 2026. doi:10.1021/acs.jpclett.5c03720. https://doi.org/10.1021/acs.jpclett.5c03720

    work page doi:10.1021/acs.jpclett.5c03720 2026

  33. [45]

    Accelerating large language model decoding with speculative sampling

    Charlie Chen, Sebastian Borgeaud, Geoffrey Irving, Jean-Baptiste Lespiau, Laurent Sifre, and John Jumper. Accelerating large language model decoding with speculative sampling. CoRR, abs/2302.01318, 2023. http://dblp.uni-trier.de/db/journals/corr/corr2302.html#abs-2302-01318

    Pith/arXiv arXiv 2023

  34. [46]

    Accelerated diffusion models via speculative sampling

    Valentin De Bortoli, Alexandre Galashov, Arthur Gretton, and Arnaud Doucet. Accelerated diffusion models via speculative sampling. In Aarti Singh, Maryam Fazel, Daniel Hsu, Simon Lacoste-Julien, Felix Berkenkamp, Tegan Maharaj, Kiri Wagstaff, and Jerry Zhu, editors, Proceedings of the 42nd International Conference on Machine Learning, volume 267 of Procee...

    2025

  35. [47]

    Kennedy, Brian J

    Simon Duane, A.D. Kennedy, Brian J. Pendleton, and Duncan Roweth. Hybrid monte carlo. Physics Letters B, 195 0 (2): 0 216--222, 1987. ISSN 0370-2693. doi:https://doi.org/10.1016/0370-2693(87)91197-X. https://www.sciencedirect.com/science/article/pii/037026938791197X

    work page doi:10.1016/0370-2693(87)91197-x 1987

  36. [48]

    Understanding molecular simulation: from algorithms to applications

    Daan Frenkel and Berend Smit. Understanding molecular simulation: from algorithms to applications. Elsevier, 2023

    2023

  37. [49]

    Learning interatomic potentials at multiple scales

    Xiang Fu, Albert Musaelian, Anders Johansson, Tommi Jaakkola, and Boris Kozinsky. Learning interatomic potentials at multiple scales. In NeurIPS 2023 AI for Science Workshop, 2023. https://openreview.net/forum?id=qFIs4hYZaZ

    2023

  38. [50]

    Levine, Meng Gao, Misko Dzamba, and C

    Xiang Fu, Brandon M Wood, Luis Barroso-Luque, Daniel S. Levine, Meng Gao, Misko Dzamba, and C. Lawrence Zitnick. Learning smooth and expressive interatomic potentials for physical property prediction. In Aarti Singh, Maryam Fazel, Daniel Hsu, Simon Lacoste-Julien, Felix Berkenkamp, Tegan Maharaj, Kiri Wagstaff, and Jerry Zhu, editors, Proceedings of the 4...

    2025

  39. [51]

    Borgwardt, Malte J

    Arthur Gretton, Karsten M. Borgwardt, Malte J. Rasch, Bernhard Sch \"o lkopf, and Alexander Smola. A kernel two-sample test. Journal of Machine Learning Research, 13 0 (25): 0 723--773, 2012. http://jmlr.org/papers/v13/gretton12a.html

    2012

  40. [52]

    A lithium superionic conductor

    Noriaki Kamaya, Kenji Homma, Yuichiro Yamakawa, Masaaki Hirayama, Ryoji Kanno, Masao Yonemura, Takashi Kamiyama, Yuki Kato, Shigenori Hama, Koji Kawamoto, and Akio Mitsui. A lithium superionic conductor. Nature Materials, 10 0 (9): 0 682--686, September 2011. ISSN 1476-4660. doi:10.1038/nmat3066. https://doi.org/10.1038/nmat3066

    work page doi:10.1038/nmat3066 2011

  41. [53]

    Timewarp: Transferable acceleration of molecular dynamics by learning time-coarsened dynamics

    Leon Klein, Andrew Foong, Tor Fjelde, Bruno Mlodozeniec, Marc Brockschmidt, Sebastian Nowozin, Frank Noe, and Ryota Tomioka. Timewarp: Transferable acceleration of molecular dynamics by learning time-coarsened dynamics. In A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, editors, Advances in Neural Information Processing Systems, volum...

    2023

  42. [54]

    Leimkuhler and Charles Matthews

    Benedict J. Leimkuhler and Charles Matthews. Robust and efficient configurational molecular sampling via langevin dynamics. The Journal of chemical physics, 138 17: 0 174102, 2013. https://api.semanticscholar.org/CorpusID:8476366

    2013

  43. [55]

    Fast inference from transformers via speculative decoding

    Yaniv Leviathan, Matan Kalman, and Yossi Matias. Fast inference from transformers via speculative decoding. In Proceedings of the 40th International Conference on Machine Learning, ICML'23. JMLR.org, 2023

    2023

  44. [56]

    Lindvall

    T. Lindvall. Lectures on the Coupling Method . Dover Books on Mathematics Series . Dover Publications, Incorporated, 2002. ISBN 978-0-486-42145-2. https://books.google.com/books?id=GUwyU1ypd1wC

    2002

  45. [57]

    How Concerted Are Ionic Hops in Inorganic Solid - State Electrolytes ? Journal of the American Chemical Society, 146 0 (12): 0 8269--8279, March 2024

    Cibrán López, Riccardo Rurali, and Claudio Cazorla. How Concerted Are Ionic Hops in Inorganic Solid - State Electrolytes ? Journal of the American Chemical Society, 146 0 (12): 0 8269--8279, March 2024. ISSN 0002-7863. doi:10.1021/jacs.3c13279. https://doi.org/10.1021/jacs.3c13279

    work page doi:10.1021/jacs.3c13279 2024

  46. [58]

    McCluskey, Alexander G

    Andrew R. McCluskey, Alexander G. Squires, Josh Dunn, Samuel W. Coles, and Benjamin J. Morgan. kinisi: Bayesian analysis of mass transport from molecular dynamics simulations. Journal of Open Source Software, 9 0 (94): 0 5984, 2024. doi:10.21105/joss.05984. https://doi.org/10.21105/joss.05984

    work page doi:10.21105/joss.05984 2024

  47. [59]

    Amusd: Asynchronous multi-device speculative decoding for llm acceleration

    Bradley McDanel. Amusd: Asynchronous multi-device speculative decoding for llm acceleration. In 2025 IEEE International Symposium on Circuits and Systems (ISCAS), pages 1--5, 2025. doi:10.1109/ISCAS56072.2025.11043575

    work page doi:10.1109/iscas56072.2025.11043575 2025

  48. [60]

    Self-learning hybrid monte carlo: A first-principles approach

    Yuki Nagai, Masahiko Okumura, Keita Kobayashi, and Motoyuki Shiga. Self-learning hybrid monte carlo: A first-principles approach. Phys. Rev. B, 102: 0 041124, Jul 2020. doi:10.1103/PhysRevB.102.041124. https://link.aps.org/doi/10.1103/PhysRevB.102.041124

    work page doi:10.1103/physrevb.102.041124 2020

  49. [61]

    Scalable parallel algorithm for graph neural network interatomic potentials in molecular dynamics simulations

    Yutack Park, Jaesun Kim, Seungwoo Hwang, and Seungwu Han. Scalable parallel algorithm for graph neural network interatomic potentials in molecular dynamics simulations. Journal of Chemical Theory and Computation, 20 0 (11): 0 4857–4868, May 2024. ISSN 1549-9626. doi:10.1021/acs.jctc.4c00190. http://dx.doi.org/10.1021/acs.jctc.4c00190

    work page doi:10.1021/acs.jctc.4c00190 2024

  50. [62]

    Orb-v3: atomistic simulation at scale, 2025

    Benjamin Rhodes, Sander Vandenhaute, Vaidotas Šimkus, James Gin, Jonathan Godwin, Tim Duignan, and Mark Neumann. Orb-v3: atomistic simulation at scale, 2025. https://arxiv.org/abs/2504.06231

    arXiv 2025

  51. [63]

    Boost MD : Accelerated molecular sampling leveraging ML force field features

    Lars Leon Schaaf, Ilyes Batatia, Jules Tilly, and Thomas D Barrett. Boost MD : Accelerated molecular sampling leveraging ML force field features. In NeurIPS 2024 Workshop on Data-driven and Differentiable Simulations, Surrogates, and Solvers, 2024. https://openreview.net/forum?id=H0USH61HnF

    2024

  52. [64]

    Score-based generative modeling through stochastic differential equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. In International Conference on Learning Representations, 2021. https://openreview.net/forum?id=PxTIG12RRHS

    2021

  53. [65]

    Lawrence Zitnick

    Anuroop Sriram, Abhishek Das, Brandon M Wood, and C. Lawrence Zitnick. Towards training billion parameter graph neural networks for atomic simulations. In International Conference on Learning Representations, 2022. https://openreview.net/forum?id=0jP2n0YFmKG

    2022

  54. [66]

    Applied Stochastic Differential Equations

    Simo Särkkä and Arno Solin. Applied Stochastic Differential Equations. Institute of Mathematical Statistics Textbooks. Cambridge University Press, 2019

    2019

  55. [67]

    Tuckerman, B

    M. Tuckerman, B. J. Berne, and G. J. Martyna. Reversible multiple time scale molecular dynamics. The Journal of Chemical Physics, 97 0 (3): 0 1990--2001, August 1992. ISSN 0021-9606. doi:10.1063/1.463137. https://doi.org/10.1063/1.463137

    work page doi:10.1063/1.463137 1990

  56. [68]

    experiments

    Loup Verlet. Computer "experiments" on classical fluids. i. thermodynamical properties of lennard-jones molecules. Phys. Rev., 159: 0 98--103, Jul 1967. doi:10.1103/PhysRev.159.98. https://link.aps.org/doi/10.1103/PhysRev.159.98

    work page doi:10.1103/physrev.159.98 1967

  57. [69]

    Boltzmann priors for implicit transfer operators

    Juan Viguera Diez, Mathias Schreiner, Ola Engkvist, and Simon Olsson. Boltzmann priors for implicit transfer operators. In Y. Yue, A. Garg, N. Peng, F. Sha, and R. Yu, editors, International Conference on Learning Representations, volume 2025, pages 38454--38482, 2025

    2025

  58. [70]

    Kitchin, Daniel S

    Brandon M Wood, Misko Dzamba, Xiang Fu, Meng Gao, Muhammed Shuaibi, Luis Barroso-Luque, Kareem Abdelmaqsoud, Vahe Gharakhanyan, John R. Kitchin, Daniel S. Levine, Kyle Michel, Anuroop Sriram, Taco Cohen, Abhishek Das, Sushree Jagriti Sahoo, Ammar Rizvi, Zachary Ward Ulissi, and C. Lawrence Zitnick. UMA : A family of universal models for atoms. In The Thir...

    2025