Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Beyond Scaling: Chemical Intuition as Emergent Ability of Universal Machine Learning Interatomic Potentials

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A universal machine-learning interatomic potential trained only on total energies and forces spontaneously learns per-bond dissociation energies that match literature values.

desk verdict A new diagnostic with a plausible but not fully pinned-down claim: per-edge energy decomposition yields BDE-like values, but gauge non-identifiability and missing per-bond statistics leave the emergence claim under-supported. read the letter →

arxiv 2506.07579 v1 pith:GHCHMELN submitted 2025-06-09 cond-mat.mtrl-sci

Shinnosuke Hattori , Kohei Shimamura , Aiichiro Nakano , Rajiv K. Kalia , Priya Vashishta , Ken-ichi Nomura This is my paper
classification cond-mat.mtrl-sci
keywords machinelearninginteratomicpotentialsbonddissociationenergyemergentabilitydecompositionE(3)-equivariantneuralnetworksAllegroscalinglawstransitionstates
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 paper argues that a machine-learning interatomic potential, trained only on global energies and forces, can spontaneously develop chemically meaningful per-bond energy terms without ever being shown a bond dissociation energy. The authors modify the Allegro architecture so that its output is a sum of per-edge energies, then train it on molecular and bulk datasets. They find that the symmetric part of each edge energy, $D_{ij} = \varepsilon_{ij} + \varepsilon_{ji}$, quantitatively matches experimental bond dissociation energies for bonds such as C-C, C-N, C-O, and C-H, including the correct ordering of single, double, and triple bonds. The effect persists even when training only on inorganic bulk materials data, and it sharpens as training data grows. The same decomposition also explains why activation-energy prediction plateaus: internal representations of C-N and C-O bonds in transition states remain diffuse even as stable-structure representations sharpen.

What carries the argument

The central machinery is the Edge-wise Emergent Decomposition (E3D) framework, built on the Allegro architecture's per-edge energy outputs $\varepsilon_{ij}$. By fixing the model's scale and shift parameters so that the cohesive energy is exactly the sum of edge energies, the paper defines the symmetric component $D_{ij} = \varepsilon_{ij} + \varepsilon_{ji}$ and the asymmetric component $A_{ij} = \varepsilon_{ij} - \varepsilon_{ji}$. The sum of $D_{ij}$ over all pairs equals the cohesive energy by construction, and $D_{ij}$ is interpreted as the bond energy, with $A_{ij}$ capturing local-environment asymmetry; two-dimensional $D_{ij}$-$A_{ij}$ histograms and their Shannon entropy serve as the diagnostic that tracks how sharply each bond type is represented as training data grows.

What would settle it

Train the same Allegro architecture on the same data with several random seeds and with at least one alternative radial cutoff, such as 4.0 Å instead of 5.2 Å, and compare the resulting $D_{ij}$ means to the reference bond dissociation energies; if the agreement scatters by more than the quoted distribution widths, or if an energy-preserving reparametrization of edge energies systematically shifts $D_{ij}$ while leaving total energies unchanged, then the correspondence between $D_{ij}$ and bond dissociation energy is not uniquely learned from the data.

Watch

Extended reading notes

Core claim

The central claim is that a universal machine learning interatomic potential trained only on total energies and forces spontaneously acquires per-edge energy components $D_{ij} = \varepsilon_{ij} + \varepsilon_{ji}$ whose means agree with literature bond dissociation energies across diverse bond types and bond orders, with no explicit supervision on bond energies. The authors demonstrate this agreement for C-H, N-H, O-H, C-C, C-N, C-O, N-O, and N-N bonds in a model trained on the SPICE 2 molecular dataset, and show that the agreement is qualitatively preserved when training exclusively on the MatPES inorganic bulk dataset. They further claim that data scaling organizes the learned representation, as measured by Shannon entropy of two-dimensional $D_{ij}$-$A_{ij}$ maps, and that the persistently high entropy of C-N and C-O transition-state representations coincides with the observed "scaling wall" in activation energy prediction. The paper thus proposes that decomposable learning of local bond energetics is an emergent ability of scaled equivariant networks, not an artifact of explicit BDE supervision.

Load-bearing premise

The claim that $D_{ij}$ is a physically meaningful bond energy assumes that the per-edge decomposition of the cohesive energy is identifiable, meaning that the specific value the network assigns to a particular C-C or C-N pair is the bond's true dissociation energy rather than an artifact of the chosen normalization, cutoff radius, or network initialization.

Editorial extensions

If this is right

  • If the claim is correct, machine learning interatomic potentials trained purely on energies and forces carry chemically interpretable per-bond information that can be read out directly without retraining or explicit labels.
  • The persistence of the emergent bond-dissociation energies after inorganic-only training implies that covalent bond energetics is a transferable representation that emerges from local atomic environments rather than from memorized molecular topologies.
  • The activation-energy scaling wall can be diagnosed and partially circumvented: training on a hybrid SPICE 2 plus MatPES dataset lowers the activation-energy mean absolute error from 0.58 eV to 0.44 eV, and the change is correlated with a reshaping of the $D_{ij}$-$A_{ij}$ maps for transition states.
  • The Shannon entropy of the $D_{ij}$-$A_{ij}$ maps provides a data-driven signal for when a bond type's internal representation is still diffuse, offering a way to spot dataset limitations for reactive structures before running expensive reference calculations.
  • The emergent per-bond energies could serve as collective variables in enhanced-sampling simulations and as real-time bond-dissociation progress monitors in molecular dynamics, as the paper suggests in its outlook.

Reading between the lines

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

  • A direct test of the framework's physical identifiability would be to train the same architecture with several random seeds and with different radial cutoffs; if the resulting $D_{ij}$ distributions all overlap within their quoted widths, the per-bond energies are robust, whereas wide scatter would indicate that the agreement with bond dissociation energies is partly a modeling convention.
  • An energy-preserving reparametrization of the edge energies, such as adding a divergence-free flow to $\varepsilon_{ij}$, could reveal whether $D_{ij}$ is gauge-invariant; total energies would stay unchanged while per-bond values shift, and a shift would show that the decomposition is not uniquely determined by the training objective.
  • The entropy transition observed near $10^5$ training structures suggests a concrete, testable prediction: models with insufficient capacity should not exhibit the sharp entropy drop even when given the full dataset, which could be checked by scanning tensor sizes across the data-size axis.
  • The same decomposition logic could be extended to other additive properties, such as decomposable dipole moments or partial charges, turning the E3D analysis from a bond-energy diagnostic into a general probe of what physics an equivariant potential has learned.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper introduces Edge-wise Emergent Decomposition (E3D), an analysis framework that decomposes the total cohesive energy of an Allegro machine-learning interatomic potential into symmetric (D_ij) and asymmetric (A_ij) per-edge components. By setting the per-species shift and scale parameters to 0 and 1 and training on E_coh, the model energy becomes a direct sum of edge energies. The authors report that the resulting D_ij distributions for C-H, N-H, O-H, C-C, C-N, C-O, N-O, and N-N bonds align with literature bond dissociation energies (BDEs) without explicit supervision. They also analyze 2D D-A histograms and their Shannon entropy H2D, observing that H2D decreases with increasing training data and that C-N and C-O bonds in transition states show weaker entropy reduction, coinciding with the observed 'scaling wall' for activation-energy prediction. A hybrid SPICE+MatPES training set is shown to improve E_a predictions. The paper interprets these results as evidence of 'emergent BDE' and as a diagnostic for scaling limitations.

Significance. If the central claim is valid, E3D provides a valuable, interpretable window into the internal representations of equivariant MLIPs and a practical metric for monitoring learning and diagnosing scaling failures. The claim that a model trained only on total energies and forces spontaneously develops quantitatively accurate bond dissociation energies is striking and, if true, would be an important contribution to the emerging understanding of MLIP generalization. The paper also proposes a concrete data-diversity strategy (hybrid training) that appears to improve reactive-property prediction. However, the significance is currently conditional: the identification of D_ij with physical BDEs rests on a gauge choice that is not ablated, and the quantitative evidence is weakened by the absence of per-bond statistics and uncertainties. These issues must be resolved for the results to be fully convincing.

major comments (3)
  1. [2.2 (Eqs. 5-7)] The central identification of D_ij with bond dissociation energy is not identifiable from the training objective. Setting mu=0 and sigma=1 makes Eq. (6) hold by construction, but any zero-sum re-partitioning of edge energies (e.g., adding a flow that cancels in the sum over neighbors) leaves E_coh and the forces invariant while changing D_ij. The paper provides no ablation of this gauge: no comparisons across random seeds, different cutoff radii, alternative decompositions (such as per-atom residuals), or models trained without the forced normalization. In addition, the definition of E_coh requires a choice of one-body reference energies E^(1)_i, which is not specified; if arbitrary offsets are used, the absolute scale of D_ij is correspondingly arbitrary. Without these controls, the agreement with reference BDEs in Fig. 2 could be an artifact of the specific gauge selected by the architecture and initialization rather than evidence that the model internalizes bond energetics.
  2. [3.1 and Fig. 3] The claim of quantitative agreement is not supported by the reported statistics. Fig. 2 shows distributions of D_ij but no numerical comparison per bond type; Fig. 3 reports only pooled metrics (Delta_BDE and sigma_BDE) averaged over all bond types, with no error bars across training seeds or test-set splits. Pooling can mask systematic deviations that cancel across bond types. The paper should report per-bond-type mean D_ij with standard deviations and confidence intervals against each reference BDE (e.g., a parity plot or a table analogous to Table S2), and ideally repeat training with a few random seeds to establish the stability of the decomposition. Without these, 'quantitatively agree' in the abstract and Section 3.1 is overstated.
  3. [3.4 and 4.3] The conclusion that E3D 'identifies the root cause of the scaling wall' is not justified by the evidence presented. The finding is a correlation: H2D for C-N and C-O bonds in transition-state structures plateaus in the same data range where E_a MAE saturates. No causal test or intervention is provided, and the same framework yields an apparent inconsistency: the hybrid dataset improves E_a accuracy while increasing H2D (Section 4.3, Table S4, and Figure S2). This suggests the relationship between representational entropy and reactive accuracy is more complex than a monotonic 'low entropy implies good reactivity' narrative. The authors should either soften the causal language to 'hypothesis' or provide a direct test (e.g., fine-tuning with selected TS structures and showing that the entropy change precedes E_a improvement).
minor comments (6)
  1. [2.2] The notation in Eq. (5) is unclear: the per-species scale sigma_Zi appears in the outer sum and sigma_ZiZj in the inner sum. Please clarify the relationship between these scales and the statement that both are set to 1. Also specify how the one-body reference energies for E_coh are computed (see also Major Comment 1).
  2. [3.2] The claim of 'robust across diverse training datasets' is based on a single additional dataset (MatPES), with only qualitative agreement (Fig. S1). Please qualify the wording or add more datasets to support the generality claim.
  3. [3.4 / Fig. 5] The 'sharp decrease in entropy around 10^5 data size' is identified by visual inspection. Please provide a statistical test or at least error bars to support the claim of a discrete transition point rather than a smooth trend.
  4. [Throughout] The manuscript contains numerous typos and formatting issues, e.g., 'trainig' in Section 4.1, incomplete sentences ('and s.' in Section 3.1), and a missing y-axis label in Figure 2. A careful proofreading pass is needed.
  5. [A.1] The hyperparameters are said to be 'based on information extracted from arXiv:2502.06073'. Please clarify whether the models used in this paper share the exact architecture of Allegro-FM or whether modifications (e.g., the final edge-energy MLP with a single hidden layer and no nonlinearity) were made. This detail is important for reproducibility.
  6. [Table S2 / Sec. 2.3] The reference BDEs are a mix of experimental, theoretical, and interpolated values, but Section 2.3 states that reference values are from '[36]' alone. Please be explicit about the source of each value, especially for bond orders where interpolation was used.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: BDE emergence is compared against external references; the only minor self-citation (Allegro-FM) is not load-bearing.

full rationale

The central claim — that an Allegro model trained on total energies/forces yields per-edge symmetric energies D_ij that match literature BDEs — is genuinely external. Equation (6) (E_coh = sum ε_ij) holds by construction because the authors set μ=0 and σ=1, and the identity sum D_ij = E_coh is definitional. But the paper never conditions a loss or fit on the reference BDEs [36,48]; the comparison in Figures 2 and S1 is a post-training evaluation against an external table. Thus the main 'emergence of BDE' result does not reduce to its inputs. The identifiability/gauge concern raised about D_ij is a validity issue (the per-edge decomposition is not unique), not a circularity: the paper does not define BDE as D_ij nor fit D_ij to BDE. The H2D-entropy/scaling-wall narrative is a post-hoc correlation between two observables (Ea MAE and map entropy) computed on the same T1x structures; it is not a derivation in which one quantity is defined in terms of the other. The only notable self-citation is [31] (Allegro-FM by the same group), used to motivate hybrid-data diversity and to provide hyperparameters. This is not load-bearing: the hybrid-dataset improvement is independently demonstrated in Table 2 of this paper, and the BDE comparison relies on external data. Section 5.4 candidly limits the claim to Allegro, which further shows the authors do not use a self-citation to rule out alternatives. No circular step meets the bar of Eq. X = Eq. Y by construction or fitted-parameter-renamed-as-prediction.

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

The central result rests on an architectural gauge choice that makes per-edge energies carry all cohesive energy, the suitability of literature BDEs as a reference for aggregated D_ij distributions, and the interpretation of histogram entropy as representational quality. No fitted parameters target the BDE values, but analysis hyperparameters such as cutoff and bin width are chosen by hand without sensitivity checks.

free parameters (2)
  • Radial cutoff r_cut = 5.2 A = 5.2 A
    Chosen from prior Allegro-FM training and not ablated; the per-edge decomposition and therefore D_ij values depend on which atom pairs are included as edges.
  • Histogram bin width for D-A maps = 0.3 eV
    Shannon entropy H2D values and their scaling trends depend on the choice of uniform bin width; no sensitivity analysis is reported.
assumptions (5)
  • domain assumption The cohesive energy can be exactly expressed as a sum of directed edge energies (Eq. 6) because Allegro's architecture and the zero-shift, unit-scale gauge force all energy into per-edge terms.
    This is an architectural constraint rather than a physical law; it makes D_ij well-defined but not automatically physical.
  • domain assumption Reference BDEs from Lange's Handbook and theoretical calculations are appropriate benchmarks for the model's D_ij distributions.
    Reference BDEs are measured or computed for specific molecules, while D_ij distributions aggregate many local environments; comparability is assumed.
  • domain assumption GAFF atom typing with manual corrections correctly assigns bond multiplicities.
    Bond type labels drive the grouping of D_ij values; manual validation is described but not quantified.
  • domain assumption Lower Shannon entropy of D-A histograms indicates a more organized and chemically meaningful representation.
    This interpretive step links an information-theoretic quantity to model quality; the link is plausible but not independently validated.
  • domain assumption Allegro training runs reached convergence such that scaling comparisons are meaningful.
    Convergence is asserted via a learning-rate scheduler and a minimum of 100 epochs, but no loss curves are shown.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Beyond Scaling: Chemical Intuition as Emergent Ability of Universal Machine Learning Interatomic Potentials." pith.science (2026). https://pith.science/paper/GHCHMELN

@misc{pith2026250607579,
  author       = {Pith},
  title        = {Pith review of: Beyond Scaling: Chemical Intuition as Emergent Ability of Universal Machine Learning Interatomic Potentials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GHCHMELN}},
  note         = {Machine review of arXiv:2506.07579}
}
read the original abstract

Machine Learning Interatomic Potentials (MLIPs) have successfully demonstrated scaling behavior, i.e. the power-law improvement in training performance, however the emergence of novel capabilities at scale remains unexplored. We have developed Edge-wise Emergent Decomposition (E3D) framework to investigate how an MLIP develops the ability to derive physically meaningful local representations of chemical bonds without explicit supervision. Employing an E(3)-equivariant network (Allegro) trained on molecular data (SPICE~2), we found that the trained MLIP spontaneously learned representations of bond dissociation energy (BDE) by decomposing the global potential energy landscape. The learned BDE values quantitatively agree with literature and its scalability are found to be robust across diverse training datasets, suggesting the presence of underlying representation that captures chemical reactions faithfully beyond given training information. Our E3D analysis utilizing Shannon's entropy reveals a close interplay between the decomposability of potential energy learning, scalability of learning, and emergent chemical reactivity, thus providing novel insights of scaling limitations and pathways toward more physically interpretable and predictive simulations.

Figures

Figures reproduced from arXiv: 2506.07579 by the authors.

Figure 1
Figure 1. Scaling behavior of prediction errors for reaction energies ( [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Emergence of BDE through E3D analysis for [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Evolution of BDE metrics with training data size from E3D analysis. (a) BDE distribution [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Evolution of learned energy representations through E3D analysis. 2 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Quantitative evolution of internal representations with respect to the training data size. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Toward Exascale AI for Science: A Scalable AI Skill for Autonomous Microkinetics Discovery

    cs.CE 2026-06 unverdicted novelty 5.0 of 10

    Introduces a scalable AI skill framework for autonomous microkinetics discovery that automates workflows and evaluates surrogate reliability.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages · cited by 1 Pith paper

  1. [1]

    Generalized neural-network representation of high-dimensional potential-energy surfaces.Phys

    Jörg Behler and Michele Parrinello. Generalized neural-network representation of high-dimensional potential-energy surfaces.Phys. Rev. Lett., 98(14):146401, April 2007

    work page 2007

  2. [2]

    Gaussian approximation potentials: the accuracy of quantum mechanics, without the electrons.Phys

    Albert P Bartók, Mike C Payne, Risi Kondor, and Gábor Csányi. Gaussian approximation potentials: the accuracy of quantum mechanics, without the electrons.Phys. Rev. Lett., 104(13):136403, April 2010

    work page 2010

  3. [3]

    A foundation model for atomistic materials chemistry.arXiv [physics.chem-ph], December 2023

    Ilyes Batatia, Philipp Benner, Yuan Chiang, Alin M Elena, Dávid P Kovács, Janosh Riebesell, Xavier R Advincula, Mark Asta, William J Baldwin, Noam Bernstein, Arghya Bhowmik, Samuel M Blau, Vlad C˘arare, James P Darby, Sandip De, Flaviano Della Pia, V olker L Deringer, Rokas Elijošius, Zakariya El-Machachi, Edvin Fako, Andrea C Ferrari, Annalena Genreith-S...

    work page 2023

  4. [4]

    Scaling deep learning for materials discovery.Nature, November 2023

    Amil Merchant, Simon Batzner, Samuel S Schoenholz, Muratahan Aykol, Gowoon Cheon, and Ekin Dogus Cubuk. Scaling deep learning for materials discovery.Nature, November 2023

    work page 2023

  5. [5]

    Scaling the leading accuracy of deep equivariant models to biomolecular simulations of realistic size

    Boris Kozinsky, Albert Musaelian, Anders Johansson, and Simon Batzner. Scaling the leading accuracy of deep equivariant models to biomolecular simulations of realistic size. InProceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, number Article 2 in SC ’23, pages 1–12, New York, NY , USA, November 20...

    work page 2023

  6. [6]

    SchNetPack 2.0: A neural network toolbox for atomistic machine learning.J

    Kristof T Schütt, Stefaan S P Hessmann, Niklas W A Gebauer, Jonas Lederer, and Michael Gastegger. SchNetPack 2.0: A neural network toolbox for atomistic machine learning.J. Chem. Phys., 158(14):144801, April 2023

    work page 2023

  7. [7]

    DeePMD-kit v2: A software package for deep potential models.J

    Jinzhe Zeng, Duo Zhang, Denghui Lu, Pinghui Mo, Zeyu Li, Yixiao Chen, Marián Rynik, Li’ang Huang, Ziyao Li, Shaochen Shi, Yingze Wang, Haotian Ye, Ping Tuo, Jiabin Yang, Ye Ding, Yifan Li, Davide Tisi, Qiyu Zeng, Han Bao, Yu Xia, Jiameng Huang, Koki Muraoka, Yibo Wang, Junhan Chang, Fengbo Yuan, Sigbjørn Løland Bore, Chun Cai, Yinnian Lin, Bo Wang, Jiayan...

    work page 2023

  8. [8]

    Performant implementation of the atomic cluster expansion (PACE) and application to copper and silicon.npj Computational Materials, 7(1):1–12, June 2021

    Yury Lysogorskiy, Cas van der Oord, Anton Bochkarev, Sarath Menon, Matteo Rinaldi, Thomas Hammer- schmidt, Matous Mrovec, Aidan Thompson, Gábor Csányi, Christoph Ortner, and Ralf Drautz. Performant implementation of the atomic cluster expansion (PACE) and application to copper and silicon.npj Computational Materials, 7(1):1–12, June 2021

    work page 2021

Show all 49 references
  1. [9]

    Atomic cluster expansion for accurate and transferable interatomic potentials.Phys

    Ralf Drautz. Atomic cluster expansion for accurate and transferable interatomic potentials.Phys. Rev. B Condens. Matter, 99(1):014104, January 2019

    2019

  2. [10]

    E(3)-equivariant graph neural networks for data- efficient and accurate interatomic potentials.Nat

    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.Nat. Commun., 13(1):2453, May 2022. 12

    2022

  3. [11]

    MACE: Higher order equivariant message passing neural networks for fast and accurate force fields.arXiv [stat.ML], June 2022

    Ilyes Batatia, Dávid Péter Kovács, Gregor N C Simm, Christoph Ortner, and Gábor Csányi. MACE: Higher order equivariant message passing neural networks for fast and accurate force fields.arXiv [stat.ML], June 2022

    2022

  4. [12]

    MACE-OFF23: Transferable machine learning force fields for organic molecules.arXiv [physics.chem-ph], December 2023

    Dávid Péter Kovács, J Harry Moore, Nicholas J Browning, Ilyes Batatia, Joshua T Horton, Venkat Kapil, William C Witt, Ioan-Bogdan Magd˘au, Daniel J Cole, and Gábor Csányi. MACE-OFF23: Transferable machine learning force fields for organic molecules.arXiv [physics.chem-ph], Dec...

    2023

  5. [13]

    CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling.Nature Machine Intelligence, 5(9):1031–1041, September 2023

    Bowen Deng, Peichen Zhong, Kyujung Jun, Janosh Riebesell, Kevin Han, Christopher J Bartel, and Gerbrand Ceder. CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling.Nature Machine Intelligence, 5(9):1031–1041, September 2023

    2023

  6. [14]

    Nutmeg and SPICE: Models and data for biomolecular machine learning.J

    Peter Eastman, Benjamin P Pritchard, John D Chodera, and Thomas E Markland. Nutmeg and SPICE: Models and data for biomolecular machine learning.J. Chem. Theory Comput., 20(19):8583–8593, October 2024

    2024

  7. [15]

    SPICE, a dataset of drug-like molecules and peptides for training machine learning potentials

    Peter Eastman, Pavan Kumar Behara, David L Dotson, Raimondas Galvelis, John E Herr, Josh T Horton, Yuezhi Mao, John D Chodera, Benjamin P Pritchard, Yuanqing Wang, Gianni De Fabritiis, and Thomas E Markland. SPICE, a dataset of drug-like molecules and peptides for training mac...

    2023

  8. [16]

    A dataset of 175k stable and metastable materials calculated with the PBEsol and SCAN functionals.Sci Data, 9(1):1–8, March 2022

    Jonathan Schmidt, Hai-Chen Wang, Tiago F T Cerqueira, Silvana Botti, and Miguel A L Marques. A dataset of 175k stable and metastable materials calculated with the PBEsol and SCAN functionals.Sci Data, 9(1):1–8, March 2022

    2022

  9. [17]

    Complexity of many-body interactions in transition metals via machine-learned force fields from the TM23 data set.Npj Comput

    Cameron J Owen, Steven B Torrisi, Yu Xie, Simon L Batzner, Kyle Bystrom, J Coulter, Albert Musaelian, Lixin Sun, and B Kozinsky. Complexity of many-body interactions in transition metals via machine-learned force fields from the TM23 data set.Npj Comput. Mater ., 10(1):1–16, F...

    2023

  10. [18]

    Open materials 2024 (OMat24) inorganic materials dataset and models.arXiv [cond-mat.mtrl-sci], October 2024

    Luis Barroso-Luque, Muhammed Shuaibi, Xiang Fu, Brandon M Wood, Misko Dzamba, Meng Gao, Ammar Rizvi, C Lawrence Zitnick, and Zachary W Ulissi. Open materials 2024 (OMat24) inorganic materials dataset and models.arXiv [cond-mat.mtrl-sci], October 2024

    2024

  11. [19]

    The open molecules 2025 (OMol25) dataset, evaluations, and models.arXiv [physics.chem-ph], May 2025

    Daniel S Levine, Muhammed Shuaibi, Evan Walter Clark Spotte-Smith, Michael G Taylor, Muhammad R Hasyim, Kyle Michel, Ilyes Batatia, Gábor Csányi, Misko Dzamba, Peter Eastman, Nathan C Frey, Xiang Fu, Vahe Gharakhanyan, Aditi S Krishnapriyan, Joshua A Rackers, Sanjeev Raja, Amm...

    2025

  12. [20]

    A foundational potential energy surface dataset for materials.arXiv [cond-mat.mtrl-sci], March 2025

    Aaron D Kaplan, Runze Liu, Ji Qi, Tsz Wai Ko, Bowen Deng, Janosh Riebesell, Gerbrand Ceder, Kristin A Persson, and Shyue Ping Ong. A foundational potential energy surface dataset for materials.arXiv [cond-mat.mtrl-sci], March 2025

    2025

  13. [21]

    Application of pretrained universal machine-learning interatomic potential for physicochemical simulation of liquid electrolytes in li-ion battery.Digit

    Suyeon Ju, Jinmu You, Gijin Kim, Yutack Park, Hyungmin An, and Seungwu Han. Application of pretrained universal machine-learning interatomic potential for physicochemical simulation of liquid electrolytes in li-ion battery.Digit. Discov., May 2025

    2025

  14. [22]

    The open catalyst 2020 (OC20) dataset and community challenges.arXiv [cond-mat.mtrl-sci], October 2020

    Lowik Chanussot, Abhishek Das, Siddharth Goyal, Thibaut Lavril, Muhammed Shuaibi, Morgane Riviere, Kevin Tran, Javier Heras-Domingo, Caleb Ho, Weihua Hu, Aini Palizhati, Anuroop Sriram, Brandon Wood, Junwoong Yoon, Devi Parikh, C Lawrence Zitnick, and Zachary Ulissi. The open ...

    2020

  15. [23]

    CatTSunami: Accelerating transition state energy calculations with pretrained graph neural networks.ACS Catal., pages 5283–5294, March 2025

    Brook Wander, Muhammed Shuaibi, John R Kitchin, Zachary W Ulissi, and C Lawrence Zitnick. CatTSunami: Accelerating transition state energy calculations with pretrained graph neural networks.ACS Catal., pages 5283–5294, March 2025

    2025

  16. [24]

    Transferable machine learning interatomic potential for bond dissociation energy prediction of drug-like molecules.J

    Elena Gelžinyt˙e, Mario Öeren, Matthew D Segall, and Gábor Csányi. Transferable machine learning interatomic potential for bond dissociation energy prediction of drug-like molecules.J. Chem. Theory Comput., 20(1):164–177, January 2024

    2024

  17. [25]

    Revisiting aspirin polymorphic stability using a machine learning potential.ACS Omega, 0(0):null, August 2024

    Shinnosuke Hattori and Qiang Zhu. Revisiting aspirin polymorphic stability using a machine learning potential.ACS Omega, 0(0):null, August 2024

    2024

  18. [26]

    Atomistic modeling of the mechanical properties: the rise of machine learning interatomic potentials.Mater

    Bohayra Mortazavi, Xiaoying Zhuang, Timon Rabczuk, and Alexander V Shapeev. Atomistic modeling of the mechanical properties: the rise of machine learning interatomic potentials.Mater . Horiz., 10(6):1956– 1968, 2023. 13

    1956

  19. [27]

    Machine learning interatomic potential for modeling the mechanical and thermal properties of naphthyl-based nanotubes.J

    Hugo X Rodrigues, Hudson R Armando, Daniel A da Silva, João Paulo J da Costa, Luiz A Ribeiro, Jr, and Marcelo L Pereira, Jr. Machine learning interatomic potential for modeling the mechanical and thermal properties of naphthyl-based nanotubes.J. Chem. Theory Comput., 21(5):261...

    2025

  20. [28]

    Thermal conductivity calculation using homogeneous non-equilibrium molecular dynamics simulation with allegro

    Kohei Shimamura, Shinnosuke Hattori, Ken-Ichi Nomura, Akihide Koura, and Fuyuki Shimojo. Thermal conductivity calculation using homogeneous non-equilibrium molecular dynamics simulation with allegro. Int. J. Heat Mass Transf., 234(126106):126106, December 2024

    2024

  21. [29]

    Exploring the frontiers of condensed-phase chemistry with a general reactive machine learning potential.Nat

    Shuhao Zhang, Małgorzata Z Mako´s, Ryan B Jadrich, Elfi Kraka, Kipton Barros, Benjamin T Nebgen, Sergei Tretiak, Olexandr Isayev, Nicholas Lubbers, Richard A Messerly, and Justin S Smith. Exploring the frontiers of condensed-phase chemistry with a general reactive machine lear...

    2024

  22. [30]

    The open catalyst 2022 (OC22) dataset and challenges for oxide electrocatalysts.ACS Catal., 13(5):3066–3084, March 2023

    Richard Tran, Janice Lan, Muhammed Shuaibi, Brandon M Wood, Siddharth Goyal, Abhishek Das, Javier Heras-Domingo, Adeesh Kolluru, Ammar Rizvi, Nima Shoghi, Anuroop Sriram, Félix Therrien, Jehad Abed, Oleksandr V oznyy, Edward H Sargent, Zachary Ulissi, and C Lawrence Zitnick. T...

    2022

  23. [31]

    Allegro-FM: Towards equivariant foundation model for exascale molecular dynamics simulations.arXiv [physics.comp-ph], February 2025

    Ken-Ichi Nomura, Shinnosuke Hattori, Satoshi Ohmura, Ikumi Kanemasu, Kohei Shimamura, Nabankur Dasgupta, Aiichiro Nakano, Rajiv K Kalia, and Priya Vashishta. Allegro-FM: Towards equivariant foundation model for exascale molecular dynamics simulations.arXiv [physics.comp-ph], F...

    2025

  24. [32]

    Taming multi-domain, -fidelity data: Towards foundation models for atomistic scale simulations.arXiv [physics.chem-ph], December 2024

    Shiota Tomoya, Ishihara Kenji, Tuan Minh Do, Mori Toshio, and Mizukami Wataru. Taming multi-domain, -fidelity data: Towards foundation models for atomistic scale simulations.arXiv [physics.chem-ph], December 2024

    2024

  25. [33]

    Transition1x - a dataset for building generalizable reactive machine learning potentials.Sci

    Mathias Schreiner, Arghya Bhowmik, Tejs Vegge, Jonas Busk, and Ole Winther. Transition1x - a dataset for building generalizable reactive machine learning potentials.Sci. Data, 9(1):779, December 2022

    2022

  26. [34]

    Learning local equivariant representations for large-scale atomistic dynamics.Nat

    Albert Musaelian, Simon Batzner, Anders Johansson, Lixin Sun, Cameron J Owen, Mordechai Kornbluth, and Boris Kozinsky. Learning local equivariant representations for large-scale atomistic dynamics.Nat. Commun., 14(1):579, February 2023

    2023

  27. [35]

    AmberTools.J

    David A Case, Hasan Metin Aktulga, Kellon Belfon, David S Cerutti, G Andrés Cisneros, Vinícius Wilian D Cruzeiro, Negin Forouzesh, Timothy J Giese, Andreas W Götz, Holger Gohlke, Saeed Izadi, Koushik Kasavajhala, Mehmet C Kaymak, Edward King, Tom Kurtzman, Tai-Sung Lee, Pengfe...

    2023

  28. [36]

    McGraw-Hill Education, Columbus, OH, 17 edition, October 2016

    James Speight.Lange’s handbook of chemistry, seventeenth edition. McGraw-Hill Education, Columbus, OH, 17 edition, October 2016

    2016

  29. [37]

    Emergent abilities of large language models.arXiv [cs.CL], June 2022

    Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, Ed H Chi, Tatsunori Hashimoto, Oriol Vinyals, Percy Liang, Jeff Dean, and William Fedus. Emergent abilities of large language models.arX...

    2022

  30. [38]

    Emergent abilities in large language models: A survey

    Leonardo Berti, Flavio Giorgi, and Gjergji Kasneci. Emergent abilities in large language models: A survey. arXiv [cs.LG], February 2025

    2025

  31. [39]

    Embedded-atom method: Derivation and application to impurities, surfaces, and other defects in metals.Phys

    Murray S Daw and M I Baskes. Embedded-atom method: Derivation and application to impurities, surfaces, and other defects in metals.Phys. Rev. B, 29(12):6443, June 1984

    1984

  32. [40]

    John, and Robert S Paton

    Shree Sowndarya, Yeonjoon Kim, Seonah Kim, Peter C St. John, and Robert S Paton. Expansion of bond dissociation prediction with machine learning to medicinally and environmentally relevant chemical space. Digit. Discov., 2(6):1900–1910, December 2023

    1900

  33. [41]

    Entropy-based guid- ance of deep neural networks for accelerated convergence and improved performance.Inf

    Mackenzie J Meni, Ryan T White, Michael L Mayo, and Kevin R Pilkiewicz. Entropy-based guid- ance of deep neural networks for accelerated convergence and improved performance.Inf. Sci. (Ny), 681(121239):121239, October 2024

    2024

  34. [42]

    Information theory unifies atomistic machine learning, uncertainty quantification, and materials thermodynamics.arXiv [cond-mat.mtrl-sci], April 2024

    Daniel Schwalbe-Koda, Sebastien Hamel, Babak Sadigh, Fei Zhou, and Vincenzo Lordi. Information theory unifies atomistic machine learning, uncertainty quantification, and materials thermodynamics.arXiv [cond-mat.mtrl-sci], April 2024. 14

    2024

  35. [43]

    Collective variables for the study of long-time kinetics from molecular trajectories: theory and methods.Curr

    Frank Noé and Cecilia Clementi. Collective variables for the study of long-time kinetics from molecular trajectories: theory and methods.Curr . Opin. Struct. Biol., 43:141–147, April 2017

    2017

  36. [44]

    Graph atomic cluster expansion for semilocal interactions beyond equivariant message passing.Phys

    Anton Bochkarev, Yury Lysogorskiy, and Ralf Drautz. Graph atomic cluster expansion for semilocal interactions beyond equivariant message passing.Phys. Rev. X., 14(2):021036, June 2024

    2024

  37. [45]

    PhysNet: A neural network for predicting energies, forces, dipole moments and partial charges.arXiv [physics.chem-ph], February 2019

    Oliver T Unke and Markus Meuwly. PhysNet: A neural network for predicting energies, forces, dipole moments and partial charges.arXiv [physics.chem-ph], February 2019

    2019

  38. [46]

    Learning molecular potentials with neural networks.Wiley Interdiscip

    Hatice Gokcan and Olexandr Isayev. Learning molecular potentials with neural networks.Wiley Interdiscip. Rev. Comput. Mol. Sci., 12(2), March 2022

    2022

  39. [47]

    Psi4 1.4: Open-source software for high-throughput quantum chemistry.J

    Daniel G A Smith, Lori A Burns, Andrew C Simmonett, Robert M Parrish, Matthew C Schieber, Raimondas Galvelis, Peter Kraus, Holger Kruse, Roberto Di Remigio, Asem Alenaizan, Andrew M James, Susi Lehtola, Jonathon P Misiewicz, Maximilian Scheurer, Robert A Shaw, Jeffrey B Schrib...

    2020

  40. [48]

    Theoretical bond energies: A critical evaluation.Journal of Physical Chemistry A, 105:3407–3416, March 2001

    Kai Exner And and P Schleyer. Theoretical bond energies: A critical evaluation.Journal of Physical Chemistry A, 105:3407–3416, March 2001

    2001

  41. [49]

    Type- Decomposition

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling laws for neural language models.arXiv [cs.LG], January 2020. 15 A Appendices A.1 Software/Hardware environment CodeThis work us...

    2020 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.