Pith. sign in

REVIEW 4 major objections 4 minor 45 references

Fine-Tuning Small Language Models for Reliable VASP INCAR Generation

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

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

Pith's one-line read A fine-tuned 4-billion-parameter model paired with a rule-based checker beats GPT-5.4 at generating VASP input files.

desk verdict The win over GPT-5.4 is a measured fact, but 'reliable' outruns the evidence when the benchmark and the post-processor share the same rule table. read the letter →

arxiv 2608.05387 v1 pith:VW6PMAKP submitted 2026-08-05 cond-mat.mtrl-sci

Xinyue Zhang , Jixiang Li , Bin Shao , Baishun Yang , Zhiyang Liu , Weichao Wang This is my paper · ORCID
classification cond-mat.mtrl-sci
keywords VASPINCARgenerationsmalllanguagemodelsfine-tuningLoRArule-basedpost-processingINCARBenchdensityfunctionaltheory
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to show that reliable VASP INCAR generation does not require a large proprietary cloud language model. It claims that a small open model, fine-tuned on reference VASP calculations and followed by a deterministic rule-based post-processor, reaches an INCAR Score of 89.88 on INCARBench, 15.55 points above GPT-5.4, while staying small enough to run on a single GPU. The payoff would be local, offline, and privacy-preserving high-throughput materials workflows with no per-call API costs. The paper also finds that model size saturates quickly: once fine-tuning and post-processing are applied, a 4-billion-parameter model slightly outperforms an 8-billion-parameter one in the same family.

What carries the argument

The central object is INCAR-SLM, a two-stage pipeline: a small language model fine-tuned with LoRA on thousands of reference VASP calculations generates an INCAR draft, and VASPGuard, a deterministic post-processor, parses the draft, checks workflow controls against the requested calculation type, and applies material-dependent corrections (DFT+U arrays, MAGMOM lengths, smearing, symmetry settings) only when the draft passes an evidence gate. The evaluation metric is the INCAR Score, the average of a Must-match component on task-critical tags and a Policy-match component on numerical settings within tolerances. VASPGuard's rules encode the same calculation conventions used to build the training set and benchmark, which is what lets the post-processor fix the errors the model leaves behind.

What would settle it

Run the INCAR files produced by INCAR-SLM and by GPT-5.4 through VASP on the INCARBench cases and compare convergence rates and physical outputs; if the small model's files do not converge or reproduce reference results at least as well as GPT-5.4's, the reported 15.55-point gap would reflect benchmark alignment rather than real INCAR quality.

Watch

Extended reading notes

Core claim

On INCARBench's 500-case evaluation, the paper reports that INCAR-SLM built on Qwen3-4B scores 89.88 on the 100-point INCAR Score, outperforming every general-purpose LLM evaluated, with GPT-5.4 at 74.33. Most of the gain comes from fine-tuning on the 4,453-case INCAR Training Set; VASPGuard then corrects the remaining workflow and material-dependent errors, such as forcing ionic-relaxation controls for geometry optimization and charge-density reuse for non-self-consistent runs. The paper further reports that performance saturates at a few billion parameters, with Qwen3-4B slightly ahead of Qwen3-8B, and that fine-tuning plus VASPGuard lifts even 0.6B-parameter models above 80 points.

Load-bearing premise

The load-bearing premise is that INCARBench's scoring is a faithful measure of real INCAR quality; the paper gives no check by running VASP or using an external benchmark, and the same conventions that define the benchmark also define VASPGuard's rules.

Editorial extensions

If this is right

  • A locally deployable model can match or exceed large proprietary cloud models on structured scientific input generation, removing per-request API costs and data-privacy constraints from high-throughput workflows.
  • Once fine-tuning and rule-based post-processing are in place, model size beyond a few billion parameters yields little: the 4-billion-parameter model slightly beats the 8-billion-parameter one.
  • Most of the reliability gain comes from fine-tuning; VASPGuard contributes a smaller but necessary correction layer, especially on DFT+U and symmetry settings.
  • The same split between learned draft generation and checkable rule correction should transfer to other input formats with explicit syntax and constraints, such as Quantum ESPRESSO or LAMMPS inputs.

Reading between the lines

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

  • A direct validation the paper does not report would run the generated INCAR files through VASP and compare convergence and physical results against expert-written files; that would test whether the INCAR Score tracks real scientific utility.
  • Because the benchmark's Must-match criteria and VASPGuard's workflow rules encode the same conventions, part of the reported advantage over GPT-5.4 could reflect benchmark alignment rather than general INCAR competence; an independent expert-written test set would separate the two.
  • The evidence-gate design means VASPGuard cannot rescue a poor model: the Llama-3.2-3B case suggests draft quality sets a ceiling, so the fine-tuning step remains the load-bearing component for reliability.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes INCAR-SLM, a locally deployable pipeline for generating VASP INCAR files from a natural-language request and a POSCAR. The pipeline combines a LoRA-fine-tuned small language model with VASPGuard, a deterministic rule-based post-processor that parses the draft, checks workflow and composition-dependent settings, and applies corrections. On INCARBench, a benchmark introduced by the same group in ref. [19], the Qwen3-4B-based INCAR-SLM achieves an INCAR Score of 89.88, exceeding GPT-5.4 by 15.55 points, while remaining small enough for single-GPU deployment. The paper reports controlled experiments separating the contributions of fine-tuning and VASPGuard across three model families, and finds that model-size scaling saturates at a few billion parameters. The authors conclude that a small adapted model with rule-based correction can match or exceed a much larger proprietary model for this task.

Significance. If the result holds, it is practically significant: it would make reliable INCAR generation feasible for local, offline, high-throughput materials workflows without relying on proprietary cloud APIs. The paper has notable strengths: it ships the training set, evaluation split, fine-tuning and inference code, VASPGuard source, and a SHA-256 manifest; the evaluation conditions in Section 2.5 are clearly described; and the decomposition of fine-tuning versus post-processing contributions is transparent and useful. The finding that fine-tuning rather than parameter count drives most of the gain is interesting and actionable. However, the central claim of "reliable" generation is measured against a benchmark authored by the same group, and the scoring criteria overlap substantially with the rules that VASPGuard itself enforces. The paper also performs no VASP runs or external validation, and it does not report uncertainty or error analysis for the headline numbers.

major comments (4)
  1. [Section 2.5, Eq. (2), and Section S6] The INCAR Score is defined as the average of Must match and Policy match, and Section S6 states that runnable checks were retained as diagnostics but excluded from the score. No VASP execution of any generated INCAR is reported anywhere in the manuscript. As a result, the headline 89.88 score measures agreement with the authors' reference conventions, not whether the generated files are runnable or physically sound. The abstract and conclusions repeatedly use the word "reliable"; that claim requires at least a subset of generated INCARs to be validated by actual VASP runs, or a clear statement that reliability is being used only in the sense of benchmark agreement.
  2. [Section 2.4, Table S5, and Section S6] The benchmark is partially self-referential. INCARBench was created by the same group (ref. [19]), and the Must-match criteria in Section S6 check exactly the workflow tags that VASPGuard enforces in Table S5 (IBRION=2/NSW=99 for relaxations, ICHARG=11 for DOS/band NSCF, static-SCF controls, and similar). Because the scoring is built from the same Materials Project/pymatgen conventions used to construct the training set and VASPGuard rules, any draft passing the evidence gate automatically earns large Must-match credit. The comparison is not wholly circular—the fine-tuned model alone scores 75.87 without VASPGuard and still exceeds GPT-5.4's 74.33—but the 15.55-point gap between INCAR-SLM and GPT-5.4 likely overstates the practical advantage when measured against an independent protocol or against runnable INCAR quality.
  3. [Section 3.4 and Figure 5] The claim that "performance saturates at a few billion parameters" is supported only by two pairwise comparisons: Qwen3-4B versus Qwen3-8B (89.88 versus 89.72, a 0.16-point difference) and Gemma3-4B versus Gemma3-12B (89.00 versus 89.40, a 0.40-point difference). These differences are far smaller than the likely case-level sampling noise on a 500-case benchmark, yet the paper reports no confidence intervals, bootstrap estimates, or repeated evaluations. The Llama3.2-3B outlier is explained by draft quality, which is reasonable, but excluding it means the saturation conclusion rests on very few points per family. The authors should provide uncertainty quantification or additional model sizes before making a definitive scaling statement.
  4. [Section 2.4 and Section S7] VASPGuard can alter or reject drafts, and the paper presents only successful correction examples (Table 1 and Section S8). There is no analysis of VASPGuard's false-correction rate, the cases where its evidence gate passes but the rule-based correction is wrong, or the failure modes that remain after post-processing. Since VASPGuard is deterministic, its errors would be systematic and could silently corrupt many generated files; quantifying this behavior is essential for the paper's reliability claim.
minor comments (4)
  1. [Title and Abstract] The title has a spacing typo in "V ASP INCAR"; it should read "VASP INCAR."
  2. [Section 3.2] The sentence "The other nine models show no such failure" is vague; specify which nine models and define the criterion for "no such failure" (for example, the number of drafts passing the evidence gate).
  3. [Section 3.3 and Figure 4] The category-level scores are shown without sample sizes or confidence intervals; a category with only a few cases could drive large percentage-point differences, so the authors should report per-category case counts.
  4. [Section S2, Table S1] The table lists per-model learning rates, but the main text gives only a pointer to Section S2; consider stating in the main text that all models use the same LoRA rank, alpha, and dropout, since these values are important for reproducibility.

Circularity Check

2 steps flagged · score 6.0 of 10

Partial circularity: VASPGuard hardcodes the same workflow tags INCARBench's Must-match rubric rewards, and the benchmark is the authors' own self-cited construction with runnable checks excluded.

  1. fitted input called prediction [Section 2.4 and 2.5; SI Table S5 and Section S6]
    "INCARBench scores each generated INCAR against a normalized reference using two components: Must match, which checks whether task-critical entries—relaxation controls, charge-inheritance settings for DOS or band workflows, DFT+U, MAGMOM, and symmetry-related parameters—are explicit and semantically correct ... VASPGuard ... enforcing ionic-relaxation controls for geometry optimization, or the fixed-charge-density convention used for DOS and band-structure calculations. ... Table S5: Geometry relaxation IBRION=2, NSW=99 ... Line-mode bands ICHARG=11, ISYM=0 ..."

    VASPGuard is a deterministic post-processor whose rule table rewrites qualifying drafts to exactly the workflow tags that INCARBench's Must-match component scores: relaxation controls (IBRION/NSW), charge inheritance for NSCF (ICHARG), symmetry control (ISYM), and DOS-grid controls (NEDOS, ISMEAR). Once a draft passes the evidence gate, Must-match credit on these task-defining entries is guaranteed by VASPGuard rather than produced by the language model. The reported 15.55-point advantage over GPT-5.4 is therefore largely a measure of overlap between the hand-coded post-processor rules and the authors' own scoring rubric, not an independent prediction of model competence.

  2. self citation load bearing [SI Section S6; Reference [19]]
    "Optional tags, extra tags, runnable checks, and KPOINTS-policy checks were retained as diagnostics but were not included in this score."

    The headline claim that INCAR-SLM 'outperforms every general-purpose LLM evaluated, exceeding GPT-5.4 by 15.55 points' is measured solely on INCARBench, whose authors (Bin Shao, Jixiang Li, Xinyue Zhang, Baishun Yang, Zhiyang Liu, Weichao Wang) are identical to those of the present paper. The benchmark score deliberately excludes runnable checks, so 'reliability' is operationalized as conformance to the same Materials Project/pymatgen conventions that the authors built into both the training set and VASPGuard. No external benchmark or actual VASP execution independently validates the central comparison; the self-citation is load-bearing because it supplies the only metric for the paper's main claim.

full rationale

The paper is not fully circular: the fine-tuned model alone reaches 75.87 on the same benchmark, narrowly exceeding GPT-5.4's 74.33, and the paper openly separates fine-tuning gains from VASPGuard post-processing. However, the headline 89.88 score and the 15.55-point margin depend on VASPGuard, whose rules enforce the same workflow tags INCARBench's Must-match component rewards, and the benchmark itself is the authors' own prior work with identical authorship. Because runnable checks are excluded from the score, the metric cannot falsify the system's own conventions; the central 'reliability' claim is thus partially constructed rather than independently validated. This warrants a partial-circularity score of 6 rather than a higher score, since some independent content remains in the fine-tuned-model comparison.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The claims are empirical and no physical constants are fitted. The main assumptions concern benchmark validity, the correctness of Materials Project-derived reference data, and the effectiveness of the deduplication fingerprint. VASPGuard is a software component, not a newly postulated physical entity.

free parameters (1)
  • LoRA hyperparameters (rank, alpha, dropout, learning rates) = rank=32, alpha=64, dropout=0.05, stage-1 LR 5e-5 to 8e-5, stage-2 LR 1.5e-5 to 2e-5
    Hand-selected standard values reported in Section S2; they affect model quality but were not tuned to the benchmark outcome, so they do not alter the qualitative claim.
assumptions (4)
  • domain assumption INCARBench Must/Policy scoring is a faithful proxy for the quality of generated INCAR files.
    The central performance claim is defined through this benchmark; the paper provides no external validation (e.g., running VASP) that higher scores correspond to more correct physics.
  • domain assumption Materials Project reference INCARs used as training targets and benchmark references are themselves correct VASP inputs.
    The training set and benchmark are derived from Materials Project records (Sections 2.2 and S4); if those references are systematically incorrect or outdated, the evaluation inherits the errors.
  • domain assumption The normalized POSCAR fingerprint exclusion is sufficient to prevent train/benchmark overlap.
    Section S4 claims no benchmark-structure overlap, but the fingerprint may not catch near-identical or compositionally similar structures, so similar examples could leak into training.
  • domain assumption Standard deep learning generalization assumptions: the fine-tuned model's performance on 500 benchmark cases reflects performance on unseen INCAR requests.
    The model is trained on 4,453 cases and evaluated on 500; no distribution-shift analysis is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fine-Tuning Small Language Models for Reliable VASP INCAR Generation." pith.science (2026). https://pith.science/paper/VW6PMAKP

@misc{pith2026260805387,
  author       = {Pith},
  title        = {Pith review of: Fine-Tuning Small Language Models for Reliable VASP INCAR Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VW6PMAKP}},
  note         = {Machine review of arXiv:2608.05387}
}
read the original abstract

Language models can prepare VASP INCAR files from natural-language requests, but so far only large proprietary cloud models come close to handling the tightly coupled, physics-sensitive settings reliably, a dependence that fits poorly with local, high-throughput materials workflows where privacy, cost, and offline deployment matter. We show that a small language model (SLM) can close this gap. The SLM is fine-tuned on reference VASP calculations and paired with VASPGuard, a deterministic post-processor that checks syntax, workflow, and material-dependent constraints; we call the combined model INCAR-SLM. On INCARBench, a benchmark for VASP INCAR generation, INCAR-SLM built on Qwen3-4B outperforms every general-purpose LLM evaluated, exceeding GPT-5.4 by 15.55 points on the 100-point INCAR Score. Most of this gain comes from fine-tuning, with VASPGuard correcting the errors that remain. We further find that model size matters less than expected: once fine-tuning and post-processing are applied, performance saturates at a few billion parameters, and Qwen3-4B outperforms larger models in the same family.

Figures

Figures reproduced from arXiv: 2608.05387 by the authors.

Figure 1
Figure 1. Overview of the INCAR-SLM workflow. Given a calculation task and a POSCAR [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Ranking of final INCAR Scores. INCAR-SLM models and general-purpose LLMs are [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Contributions of domain fine-tuning and VASPGuard to the INCAR Score across [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Decomposition of the INCAR Score across challenging categories for Qwen3-4B. Each [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Model scaling trends under three states: foundation, domain fine-tuning, and the [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 14 canonical work pages

  1. [19]

    Incarbench: A benchmark for scientific configuration in vasp incar by large language models,

    Bin Shao, Jixiang Li, Xinyue Zhang, Baishun Yang, Zhiyang Liu, and Weichao Wang. Incarbench: A benchmark for scientific configuration in vasp incar by large language models,

    work page

  2. [1]

    Ge Lei, Ronan Docherty, and Samuel J. Cooper. Materials science in the era of large language models: a perspective.Digital Discovery, 3(7):1257–1272, 2024. doi: 10.1039/ D4DD00074A

    work page 2024

  3. [2]

    Prince, Henry Chan, Aikaterini Vriza, Tao Zhou, Varuni K

    Michael H. Prince, Henry Chan, Aikaterini Vriza, Tao Zhou, Varuni K. Sastry, Yanqi Luo, et al. Opportunities for retrieval and tool augmented large language models in scientific facilities.npj Computational Materials, 10(1):251, 2024. doi: 10.1038/s41524-024-01423-2

    doi:10.1038/s41524-024-01423-2 2024

  4. [3]

    Lever- aging large language models for predictive chemistry.Nature Machine Intelligence, 6(2): 161–169, 2024

    KevinMaikJablonka, PhilippeSchwaller, AndresOrtega-Guerrero, andBerendSmit. Lever- aging large language models for predictive chemistry.Nature Machine Intelligence, 6(2): 161–169, 2024. doi: 10.1038/s42256-023-00788-1

    doi:10.1038/s42256-023-00788-1 2024

  5. [4]

    Ap- plications of natural language processing and large language models in materials discovery

    Xue Jiang, Weiren Wang, Shaohan Tian, Hao Wang, Turab Lookman, and Yanjing Su. Ap- plications of natural language processing and large language models in materials discovery. npj Computational Materials, 11(1):79, 2025. doi: 10.1038/s41524-025-01554-0

    doi:10.1038/s41524-025-01554-0 2025

  6. [5]

    Pyzer-Knapp, Matteo Manica, Peter Staar, Lucas Morin, Patrick Ruch, Teodoro Laino, et al

    Edward O. Pyzer-Knapp, Matteo Manica, Peter Staar, Lucas Morin, Patrick Ruch, Teodoro Laino, et al. Foundation models for materials discovery – current state and future directions. npj Computational Materials, 11(1):61, 2025. doi: 10.1038/s41524-025-01538-0

    doi:10.1038/s41524-025-01538-0 2025

  7. [7]

    Shyue Ping Ong, Shreyas Cholia, Anubhav Jain, Michael Brafman, Dan Gunter, Ger- brand Ceder, and Kristin A. Persson. The materials application programming interface (api): a simple, flexible and efficient api for materials data based on representational state transfer (rest) principles.Computational Materials Science, 97:209–215, 2015. doi: 10.1016/j.comm...

    doi:10.1016/j.commatsci.2014.10.037 2015

  8. [8]

    Schoenholz, Muratahan Aykol, Gowoon Cheon, and Ekin D

    Amil Merchant, Simon Batzner, Samuel S. Schoenholz, Muratahan Aykol, Gowoon Cheon, and Ekin D. Cubuk. Scaling deep learning for materials discovery.Nature, 624(7990):80–85,

    work page

Show all 45 references
  1. [9]

    Python materials genomics (pymatgen): a robust, open- source python library for materials analysis.Computational Materials Science, 68:314–319,

    Shyue Ping Ong, William Davidson Richards, Anubhav Jain, Geoffroy Hautier, Michael Kocher, Shreyas Cholia, et al. Python materials genomics (pymatgen): a robust, open- source python library for materials analysis.Computational Materials Science, 68:314–319,

  2. [10]

    Ganose, Hrushikesh Sahasrabuddhe, Mark Asta, Kevin Beck, Tathagata Biswas, Alexander Bonkowski, et al

    Alex M. Ganose, Hrushikesh Sahasrabuddhe, Mark Asta, Kevin Beck, Tathagata Biswas, Alexander Bonkowski, et al. Atomate2: modular workflows for materials science.Digital Discovery, 4(7):1944–1973, 2025. doi: 10.1039/D5DD00019J

    1944 doi

  3. [11]

    Bocarsly, Andres M

    Kevin Maik Jablonka, Qianxiang Ai, Alexander Al-Feghali, Shruti Badhwar, Joshua D. Bocarsly, Andres M. Bran, et al. 14 examples of how LLMs can transform materials science and chemistry: a reflection on a large language model hackathon.Digital Discovery, 2(5): 1233–1250, 2023....

    2023 doi

  4. [12]

    Collison, and Andrew D

    Mayk Caldas Ramos, Christopher J. Collison, and Andrew D. White. A review of large language models and autonomous agents in chemistry.Chemical Science, 16(6):2514–2572,

  5. [13]

    Hohenberg and W

    P. Hohenberg and W. Kohn. Inhomogeneous electron gas.Physical Review, 136(3B):B864– B871, 1964. doi: 10.1103/PhysRev.136.B864

    1964 doi

  6. [14]

    Kohn and L

    W. Kohn and L. J. Sham. Self-consistent equations including exchange and correlation effects.Physical Review, 140(4A):A1133–A1138, 1965. doi: 10.1103/PhysRev.140.A1133

    1965 doi

  7. [15]

    Kresse and J

    G. Kresse and J. Hafner. Ab initio molecular-dynamics simulation of the liquid-metal– amorphous-semiconductor transition in germanium.Physical Review B, 49(20):14251– 14269, 1994. doi: 10.1103/PhysRevB.49.14251

    1994 doi

  8. [16]

    Kresse and J

    G. Kresse and J. Furthmüller. Efficient iterative schemes for ab initio total-energy calcu- lations using a plane-wave basis set.Physical Review B, 54(16):11169–11186, 1996. doi: 10.1103/PhysRevB.54.11169

    1996 doi

  9. [17]

    Kresse and J

    G. Kresse and J. Furthmüller. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set.Computational Materials Science, 6(1): 15–50, 1996. doi: 10.1016/0927-0256(96)00008-0

    1996 doi

  10. [18]

    INCAR — VASP wiki.https://www.vasp.at/wiki/index.php/ INCAR, 2025

    VASP Software GmbH. INCAR — VASP wiki.https://www.vasp.at/wiki/index.php/ INCAR, 2025. Accessed 2026-06-28

    2025

  11. [20]

    P. E. Blöchl. Projector augmented-wave method.Physical Review B, 50(24):17953–17979,

  12. [21]

    Fromultrasoftpseudopotentialstotheprojectoraugmented-wave method.Physical Review B, 59(3):1758–1775, 1999

    G.KresseandD.Joubert. Fromultrasoftpseudopotentialstotheprojectoraugmented-wave method.Physical Review B, 59(3):1758–1775, 1999. doi: 10.1103/PhysRevB.59.1758

    1999 doi

  13. [22]

    Perdew, Kieron Burke, and Matthias Ernzerhof

    John P. Perdew, Kieron Burke, and Matthias Ernzerhof. Generalized gradient approx- imation made simple.Physical Review Letters, 77(18):3865–3868, 1996. doi: 10.1103/ PhysRevLett.77.3865

    1996

  14. [23]

    Mohd Zaki, Jayadeva, Mausam, and N. M. Anoop Krishnan. MaScQA: investigating ma- terials science knowledge of large language models.Digital Discovery, 3(2):313–327, 2024. doi: 10.1039/D3DD00188A

    2024 doi

  15. [24]

    Exploring the expertise of large language mod- els in materials science and metallurgical engineering.Digital Discovery, 4(2):500–512, 2025

    Christophe Bajan and Guillaume Lambard. Exploring the expertise of large language mod- els in materials science and metallurgical engineering.Digital Discovery, 4(2):500–512, 2025. doi: 10.1039/D4DD00319E

    2025 doi

  16. [25]

    Anker, Mehrdad Asgari, et al

    Joren Van Herck, María Victoria Gil, Kevin Maik Jablonka, Alex Abrudan, Andy S. Anker, Mehrdad Asgari, et al. Assessment of fine-tuned large language models for real-world chemistry and material science applications.Chemical Science, 16(2):670–684, 2025. doi: 10.1039/D4SC04401K

    2025 doi

  17. [26]

    Accelerating materials language processing with large lan- guage models.Communications Materials, 5(1):13, 2024

    Jaewoong Choi and Byungju Lee. Accelerating materials language processing with large lan- guage models.Communications Materials, 5(1):13, 2024. doi: 10.1038/s43246-024-00449-9

    2024 doi

  18. [27]

    S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, and A. P. Sutton. Electron- energy-loss spectra and the structural stability of nickel oxide: An LSDA+U study.Physical Review B, 57(3):1505–1509, 1998. doi: 10.1103/PhysRevB.57.1505

    1998 doi

  19. [28]

    Stefan Grimme, Jens Antony, Stephan Ehrlich, and Helge Krieg. A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu.The Journal of Chemical Physics, 132(15):154104, 2010. doi: 10.1063/1. 3382344

    2010 doi

  20. [29]

    Bowler, and Angelos Michaelides

    Jiří Klimeš, David R. Bowler, and Angelos Michaelides. Van der Waals density functionals applied to solids.Physical Review B, 83(19):195131, 2011. doi: 10.1103/PhysRevB.83. 195131

    2011 doi

  21. [30]

    Monkhorst and James D

    Hendrik J. Monkhorst and James D. Pack. Special points for Brillouin-zone integrations. Physical Review B, 13(12):5188–5192, 1976. doi: 10.1103/PhysRevB.13.5188

    1976 doi

  22. [31]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, et al

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, et al. LoRA: Low-rank adaptation of large language models, 2021

    2021

  23. [32]

    Qwen3 technical report, 2025

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, et al. Qwen3 technical report, 2025

    2025

  24. [33]

    The Llama 3 herd of models, 2024

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, et al. The Llama 3 herd of models, 2024

    2024

  25. [34]

    Gemma 3 technical report, 2025

    Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, et al. Gemma 3 technical report, 2025

    2025

  26. [35]

    Montoya, Alireza Faghaninia, Shyam Dwarakanath, Muratahan Aykol, Hanmei Tang, et al

    Kiran Mathew, Joseph H. Montoya, Alireza Faghaninia, Shyam Dwarakanath, Muratahan Aykol, Hanmei Tang, et al. Atomate: a high-level interface to generate, execute, and analyze computational materials science workflows.Computational Materials Science, 139:140–152,

  27. [36]

    Wataru Takahara, Yuichi Yamaguchi, Mai Ogano, Fuga Kakami, Yosuke Harashima, To- moaki Takayama, et al. Materials dual-source knowledge retrieval-augmented generation for local large language models in photocatalysts.Journal of Chemical Information and Modeling, 65(24):13098–1...

    2025 doi

  28. [37]

    Giannozzi, O

    P. Giannozzi, O. Andreussi, T. Brumme, O. Bunau, M. Buongiorno Nardelli, M. Calandra, et al. Advanced capabilities for materials modelling with Quantum ESPRESSO.Journal of Physics: Condensed Matter, 29(46):465901, 2017. doi: 10.1088/1361-648X/aa8f79

    2017 doi

  29. [38]

    Gonze, B

    X. Gonze, B. Amadon, P.-M. Anglade, J.-M. Beuken, F. Bottin, P. Boulanger, et al. ABINIT: first-principles approach to material and nanosystem properties.Computer Physics Communications, 180(12):2582–2615, 2009. doi: 10.1016/j.cpc.2009.07.007

    2009 doi

  30. [39]

    Thompson, H

    Aidan P. Thompson, H. Metin Aktulga, Richard Berger, Dan S. Bolintineanu, W. Michael Brown, Paul S. Crozier, et al. LAMMPS — a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales.Computer Physics Commu- nications, 271:108171...

    2022

  31. [40]

    H. G. Weller, G. Tabor, H. Jasak, and C. A. Fureby. A tensorial approach to computational continuum mechanics using object-oriented techniques.Computers in Physics, 12(6):620– 631, 1998. doi: 10.1063/1.168744. 14 Supplementary Information S1 Scope This document provides the in...

    1998 doi

  32. [1994]

    doi: 10.1103/PhysRevB.50.17953. 12

    doi

  33. [2013]

    doi: 10.1016/j.commatsci.2012.10.028

    2012 doi

  34. [2017]

    doi: 10.1016/j.commatsci.2017.07.030. 13

    2017 doi

  35. [2023]

    doi: 10.1038/s41586-023-06735-9

    doi

  36. [2025]

    doi: 10.1039/D4SC03921A

    doi

  37. [2026]

    Benchmark data athttps://github.com/liuzhiyangnku/incarbench

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.