pith. sign in

arxiv: 2502.02189 · v4 · submitted 2025-02-04 · 💻 cs.LG

deCIFer: Crystal Structure Prediction from Powder Diffraction Data using Autoregressive Language Models

Frederik Lizak Johansen , Ulrik Friis-Jensen , Erik Bj{\o}rnager Dam , Kirsten Marie {\O}rnsbjerg Jensen , Roc\'io Mercado , Raghavendra Selvan This is my paper

Pith reviewed 2026-05-23 03:48 UTC · model grok-4.3

classification 💻 cs.LG
keywords crystal structure predictionpowder X-ray diffractionautoregressive language modelsCIF formatPXRD-CSPinorganic materials
0
0 comments X

The pith

deCIFer autoregressive model generates crystal structures in CIF format from PXRD data at 94 percent match rate.

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

The paper presents deCIFer, an autoregressive language model trained to predict crystal structures conditioned directly on powder X-ray diffraction patterns. It generates output in standard CIF format after exposure to nearly 2.3 million structures whose diffraction signals were modified by added Gaussian noise and instrumental peak broadening. Validation on multiple synthetic datasets that represent difficult inorganic materials yields a 94 percent structural match rate, evaluated with the residual weighted profile R_wp and the match rate metric. The approach differs from earlier crystal structure prediction methods by making the diffraction data itself the primary conditioning input rather than relying mainly on composition or symmetry rules. This establishes a baseline for moving computational predictions closer to direct use of experimental diffraction measurements.

Core claim

deCIFer is an autoregressive language model for PXRD-conditioned crystal structure prediction that directly outputs crystal structures in CIF format; after training on nearly 2.3 million structures with PXRD patterns augmented by Gaussian noise and peak broadening, the model reaches a 94 percent structural match rate on diverse synthetic datasets of challenging inorganic materials when assessed by R_wp and match-rate metrics.

What carries the argument

Autoregressive language model that generates CIF token sequences conditioned on PXRD patterns augmented with Gaussian noise and instrumental peak broadening.

If this is right

  • Provides an alternative to composition- or symmetry-driven CSP methods by conditioning directly on diffraction data.
  • Produces structures in the standard CIF format that can be used immediately in downstream materials workflows.
  • Achieves its reported match rate using metrics chosen for practical relevance in the underdetermined PXRD problem.
  • Sets an explicit baseline that future work can extend to more complex experimental conditions.

Where Pith is reading between the lines

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

  • Testing the trained model on actual laboratory PXRD measurements would show whether the limited synthetic augmentations suffice for real-world generalization.
  • Extending the conditioning to include additional experimental effects such as preferred orientation or sample displacement could improve robustness without changing the core architecture.
  • Pairing deCIFer outputs with rapid Rietveld refinement might reduce the number of candidate structures that need manual inspection.

Load-bearing premise

Augmenting PXRD conditioning with only Gaussian noise and instrumental peak broadening produces training signals representative enough for the model to generalize to real experimental powder diffraction data.

What would settle it

Apply deCIFer to a collection of real experimental PXRD patterns from structures whose ground-truth CIFs are already known and measure whether the structural match rate remains near 94 percent.

Figures

Figures reproduced from arXiv: 2502.02189 by Erik Bj{\o}rnager Dam, Frederik Lizak Johansen, Kirsten Marie {\O}rnsbjerg Jensen, Raghavendra Selvan, Roc\'io Mercado, Ulrik Friis-Jensen.

Figure 1
Figure 1. Figure 1: (a) Overview of the deCIFer model, which performs autoregressive crystal structure prediction [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Evaluation pipeline: A test set CIF gener [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Left: Distribution of Rwp for deCIFer and U-deCIFer on the NOMA test set with boxplots. Lower Rwp indicates better CIF alignment. Right: Performance for 20K NOMA test samples using deCIFer and U-deCIFer with different descriptors: none (no descriptors), comp. (composition), and comp.+ s.g. (composition + space group). Metrics include validity (Val.) and match rate (MR). positions (Å). To ensure a fair and … view at source ↗
Figure 4
Figure 4. Figure 4: Left: Average Rwp by crystal system for deCIFer on the NOMA test set shows better performance for common high symmetry systems and higher Rwp for rare low symmetry systems. Right: Examples from the NOMA test set highlight this trend with predicted structures from PXRD and composition maintaining reasonable matches even for low symmetry systems with higher Rwp. Robustness to Perturbations in PXRD Conditioni… view at source ↗
Figure 5
Figure 5. Figure 5: Left: Distribution of Rwp for deCIFer on NOMA and CHILI-100K using PXRD conditioning and composition (comp.) across three scenarios: clean, high noise/broadening, and out-of-distribution (OOD), with noise and FWHM values indicated. Right: Corresponding table of performance metrics, including Rwp, overall validity (Val.), and match rate (MR) for each scenario. consistently achieves strong performance on the… view at source ↗
Figure 6
Figure 6. Figure 6: deCIFer-sampled structures for a monoclinic Sr [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Illustration of a CIF after applying the pre-processing and standardization steps described. [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Visualization of the attention masking strategy, showing the log-mean attention weights (averaged [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Simulated PXRD profiles with fixed transformation of FWHM and [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Statistical overview of the NOMA (Antunes et al., 2024) dataset (2,283,346 total samples), showing the distribution of space group frequencies, the number of elements per unit cell, elemental occurrences and CIF token lengths (indicating the percentage of CIFs with larger token sequences than the context length of 3076) • ltol = 0.3: fractional length tolerance, meaning the lattice parameters can differ b… view at source ↗
Figure 11
Figure 11. Figure 11: Statistical overview of the curated CHILI-100K ( [PITH_FULL_IMAGE:figures/full_fig_p019_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Cross-entropy loss curves for U-deCIFer and deCIFer over 50,000 training iterations, showing [PITH_FULL_IMAGE:figures/full_fig_p020_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: 2D PCA projection of learned PXRD embeddings for 500K training-set samples from NOMA. [PITH_FULL_IMAGE:figures/full_fig_p020_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Average metric values by crystal systems for deCIFer on the NOMA test set under two in [PITH_FULL_IMAGE:figures/full_fig_p021_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Average metric values by crystal systems for deCIFer on the CHILI-100K test set show better [PITH_FULL_IMAGE:figures/full_fig_p022_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Distribution of Rwp for deCIFer on the NOMA- and CHILI-100K test set, presented as violin plots with overlain boxplots; the median is shown for each distribution. Presented are four in-distribution transformations of the input PXRD profiles and three out-of-distribution transformations. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_16.png] view at source ↗
read the original abstract

Novel materials drive advancements in fields ranging from energy storage to electronics, with crystal structure characterization forming a crucial yet challenging step in materials discovery. In this work, we introduce \emph{deCIFer}, an autoregressive language model designed for powder X-ray diffraction (PXRD)-conditioned crystal structure prediction (PXRD-CSP). Unlike traditional CSP methods that rely primarily on composition or symmetry constraints, deCIFer explicitly incorporates PXRD data, directly generating crystal structures in the widely adopted Crystallographic Information File (CIF) format. The model is trained on nearly 2.3 million crystal structures, with PXRD conditioning augmented by basic forms of synthetic experimental artifacts, specifically Gaussian noise and instrumental peak broadening, to reflect fundamental real-world conditions. Validated across diverse synthetic datasets representative of challenging inorganic materials, deCIFer achieves a 94\% structural match rate. The evaluation is based on metrics such as the residual weighted profile ($R_{wp}$) and structural match rate (MR), chosen explicitly for their practical relevance in this inherently underdetermined problem. deCIFer establishes a robust baseline for future expansion toward more complex experimental scenarios, bridging the gap between computational predictions and experimental crystal structure determination.

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 / 0 minor

Summary. The paper introduces deCIFer, an autoregressive language model for PXRD-conditioned crystal structure prediction that directly generates CIF files. Trained on ~2.3 million structures with PXRD patterns augmented by Gaussian noise and peak broadening, it reports a 94% structural match rate (MR) on diverse synthetic test sets of inorganic materials, using R_wp and MR as primary metrics. The work positions the model as a baseline for future experimental PXRD scenarios.

Significance. A working autoregressive CIF generator conditioned on PXRD would address an important inverse problem in materials science. The training scale and choice of practically relevant metrics (R_wp, MR) are positive features. However, because all reported results use only synthetic data with limited augmentations, the practical significance for real experimental data remains unproven.

major comments (2)
  1. [Abstract] Abstract: the central performance claim of a 94% structural match rate is shown exclusively on synthetic PXRD patterns that incorporate only Gaussian noise and instrumental peak broadening. No experiments on measured laboratory diffractograms are reported, which directly affects the claim that the method 'bridges the gap between computational predictions and experimental crystal structure determination.'
  2. [Abstract] Abstract (training description paragraph): the assumption that the chosen augmentations suffice to simulate the underdetermined inverse problem of real PXRD is load-bearing for generalization claims, yet the manuscript provides no ablation or sensitivity analysis on additional common experimental effects such as preferred orientation, sample displacement, or impurity phases.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below, with revisions where appropriate to better reflect the scope of the work as a synthetic-data baseline.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central performance claim of a 94% structural match rate is shown exclusively on synthetic PXRD patterns that incorporate only Gaussian noise and instrumental peak broadening. No experiments on measured laboratory diffractograms are reported, which directly affects the claim that the method 'bridges the gap between computational predictions and experimental crystal structure determination.'

    Authors: We agree that all quantitative results, including the 94% structural match rate, are obtained exclusively on synthetic PXRD patterns with Gaussian noise and peak broadening. The manuscript already describes deCIFer as establishing a baseline for future expansion toward experimental scenarios. To address the concern, we will revise the abstract to explicitly state that current performance is demonstrated on synthetic data and that bridging to experimental crystal structure determination remains a direction for future work rather than a completed achievement. revision: yes

  2. Referee: [Abstract] Abstract (training description paragraph): the assumption that the chosen augmentations suffice to simulate the underdetermined inverse problem of real PXRD is load-bearing for generalization claims, yet the manuscript provides no ablation or sensitivity analysis on additional common experimental effects such as preferred orientation, sample displacement, or impurity phases.

    Authors: The chosen augmentations capture core aspects of experimental PXRD (noise and broadening) to create a practical baseline. We acknowledge that no ablation studies on additional effects such as preferred orientation, sample displacement, or impurity phases are included. As the work is framed as an initial baseline rather than a comprehensive simulation of all experimental artifacts, we do not view exhaustive ablations as necessary for the present contribution. We will add a short clarifying statement in the abstract and discussion sections noting these unmodeled effects as important avenues for future research. revision: partial

standing simulated objections not resolved
  • Results on measured laboratory diffractograms cannot be provided, as the current study contains only synthetic data experiments.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper trains an autoregressive LM on 2.3M crystal structures with synthetic PXRD augmentations (Gaussian noise + peak broadening) and reports an empirical 94% structural match rate on held-out synthetic validation sets using standard metrics (R_wp, MR). No derivation chain, equation, or claim reduces the reported performance to a fitted input, self-citation, or definition by construction. The result is an independent empirical evaluation on generated vs. ground-truth structures and remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard machine-learning assumptions plus a domain assumption about synthetic data sufficiency; no new physical entities are introduced.

free parameters (2)
  • autoregressive LM hyperparameters
    Large number of training hyperparameters (learning rate, architecture size, tokenization choices) that are fitted or chosen during model development.
  • synthetic artifact parameters
    Gaussian noise variance and peak-broadening widths chosen to augment the conditioning data.
axioms (1)
  • domain assumption The collection of 2.3 million crystal structures is representative of the distribution of inorganic materials of interest.
    Invoked when stating that the model was trained on this corpus and validated on diverse synthetic datasets.

pith-pipeline@v0.9.0 · 5781 in / 1336 out tokens · 70988 ms · 2026-05-23T03:48:38.045497+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

56 extracted references · 56 canonical work pages · 4 internal anchors

  1. [1]

    L. M. Antunes, K. T. Butler, and R. Grau-Crespo. Crystal structure generation with autoregressive large language modeling. Nature Communications, 15 0 (1): 0 10570, 2024. ISSN 2041-1723. doi:10.1038/s41467-024-54639-7

    work page doi:10.1038/s41467-024-54639-7 2024

  2. [2]

    S. Baird. mp-time-split. accessed in 2024. https://github.com/sparks-baird/mp-time-split, 2023

    work page 2024

  3. [3]

    J. K. Bunn, J. Hu, and J. R. Hattrick-Simpers. Semi-supervised approach to phase identification from combinatorial sample diffraction patterns. JOM, 68 0 (8): 0 2116--2125, Aug 2016. ISSN 1543-1851. doi:10.1007/s11837-016-2033-8

    work page doi:10.1007/s11837-016-2033-8 2016

  4. [4]

    I. E. Castelli, D. D. Landis, K. S. Thygesen, S. Dahl, I. Chorkendorff, T. F. Jaramillo, and K. W. Jacobsen. New cubic perovskites for one- and two-photon water splitting using the computational materials repository. Energy Environ. Sci., 5: 0 9034--9043, 2012 a . doi:10.1039/C2EE22341D

    work page doi:10.1039/c2ee22341d 2012

  5. [5]

    I. E. Castelli, T. Olsen, S. Datta, D. D. Landis, S. Dahl, K. S. Thygesen, and K. W. Jacobsen. Computational screening of perovskite metal oxides for optimal solar light capture. Energy Environ. Sci., 5: 0 5814--5819, 2012 b . doi:10.1039/C1EE02717D

    work page doi:10.1039/c1ee02717d 2012

  6. [6]

    A. K. Cheetham and A. L. Goodwin. Crystallography with powders. Nature Materials, 13 0 (8): 0 760--762, Aug 2014

    work page 2014

  7. [7]

    A. K. Cheetham and R. Seshadri. Artificial intelligence driving materials discovery? perspective on the article: Scaling deep learning for materials discovery. Chemistry of Materials, 36 0 (8): 0 3490--3495, 2024. doi:10.1021/acs.chemmater.4c00643

    work page doi:10.1021/acs.chemmater.4c00643 2024

  8. [8]

    Dagdelen, A

    J. Dagdelen, A. Dunn, S. Lee, N. Walker, A. S. Rosen, G. Ceder, K. A. Persson, and A. Jain. Structured information extraction from scientific text with large language models. Nature Communications, 15 0 (1): 0 1418, 2024

    work page 2024

  9. [9]

    Davariashtiyani, B

    A. Davariashtiyani, B. Wang, S. Hajinazar, E. Zurek, and S. Kadkhodaei. Impact of data bias on machine learning for crystal compound synthesizability predictions. Machine Learning: Science and Technology, 5 0 (4): 0 040501, nov 2024. doi:10.1088/2632-2153/ad9378

    work page doi:10.1088/2632-2153/ad9378 2024

  10. [10]

    Dettmers, A

    T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer. Qlora: Efficient finetuning of quantized llms. Advances in Neural Information Processing Systems, 36, 2024

    work page 2024

  11. [11]

    Draxl and M

    C. Draxl and M. Scheffler. The nomad laboratory: From data sharing to artificial intelligence. Journal of Physics: Materials, 2, 05 2019. doi:10.1088/2515-7639/ab13bb

    work page doi:10.1088/2515-7639/ab13bb 2019

  12. [12]

    The Llama 3 Herd of Models

    A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, A. Mathur, A. Schelten, A. Yang, A. Fan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  13. [13]

    Friis-Jensen, F

    U. Friis-Jensen, F. L. Johansen, A. S. Anker, E. B. Dam, K. M. O. Jensen, and R. Selvan. Chili: Chemically-informed large-scale inorganic nanomaterials dataset for advancing graph machine learning. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD '24, page 4962–4973, New York, NY, USA, 2024. Association for Comp...

    work page doi:10.1145/3637528.3671538 2024

  14. [14]

    Gazzarrini, R

    E. Gazzarrini, R. K. Cersonsky, M. Bercx, C. S. Adorf, and N. Marzari. The rule of four: anomalous distributions in the stoichiometries of inorganic compounds. npj Computational Materials, 10 0 (1): 0 73, Apr 2024. ISSN 2057-3960. doi:10.1038/s41524-024-01248-z

    work page doi:10.1038/s41524-024-01248-z 2024

  15. [15]

    Gra z ulis, D

    S. Gra z ulis, D. Chateigner, R. T. Downs, A. F. T. Yokochi, M. Quir \' o s, L. Lutterotti, E. Manakova, J. Butkus, P. Moeck, and A. Le Bail. Crystallography Open Database -- an open-access collection of crystal structures . Journal of Applied Crystallography, 42 0 (4): 0 726--729, Aug 2009. doi:10.1107/S0021889809016690

    work page doi:10.1107/s0021889809016690 2009

  16. [16]

    Gruver, A

    N. Gruver, A. Sriram, A. Madotto, A. G. Wilson, C. L. Zitnick, and Z. W. Ulissi. Fine-tuned language models generate stable inorganic materials as text. In The Twelfth International Conference on Learning Representations (ICLR), 2024

    work page 2024

  17. [17]

    G. Guo, T. Saidi, M. Terban, M. Valsecchi, S. J. Billinge, and H. Lipson. Ab initio structure solutions from nanocrystalline powder diffraction data, 2024

    work page 2024

  18. [18]

    Gupta, M

    T. Gupta, M. Zaki, N. A. Krishnan, and Mausam. Matscibert: A materials domain language model for text mining and information extraction. npj Computational Materials, 8 0 (1): 0 102, 2022

    work page 2022

  19. [19]

    G. M. Hocky and A. D. White. Natural language processing models that automate programming will transform chemistry research and teaching. Digital discovery, 1 0 (2): 0 79--83, 2022

    work page 2022

  20. [20]

    K. M. Jablonka, P. Schwaller, A. Ortega-Guerrero, and B. Smit. Leveraging large language models for predictive chemistry. Nature Machine Intelligence, 6 0 (2): 0 161--169, 2024

    work page 2024

  21. [22]

    A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, and K. A. Persson. Commentary: The materials project: A materials genome approach to accelerating materials innovation. APL Materials, 1 0 (1): 0 011002, 07 2013 b . ISSN 2166-532X. doi:10.1063/1.4812323

    work page doi:10.1063/1.4812323 2013

  22. [23]

    R. Jiao, W. Huang, P. Lin, J. Han, P. Chen, Y. Lu, and Y. Liu. Crystal structure prediction by joint equivariant diffusion. Advances in Neural Information Processing Systems, 36: 0 17464--17497, 2023

    work page 2023

  23. [24]

    Kirklin, J

    S. Kirklin, J. E. Saal, B. Meredig, A. Thompson, J. W. Doak, M. Aykol, S. R \"u hl, and C. Wolverton. The open quantum materials database (oqmd): assessing the accuracy of dft formation energies. npj Computational Materials, 1 0 (1): 0 1--15, 2015

    work page 2015

  24. [25]

    E. T. S. Kjær, A. S. Anker, M. N. Weng, S. J. L. Billinge, R. Selvan, and K. M. . Jensen. Deepstruc: towards structure solution from pair distribution function data using deep generative models. Digital Discovery, 2: 0 69--80, 2023. doi:10.1039/D2DD00086E

    work page doi:10.1039/d2dd00086e 2023

  25. [26]

    Kosec, S

    M. Kosec, S. Fu, and M. M. Krell. Packing: Towards 2x NLP BERT acceleration. CoRR, abs/2107.02027, 2021

    work page arXiv 2021

  26. [27]

    Q. Lai, F. Xu, L. Yao, Z. Gao, S. Liu, H. Wang, S. Lu, D. He, L. Wang, L. Zhang, C. Wang, and G. Ke. End-to-end crystal structure prediction from powder x-ray diffraction. Advanced Science, page 2410722, 2025

    work page 2025

  27. [28]

    Decoupled Weight Decay Regularization

    I. Loshchilov and F. Hutter. Fixing weight decay regularization in adam. CoRR, abs/1711.05101, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  28. [29]

    A. M. Bran, S. Cox, O. Schilter, C. Baldassari, A. D. White, and P. Schwaller. Augmenting large language models with chemistry tools. Nature Machine Intelligence, pages 1--11, 2024

    work page 2024

  29. [30]

    B. K. Miller, R. T. Chen, A. Sriram, and B. M. Wood. Flowmm: Generating materials with riemannian flow matching. In Forty-first International Conference on Machine Learning, 2024

    work page 2024

  30. [31]

    Mohanty, M

    T. Mohanty, M. Mehta, H. M. Sayeed, V. Srikumar, and T. D. Sparks. Crystext: A generative ai approach for text-conditioned crystal structure generation using llm. ChemRxiv, 2024. doi:10.26434/chemrxiv-2024-gjhpq. This content is a preprint and has not been peer-reviewed

    work page doi:10.26434/chemrxiv-2024-gjhpq 2024

  31. [32]

    Momma and F

    K. Momma and F. Izumi. VESTA : a three-dimensional visualization system for electronic and structural analysis . Journal of Applied Crystallography, 41 0 (3): 0 653--658, Jun 2008. doi:10.1107/S0021889808012016

    work page doi:10.1107/s0021889808012016 2008

  32. [33]

    S. P. Ong, W. D. Richards, A. Jain, G. Hautier, M. Kocher, S. Cholia, D. Gunter, V. Chevrier, K. A. Persson, and G. Ceder. Python materials genomics (pymatgen): A robust, open-source python library for materials analysis. Computational Materials Science, 68: 0 314--319, 2013. doi:10.1016/j.commatsci.2012.10.028

    work page doi:10.1016/j.commatsci.2012.10.028 2013

  33. [34]

    Oviedo, Z

    F. Oviedo, Z. Ren, S. Sun, C. Settens, Z. Liu, N. T. P. Hartono, S. Ramasamy, B. L. DeCost, S. I. P. Tian, G. Romano, A. Gilad Kusne, and T. Buonassisi. Fast and interpretable classification of small x-ray diffraction datasets using data augmentation and deep neural networks. npj Computational Materials, 5 0 (1): 0 60, May 2019. ISSN 2057-3960. doi:10.103...

    work page doi:10.1038/s41524-019-0196-x 2019

  34. [35]

    Palgrave

    R. Palgrave. An explanation for the rule of four in inorganic materials, 2024. Preprint, not peer-reviewed

    work page 2024

  35. [36]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. Kopf, E. Yang, Z. DeVito, M. Raison, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala. Pytorch: An imperative style, high-performance deep learning library. In H. Wallach, H. Larochelle, A. Beygelzimer, F....

    work page 2019

  36. [37]

    A. L. Patterson. Ambiguities in the x-ray analysis of crystal structures. Phys. Rev., 65: 0 195--201, Mar 1944. doi:10.1103/PhysRev.65.195

    work page doi:10.1103/physrev.65.195 1944

  37. [38]

    C. J. Pickard. AIRSS Data for Carbon at 10GPa and the C+N+H+O System at 1GPa . https://archive.materialscloud.org/record/2020.0026/v1, 2020

    work page arXiv 2020

  38. [39]

    M. P. Polak and D. Morgan. Extracting accurate materials data from research papers with conversational language models and prompt engineering. Nature Communications, 15 0 (1): 0 1569, 2024

    work page 2024

  39. [40]

    E. A. Riesel, T. Mackey, H. Nilforoshan, M. Xu, C. K. Badding, A. B. Altman, J. Leskovec, and D. E. Freedman. Crystal structure determination from powder diffraction patterns with generative machine learning. Journal of the American Chemical Society, 146 0 (44): 0 30340--30348, 2024. doi:10.1021/jacs.4c10244. PMID: 39298266

    work page doi:10.1021/jacs.4c10244 2024

  40. [41]

    A. N. Rubungo, K. Li, J. Hattrick-Simpers, and A. B. Dieng. LLM 4mat-bench: Benchmarking large language models for materials property prediction. In AI for Accelerated Materials Design - NeurIPS 2024, 2024

    work page 2024

  41. [42]

    Schilling-Wilhelmi, M

    M. Schilling-Wilhelmi, M. R \' os-Garc \' a, S. Shabih, M. V. Gil, S. Miret, C. T. Koch, J. A. M \'a rquez, and K. M. Jablonka. From text to insight: large language models for chemical data extraction. Chemical Society Reviews, 2025

    work page 2025

  42. [43]

    M. N. Schneider, M. Seibald, P. Lagally, and O. Oeckler. Ambiguities in the structure determination of antimony tellurides arising from almost homometric structure models and stacking disorder . Journal of Applied Crystallography, 43 0 (5 Part 1): 0 1012--1020, Oct 2010. doi:10.1107/S0021889810032644

    work page doi:10.1107/s0021889810032644 2010

  43. [44]

    Y. Shen, Y. Jiang, J. Lin, C. Wang, and J. Sun. A general method for searching for homometric structures . Acta Crystallographica Section B, 78 0 (1): 0 14--19, Feb 2022. doi:10.1107/S2052520621011859

    work page doi:10.1107/s2052520621011859 2022

  44. [45]

    N. J. Szymanski, B. Rendy, Y. Fei, R. E. Kumar, T. He, D. Milsted, M. J. McDermott, M. Gallant, E. D. Cubuk, A. Merchant, et al. An autonomous laboratory for the accelerated synthesis of novel materials. Nature, 624 0 (7990): 0 86--91, 2023

    work page 2023

  45. [46]

    M. Tatlier. Artificial neural network methods for the prediction of framework crystal structures of zeolites from xrd data. Neural Computing and Applications, 20 0 (3): 0 365--371, Apr 2011. ISSN 1433-3058. doi:10.1007/s00521-010-0386-4

    work page doi:10.1007/s00521-010-0386-4 2011

  46. [47]

    Togo and I

    A. Togo and I. Tanaka. Spglib : a software library for crystal symmetry search , 2018

    work page 2018

  47. [48]

    Llama 2: Open Foundation and Fine-Tuned Chat Models

    H. Touvron, L. Martin, K. Stone, P. Albert, A. Almahairi, Y. Babaei, N. Bashlykov, S. Batra, P. Bhargava, S. Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  48. [49]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. u. Kaiser, and I. Polosukhin. Attention is all you need. In I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc., 2017

    work page 2017

  49. [50]

    H. Wang, Y. Xie, D. Li, H. Deng, Y. Zhao, M. Xin, and J. Lin. Rapid identification of x-ray diffraction patterns based on very limited data by interpretable convolutional neural networks. Journal of Chemical Information and Modeling, 60 0 (4): 0 2004--2011, Apr 2020. ISSN 1549-9596. doi:10.1021/acs.jcim.0c00020

    work page doi:10.1021/acs.jcim.0c00020 2004

  50. [51]

    A. R. West. Solid State Chemistry and its Applications. Wiley, 2nd edition, 2014

    work page 2014

  51. [52]

    T. Xie, X. Fu, O.-E. Ganea, R. Barzilay, and T. Jaakkola. Crystal diffusion variational autoencoder for periodic material generation, 2022

    work page 2022

  52. [53]

    R. Young. The Rietveld Method. IUCr monographs on crystallography. Oxford University Press, 1995. ISBN 9780198559122

    work page 1995

  53. [54]

    C. Zeni, R. Pinsler, D. Z \"u gner, A. Fowler, M. Horton, X. Fu, Z. Wang, A. Shysheya, J. Crabb \'e , S. Ueda, et al. A generative model for inorganic materials design. Nature, pages 1--3, 2025

    work page 2025

  54. [55]

    Zhang, X

    D. Zhang, X. Liu, X. Zhang, C. Zhang, C. Cai, H. Bi, Y. Du, X. Qin, A. Peng, J. Huang, B. Li, Y. Shan, J. Zeng, Y. Zhang, S. Liu, Y. Li, J. Chang, X. Wang, S. Zhou, J. Liu, X. Luo, Z. Wang, W. Jiang, J. Wu, Y. Yang, J. Yang, M. Yang, F.-Q. Gong, L. Zhang, M. Shi, F.-Z. Dai, D. M. York, S. Liu, T. Zhu, Z. Zhong, J. Lv, J. Cheng, W. Jia, M. Chen, G. Ke, W. ...

    work page doi:10.1038/s41524-024-01493-2 2024

  55. [56]

    Zhang, W

    H. Zhang, W. W. Chen, J. M. Rondinelli, and W. Chen. Et-al: Entropy-targeted active learning for bias mitigation in materials data. Applied Physics Reviews, 10 0 (2): 0 021403, 04 2023. ISSN 1931-9401. doi:10.1063/5.0138913

    work page doi:10.1063/5.0138913 2023

  56. [57]

    D. M. Ziegler, N. Stiennon, J. Wu, T. B. Brown, A. Radford, D. Amodei, P. Christiano, and G. Irving. Fine-tuning language models from human preferences. arXiv preprint arXiv:1909.08593, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1909