Pith. sign in

REVIEW 5 major objections 6 minor 39 references

Toward Greater Autonomy in Materials Discovery Agents: Unifying Planning, Physics, and Scientists

T0 review · 5 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 MAPPS is a multi-agent system that plans its own materials-discovery workflows from high-level goals and human intuition, and the paper claims it raises the rate of stable, unique, novel crystals on MP-20 roughly five-fold over prior…

desk verdict A well-structured Level-2 agent framework undercut by an uncontrolled retrieval advantage; the five-fold improvement claim does not survive scrutiny, but the autonomy taxonomy and the Level-3 failure study are worth engaging. read the letter →

arxiv 2506.05616 v2 pith:SJJ5ZLE3 submitted 2025-06-05 cs.AI cond-mat.mtrl-sciphysics.comp-ph

Lianhao Zhou , Hongyi Ling , Keqiang Yan , Kaiji Zhao , Xiaoning Qian , Raymundo Arróyave , Xiaofeng Qian , Shuiwang Ji This is my paper
classification cs.AIcond-mat.mtrl-sciphysics.comp-ph
keywords largelanguagemodelagentsmaterialsdiscoverycrystalstructuregenerationpredictionproperty-guidedmachinelearningforcefieldsworkflowplanninghuman-in-the-loop
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 tries to show that large language models can be more than step-executing tools in materials discovery: given a high-level goal and human scientific intuition, an LLM-based agent can plan a multi-step discovery workflow, write the code to carry it out, incorporate physics via machine-learning force fields, and revise its own errors. If this holds, the practical consequence is that generative materials AI need not be retrained for each new task; a planning agent can steer existing tools toward stable and novel crystals, predicting structures for given compositions, and designing materials with target bandgaps. The headline evidence is a five-fold improvement in the rate of stable, unique, novel crystals generated on MP-20 compared with prior generative models, with DFT-confirmed stability at 34.3%.

What carries the argument

The load-bearing component is the three-agent control loop: a Workflow Planner that turns a high-level task and a scientist's intuition into a five-step plan; a Tool Code Generator that writes executable Python for each step, drawing on a physics toolbox of ML force fields (e.g. CHGNet, M3GNet), pymatgen, and space-group analysis; and a Scientific Mediator that carries intermediate results between steps, requests human approval at decision points, and triggers error-based code revision. Together they implement what the paper calls Level 2 autonomy: the agent designs its own workflow, but a human still shapes it and checks results.

What would settle it

Run the strongest baseline (e.g., FlowLLM) with the same retrieval database, ML-force-field relaxation, and novelty filtering that MAPPS uses; if its S.U.N. rate approaches 24.9%, the claimed five-fold improvement is mostly retrieval and relaxation, not planning. A direct check would also count how many of MAPPS's 'novel' structures are relaxed copies of retrieved database entries.

Watch

Extended reading notes

Core claim

The paper argues that a language-agent system can move from executing fixed, human-designed steps (Level 1) to planning its own multi-step scientific workflows under human guidance (Level 2), and that this added autonomy produces markedly better materials. On MP-20, MAPPS attains a DFT stability rate of 34.3% and a stability-uniqueness-novelty (S.U.N.) rate of 24.9%, compared with 17.8% and 4.92% for FlowLLM; on crystal structure prediction it reaches 63.9% match rate on MP-20, 27.6% on MPTS-52, and 31.0% on a recent-discovery challenge set, without training any new generative model. The system's workflows start from a task description plus human intuition, are approved by a scientist, and are executed as generated Python code that calls machine-learning force fields, space-group analyzers, and structure-matching tools, with self-reflection to recover from errors.

Load-bearing premise

The comparison assumes MAPPS's edge comes from its planning agents rather than from being allowed to retrieve from the full MP-20 dataset while baselines only train on it; no retrieval-augmented baseline is tested.

Editorial extensions

If this is right

  • If MAPPS's numbers hold, generative materials models can be outperformed by an agent that does not train new weights but retrieves candidate prototypes, relaxes them with ML force fields, and filters by symmetry and stability.
  • The same framework transfers across three task families—unconditional generation, composition-to-structure prediction, and bandgap-constrained design—suggesting one agent architecture can replace task-specific pipelines.
  • Because the property-guided setting reaches 74.6% (high bandgap) and 92.2% (low bandgap) DFT satisfaction, agentic planning is a viable route to inverse design of electronic properties.
  • The paper's Level-3 study indicates current reasoning LLMs cannot yet produce valid workflows without human intuition, so human-in-the-loop remains necessary for reliability.

Reading between the lines

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

  • The S.U.N. comparison is not apples-to-apples: MAPPS retrieves candidates from the full MP-20 database and relaxes them, while baselines must generate from training data, so part of the five-fold gain is attributable to retrieval plus physics relaxation rather than to planning. A retrieval-augmented baseline would settle how much.
  • A natural extension is to measure the contribution of each component separately—planner, self-reflection, and human feedback—by ablating them, and to test whether the same agent transfers to molecules or polymers once a suitable physics toolbox exists.
  • The paper's own Table 5 suggests that the human intuition input, not the language model, is the main source of workflow validity; this makes 'autonomy' largely conditional on expert guidance.
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

5 major / 6 minor

Summary. The paper introduces MAPPS, a multi-agent LLM framework for materials discovery that combines a Workflow Planner, a Tool Code Generator, and a Scientific Mediator, with human scientists providing intuition and feedback. The system is designed to achieve Level 2 autonomy: the agent proposes workflows while humans provide high-level goals, domain heuristics, and step-by-step approval. Experiments are reported on crystal structure generation (MP-20, Matbench), crystal structure prediction (MP-20, MPTS-52, and a challenge set), property-guided generation (bandgap targets), and a workflow-validity study. The headline claim is a five-fold improvement in stability, uniqueness, and novelty rates over prior generative models on MP-20.

Significance. If the evaluation were sound, the paper would make a useful contribution by showing that LLM-driven workflow planning, combined with ML force fields and human guidance, can produce high-quality crystal candidates and competitive structure-prediction results without training a purpose-built generative model. The autonomy-level taxonomy is a helpful framing, and the inclusion of DFT-based stability evaluation and a workflow-validity analysis is a strength. However, the central quantitative claims are not currently supported because the evaluation design confounds the effect of retrieval from the MP-20 database with the effect of planning and physics; moreover, the primary LLM is not identified, and no error bars or retrieval-augmented baselines are provided. The framework is promising, but the evidence as presented is not sufficient to establish the claimed advantages.

major comments (5)
  1. [Abstract, Section 4.1, Table 1] The claim of a 'five-fold improvement in stability, uniqueness, and novelty rates' is not supported by the data in Table 1. The DFT stability rate of MAPPS is 34.3% versus 17.8% for FlowLLM, which is a factor of approximately 1.93, not five. The only metric that is roughly five-fold higher is the composite S.U.N. rate (24.9% vs 4.92%). Please report the per-component uniqueness and novelty rates separately, and rephrase the claim so that it refers specifically to the composite S.U.N. rate or to the individual components that actually show such improvement.
  2. [Section 4.1] The comparison with generative baselines is confounded by retrieval. The text states that MP-20 is 'used as the retrieval database of our method and the training set of the baselines.' Because MP-20 contains only stable or near-stable materials, MAPPS can start from known stable prototypes, relax them with CHGNet/M3GNet, and submit them to DFT, while CDVAE, DiffCSP, FlowMM, and FlowLLM must generate structures from learned distributions without retrieval. A retrieval-augmented baseline that randomly samples structures from the same database and applies the identical MLFF relaxation and DFT evaluation pipeline is needed to control for the contribution of the database itself. Without such a control, the reported S.U.N. rate and the headline improvement cannot be attributed to MAPPS's planning, physics, or autonomy.
  3. [Section 4.2] The same retrieval confound affects the crystal structure prediction results. The sentence 'To ensure a fair comparison, we also provide our agents with access to the corresponding training data for retrieval' indicates that only MAPPS (and not the baselines) is allowed to retrieve candidate prototypes during inference. For CSP, this means MAPPS can retrieve a structurally similar prototype for a given composition and then substitute atoms, which is a substantially easier task than generating a structure from scratch. The match-rate improvements (63.9% vs 61.3% on MP-20, 27.6% vs 23.1% on MPTS-52, and 31.0% vs 22.4% on the challenge set) may therefore reflect the retrieval mechanism rather than workflow planning. A baseline that retrieves a random or similarity-based prototype from the training data and proceeds with the same relaxation pipeline should be included.
  4. [Section 4] The main experimental results do not identify which LLM or reasoning model is used for the Workflow Planner and Tool Code Generator. The text says only 'We conduct our experiments using OpenAI API and a single NVIDIA A100 GPU.' Given that Section 4.4 shows that workflow validity varies from 0% to 100% depending on the model (GPT-4o-mini, GPT-4o, O3-mini), the Tables 1-4 results cannot be reproduced or assessed without naming the exact model and version, along with sampling parameters. Additionally, no error bars, standard deviations, or multi-seed runs are reported for any of the main metrics, so it is unclear whether the reported differences are statistically meaningful.
  5. [Appendix C] The definition of novelty is partially circular with the retrieval setup. Appendix C states that a crystal is novel if it does not match any existing structure in the original dataset, and the same dataset is used as MAPPS's retrieval source. A structure retrieved from MP-20 and then relaxed or slightly modified can be counted as novel according to the StructureMatcher criterion while retaining the thermodynamic stability of the original prototype. This inflates the S.U.N. rate in a way that cannot be separated from retrieval. Please either compute novelty against a held-out set (for example, a newer release of the Materials Project or a temporally later test set) or provide an analysis that shows how many of the 'novel' structures originate from retrieved prototypes and how much perturbation was applied.
minor comments (6)
  1. [Section 2.2] The heading contains a typo: 'Planing' should be 'Planning'.
  2. [References] References [30] and [32] are duplicates; both cite the same Pymatgen paper. Please remove the duplicate and renumber.
  3. [Table 1] The row for CrystalTextLLM has a missing stability value and the column header for the M3GNet metric is split across lines; please format the table so all entries are complete and aligned.
  4. [Section 4.3] For the bandgap-constrained generation results, please clarify how many independent runs or random seeds were used to obtain the 74.6% and 92.2% satisfaction rates, and report confidence intervals or standard errors.
  5. [Appendix B, Figure 4] Please add axis labels and a caption note indicating whether the histograms are over the full 500 generated structures or only the valid/unique subset, and state the bin width used.
  6. [Section 5] The limitations paragraph acknowledges that MAPPS does not reach Level 3 autonomy and focuses only on materials tasks; this is appropriate, but the abstract and introduction should more consistently refer to 'Level 2 autonomy' rather than 'greater autonomy' in a way that might overstate the result.

Circularity Check

2 steps flagged · score 6.0 of 10

S.U.N. and bandgap "predictions" reduce to retrieval from the same database used for evaluation; no retrieval-augmented baseline is provided.

  1. self definitional [Section 4.1 (Crystal Structure Generation, Setup) and Appendix C (Novelty); cf. Section 2 (task formulation)]
    ""The datasets are used as the retrieval database of our method and the training set of the baselines." (Sec. 4.1); "A crystal is considered novel if it does not match any existing structure in the original dataset, again based on the StructureMatcher." (App. C); "The process may start from an empty structure or from a candidate retrieved from a database." (Sec. 2)"

    MAPPS is explicitly allowed to start from candidates retrieved from MP-20/Matbench, while the baselines are only trained on that same data. Stability can therefore be inherited from known near-hull or on-hull prototypes, and novelty is judged against the very database from which the candidates were drawn. Because novelty excludes only exact matches, a retrieved structure that is relaxed or lightly modified counts as novel. The five-fold S.U.N./stability advantage over generative models is thus partly a property of the evaluation protocol (retrieval vs. no retrieval), not a demonstrated effect of planning or physics; no retrieval-augmented baseline is reported.

  2. self definitional [Section 4.3 (Discovering crystal structures with desired properties, Setup and Results)]
    ""For retrieval, we use the JARVIS-DFT dataset [1], which contains 61,541 crystal structures along with their corresponding bandgap values." (Sec. 4.3); "74.6% of the generated crystals have band gap values greater than 3 eV." (Sec. 4.3 Results)"

    The property-satisfaction 'prediction' is made after retrieving candidates from a database that already contains the target property labels. Since the retrieval source is chosen precisely because it carries bandgap values, and the generation condition is defined by those values, the reported DFT condition-satisfaction rates (e.g., 74.6% >3 eV) largely measure retrieval selection from a property-labeled database rather than de novo generative or planning ability. The result is forced by the retrieval input by construction, and no baseline that retrieves without MAPPS planning is given to separate the two effects.

full rationale

The core derivation for the headline 'five-fold improvement' is: retrieve from MP-20/Matbench (or JARVIS-DFT), relax with ML force fields, validate with DFT, then measure stability, uniqueness, novelty, or bandgap satisfaction against the same source used for retrieval. Because the retrieval source equals the evaluation reference (MP-20 for S.U.N.; JARVIS-DFT for bandgap), the reported rates are partially constructed by the choice of retrieval database, not by the planning/physics pipeline. The paper provides no retrieval-augmented baseline (e.g., direct retrieval plus relaxation without MAPPS workflow reasoning), so the attribution of the gain to MAPPS's planning is not established. Section 4.4's finding that all workflows are 0% valid without human intuition is a limitation on the autonomy claim, but it is not circularity. No load-bearing self-citation or uniqueness-theorem smuggling was found: prior work by the same authors appears as background, and the DFT/MLFF tools are external benchmarks. The most severe issue is the retrieval/evaluation equivalence for the headline metrics, which makes the claimed improvement partly definitional rather than empirical. Score 6 reflects this partial circularity.

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

The central claim rests on the reliability of ML surrogates (CHGNet/M3GNet), the correctness of the convex hull reference, the LLM's code-generation ability under self-reflection, and the authors' hand-crafted human intuition. No new physical entities are introduced. The potential circularity in the novelty metric is discussed under circularity score.

assumptions (5)
  • domain assumption CHGNet and M3GNet provide reliable estimates of decomposition energy for screening
    Used in Section 4.1 as surrogate models to compute metastability; the paper does not validate these surrogates on the generated structures beyond prior literature.
  • domain assumption The MP-20 dataset is an appropriate retrieval database and its convex hull (via Matbench Discovery) is the correct reference for stability evaluation
    Section 4.1/D: DFT energy above hull is computed against the Matbench Discovery convex hull; if the convex hull is not consistent with the retrieval set, stability rates could be misestimated.
  • domain assumption The LLM can generate executable code given a workflow step and error feedback
    Section 2.5 and Appendix A.2; the paper's self-reflection loop assumes revision after errors converges to working code, but no success rate is reported.
  • ad hoc to paper Human intuition is necessary and sufficient for workflow validity
    Section 4.4, Table 5: without human intuition, all workflows are invalid. The paper uses human-authored intuition prompts as an input, which is central to the method but comes from the authors, not from the agent.
  • domain assumption StructureMatcher tolerances define 'novel' correctly
    Section C: novelty is determined by pymatgen StructureMatcher; the choice of tolerances is not justified and directly affects the S.U.N. rate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Toward Greater Autonomy in Materials Discovery Agents: Unifying Planning, Physics, and Scientists." pith.science (2026). https://pith.science/paper/SJJ5ZLE3

@misc{pith2026250605616,
  author       = {Pith},
  title        = {Pith review of: Toward Greater Autonomy in Materials Discovery Agents: Unifying Planning, Physics, and Scientists},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SJJ5ZLE3}},
  note         = {Machine review of arXiv:2506.05616}
}
read the original abstract

We aim at designing language agents with greater autonomy for crystal materials discovery. While most of existing studies restrict the agents to perform specific tasks within predefined workflows, we aim to automate workflow planning given high-level goals and scientist intuition. To this end, we propose Materials Agent unifying Planning, Physics, and Scientists, known as MAPPS. MAPPS consists of a Workflow Planner, a Tool Code Generator, and a Scientific Mediator. The Workflow Planner uses large language models (LLMs) to generate structured and multi-step workflows. The Tool Code Generator synthesizes executable Python code for various tasks, including invoking a force field foundation model that encodes physics. The Scientific Mediator coordinates communications, facilitates scientist feedback, and ensures robustness through error reflection and recovery. By unifying planning, physics, and scientists, MAPPS enables flexible and reliable materials discovery with greater autonomy, achieving a five-fold improvement in stability, uniqueness, and novelty rates compared with prior generative models when evaluated on the MP-20 data. We provide extensive experiments across diverse tasks to show that MAPPS is a promising framework for autonomous materials discovery.

Figures

Figures reproduced from arXiv: 2506.05616 by the authors.

Figure 1
Figure 1. MAPPS Agent Framework. The MAPPS framework consists of three key modules: the [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Examples of generated material structures in the Crystal Structure Generation task [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Examples of generated material structures in the Crystal Structure Prediction task [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: DFT-computed bandgap distributions under two generation settings. [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 19 canonical work pages

  1. [1]

    The joint automated repository for various integrated simulations (jarvis) for data-driven materials design.npj computational materials, 6(1):173, 2020

    Kamal Choudhary, Kevin F Garrity, Andrew CE Reid, Brian DeCost, Adam J Biacchi, Angela R Hight Walker, Zachary Trautt, Jason Hattrick-Simpers, A Gilad Kusne, Andrea Centrone, et al. The joint automated repository for various integrated simulations (jarvis) for data-driven materials design.npj computational materials, 6(1):173, 2020

    2020

  2. [2]

    Periodic graph transformers for crystal material property prediction

    Keqiang Yan, Yi Liu, Yuchao Lin, and Shuiwang Ji. Periodic graph transformers for crystal material property prediction. InThe 36th Annual Conference on Neural Information Processing Systems, pages 15066–15080, 2022

    work page 2022

  3. [3]

    A space group symmetry informed network for O(3) equivariant crystal tensor prediction

    Keqiang Yan, Alexandra Saxton, Xiaofeng Qian, Xiaoning Qian, and Shuiwang Ji. A space group symmetry informed network for O(3) equivariant crystal tensor prediction. InProceedings of the 41st International Conference on Machine Learning, pages 55797–55813, 2024

    work page 2024

  4. [4]

    Complete and efficient graph transformers for crystal material property prediction

    Keqiang Yan, Cong Fu, Xiaofeng Qian, Xiaoning Qian, and Shuiwang Ji. Complete and efficient graph transformers for crystal material property prediction. InInternational Conference on Learning Representations, 2024

    work page 2024

  5. [5]

    JARVIS- Leaderboard: a large scale benchmark of materials design methods.npj Computational Materi- als, 10(1):93, 2024

    Kamal Choudhary, Daniel Wines, Kangming Li, Kevin F Garrity, Vishu Gupta, Aldo H Romero, Jaron T Krogel, Kayahan Saritas, Addis Fuhr, Panchapakesan Ganesh, et al. JARVIS- Leaderboard: a large scale benchmark of materials design methods.npj Computational Materi- als, 10(1):93, 2024

    work page 2024

  6. [6]

    Crystal structure prediction by joint equivariant diffusion.Advances in Neural Information Processing Systems, 2023

    Rui Jiao, Wenbing Huang, Peijia Lin, Jiaqi Han, Pin Chen, Yutong Lu, and Yang Liu. Crystal structure prediction by joint equivariant diffusion.Advances in Neural Information Processing Systems, 2023

    work page 2023

  7. [7]

    Crystal structure generation with autoregressive large language modeling.Nature Communications, 15(1):1–16, 2024

    Luis M Antunes, Keith T Butler, and Ricardo Grau-Crespo. Crystal structure generation with autoregressive large language modeling.Nature Communications, 15(1):1–16, 2024

    2024

  8. [8]

    Invariant tokenization of crystalline materials for language model enabled generation.Advances in Neural Information Processing Systems, 37:125050–125072, 2024

    Keqiang Yan, Xiner Li, Hongyi Ling, Kenna Ashen, Carl Edwards, Raymundo Arróyave, Marinka Zitnik, Heng Ji, Xiaofeng Qian, Xiaoning Qian, et al. Invariant tokenization of crystalline materials for language model enabled generation.Advances in Neural Information Processing Systems, 37:125050–125072, 2024. 10

    work page 2024

Show all 39 references
  1. [9]

    Hofgard, Aria Mansouri Tehrani, Rui Wang, Ameya Daigavane, Montgomery Bohde, Jerry Kurtin, Qian Huang, Tuong Phung, Minkai Xu, Chaitanya K

    Xuan Zhang, Limei Wang, Jacob Helwig, Youzhi Luo, Cong Fu, Yaochen Xie, Meng Liu, Yuchao Lin, Zhao Xu, Keqiang Yan, Keir Adams, Maurice Weiler, Xiner Li, Tianfan Fu, Yucheng Wang, Haiyang Yu, YuQing Xie, Xiang Fu, Alex Strasser, Shenglong Xu, Yi Liu, Yuanqi Du, Alexandra Saxto...

    2023 arXiv

  2. [10]

    Llmatdesign: Autonomous materials discovery with large language models.arXiv preprint arXiv:2406.13163, 2024

    Shuyi Jia, Chao Zhang, and Victor Fung. Llmatdesign: Autonomous materials discovery with large language models.arXiv preprint arXiv:2406.13163, 2024

    2024 arXiv

  3. [11]

    A comprehensive survey of scientific large language models and their applications in scientific discovery

    Yu Zhang, Xiusi Chen, Bowen Jin, Sheng Wang, Shuiwang Ji, Wei Wang, and Jiawei Han. A comprehensive survey of scientific large language models and their applications in scientific discovery. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processi...

    2024

  4. [12]

    Crystal Dif- fusion Variational Autoencoder for Periodic Material Generation

    Tian Xie, Xiang Fu, Octavian-Eugen Ganea, Regina Barzilay, and Tommi Jaakkola. Crystal Dif- fusion Variational Autoencoder for Periodic Material Generation. InInternational Conference on Learning Representations, 2022

    2022

  5. [13]

    Automating alloy design and discovery with physics-aware multimodal multiagent ai.Proceedings of the National Academy of Sciences, 122(4):e2414074122, 2025

    Alireza Ghafarollahi and Markus J Buehler. Automating alloy design and discovery with physics-aware multimodal multiagent ai.Proceedings of the National Academy of Sciences, 122(4):e2414074122, 2025

    2025

  6. [14]

    Osda agent: Leveraging large language models for de novo design of organic structure directing agents

    Zhaolin Hu, Yixiao Zhou, Zhongan Wang, Xin Li, Weimin Yang, Hehe Fan, and Yi Yang. Osda agent: Leveraging large language models for de novo design of organic structure directing agents. InThe Thirteenth International Conference on Learning Representations, 2025

    2025

  7. [15]

    Large language models are innate crystal structure generators

    Jingru Gan, Peichen Zhong, Yuanqi Du, Yanqiao Zhu, Chenru Duan, Haorui Wang, Daniel Schwalbe-Koda, Carla P Gomes, Kristin Persson, and Wei Wang. Large language models are innate crystal structure generators. InAI for Accelerated Materials Design-ICLR, 2025

    2025

  8. [16]

    Alireza Ghafarollahi and Markus J. Buehler. Protagents: Protein discovery via large language model multi-agent collaborations combining physics and machine learning.arXiv preprint arXiv:2402.04268, 2024

    2024 arXiv

  9. [17]

    Drugagent: Automating ai-aided drug discovery programming through llm multi-agent collaboration.arXiv preprint arXiv:2411.15692, 2024

    Sizhe Liu, Yizhou Lu, Siyu Chen, Xiyang Hu, Jieyu Zhao, Tianfan Fu, and Yue Zhao. Drugagent: Automating ai-aided drug discovery programming through llm multi-agent collaboration.arXiv preprint arXiv:2411.15692, 2024

    2024 arXiv

  10. [18]

    Dohhyun Kang and Joy D. Kim. Chatmof: An autonomous ai system for predicting and generating metal-organic frameworks.arXiv preprint arXiv:2308.01423, 2023

    2023 arXiv

  11. [19]

    Toward a team of ai-made scientists for scientific discovery from gene expression data.arXiv preprint arXiv:2402.12391, 2024

    Haoyang Liu, Yijiang Li, Jinglin Jian, et al. Toward a team of ai-made scientists for scientific discovery from gene expression data.arXiv preprint arXiv:2402.12391, 2024

    2024 arXiv

  12. [20]

    Crispr-gpt: An llm agent for automated design of gene-editing experiments.arXiv preprint arXiv:2404.18021, 2024

    Yixin Zhao, Rui Wang, Lixin Zhang, et al. Crispr-gpt: An llm agent for automated design of gene-editing experiments.arXiv preprint arXiv:2404.18021, 2024

    2024 arXiv

  13. [21]

    Coleman, et al

    Mihailo Bran, Andrew White, Connor W. Coleman, et al. Chemcrow: Augmenting large- language models with chemistry tools.arXiv preprint arXiv:2304.05376, 2023

    2023 arXiv

  14. [22]

    Commentary: The Materials Project: A materials genome approach to accelerating materials innovation.APL Materials, 1(1):011002, 2013

    Anubhav Jain, Shyue Ping Ong, Geoffroy Hautier, Wei Chen, William Davidson Richards, Stephen Dacek, Shreyas Cholia, Dan Gunter, David Skinner, Gerbrand Ceder, et al. Commentary: The Materials Project: A materials genome approach to accelerating materials innovation.APL Materia...

    2013

  15. [23]

    Benchmarking materials property prediction methods: the matbench test set and automatminer reference algorithm.npj Computational Materials, 6(1):138, 2020

    Alexander Dunn, Qi Wang, Alex Ganose, Daniel Dopp, and Anubhav Jain. Benchmarking materials property prediction methods: the matbench test set and automatminer reference algorithm.npj Computational Materials, 6(1):138, 2020

    2020

  16. [24]

    3-d inorganic crystal structure generation and property prediction via representation learning.Journal of Chemical Information and Modeling, 60(10):4518–4535, 2020

    Callum J Court, Batuhan Yildirim, Apoorv Jain, and Jacqueline M Cole. 3-d inorganic crystal structure generation and property prediction via representation learning.Journal of Chemical Information and Modeling, 60(10):4518–4535, 2020

    2020

  17. [25]

    Flowmm: Generating materials with riemannian flow matching

    Benjamin Kurt Miller, Ricky TQ Chen, Anuroop Sriram, and Brandon M Wood. Flowmm: Generating materials with riemannian flow matching. InForty-first International Conference on Machine Learning, 2024

    2024

  18. [26]

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

    2023

  19. [27]

    A universal graph deep learning interatomic potential for the periodic table.Nature Computational Science, 2(11):718–728, 2022

    Chi Chen and Shyue Ping Ong. A universal graph deep learning interatomic potential for the periodic table.Nature Computational Science, 2(11):718–728, 2022

    2022

  20. [28]

    Fine-tuned language models generate stable inorganic materials as text

    Nate Gruver, Anuroop Sriram, Andrea Madotto, Andrew Gordon Wilson, C Lawrence Zitnick, and Zachary Ulissi. Fine-tuned language models generate stable inorganic materials as text. In International Conference on Learning Representations, 2024

    2024

  21. [29]

    Flowllm: Flow matching for material generation with large language models as base distributions.Advances in Neural Information Processing Systems, 2024

    Anuroop Sriram, Benjamin Miller, Ricky TQ Chen, and Brandon Wood. Flowllm: Flow matching for material generation with large language models as base distributions.Advances in Neural Information Processing Systems, 2024

    2024

  22. [30]

    Python Materials Genomics (pymatgen): A robust, open-source python library for materials analysis.Computational Materials Science, 68:314–319, 2013

    Shyue Ping Ong, William Davidson Richards, Anubhav Jain, Geoffroy Hautier, Michael Kocher, Shreyas Cholia, Dan Gunter, Vincent L Chevrier, Kristin A Persson, and Gerbrand Ceder. Python Materials Genomics (pymatgen): A robust, open-source python library for materials analysis.C...

    2013

  23. [31]

    Smact: Semiconducting materials by analogy and chemical theory.Journal of Open Source Software, 4(38):1361, 2019

    Daniel W Davies, Keith T Butler, Adam J Jackson, Jonathan M Skelton, Kazuki Morita, and Aron Walsh. Smact: Semiconducting materials by analogy and chemical theory.Journal of Open Source Software, 4(38):1361, 2019

    2019

  24. [32]

    Python materials genomics (pymatgen): A robust, open-source python library for materials analysis.Computational Materials Science, 68:314–319, 2013

    Shyue Ping Ong, William Davidson Richards, Anubhav Jain, Geoffroy Hautier, Michael Kocher, Shreyas Cholia, Dan Gunter, Vincent L Chevrier, Kristin A Persson, and Gerbrand Ceder. Python materials genomics (pymatgen): A robust, open-source python library for materials analysis.C...

    2013

  25. [33]

    Inhomogeneous electron gas.Phys

    Pierre Hohenberg and Walter Kohn. Inhomogeneous electron gas.Phys. Rev., 136(1964):B864, 1964

    1964

  26. [34]

    Self-consistent equations including exchange and correlation effects.Physical Review, 140(4A):A1133, 1965

    Walter Kohn and Lu Jeu Sham. Self-consistent equations including exchange and correlation effects.Physical Review, 140(4A):A1133, 1965

    1965

  27. [35]

    Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set.Physical Review B, 54(16):11169, 1996

    Georg Kresse and Jürgen Furthmüller. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set.Physical Review B, 54(16):11169, 1996

    1996

  28. [36]

    Generalized gradient approximation made simple.Physical Review Letters, 77(18):3865, 1996

    John P Perdew, Kieron Burke, and Matthias Ernzerhof. Generalized gradient approximation made simple.Physical Review Letters, 77(18):3865, 1996

    1996

  29. [37]

    Beyond the local-density approximation in calculations of ground-state electronic properties.Physical Review B, 28(4):1809, 1983

    David C Langreth and MJ Mehl. Beyond the local-density approximation in calculations of ground-state electronic properties.Physical Review B, 28(4):1809, 1983

    1983

  30. [38]

    Matbench discovery–a framework to evaluate machine learning crystal stability predictions.arXiv preprint arXiv:2308.14920, 2023

    Janosh Riebesell, Rhys EA Goodall, Philipp Benner, Yuan Chiang, Bowen Deng, Alpha A Lee, Anubhav Jain, and Kristin A Persson. Matbench discovery–a framework to evaluate machine learning crystal stability predictions.arXiv preprint arXiv:2308.14920, 2023

    2023 arXiv

  31. [39]

    Please predict the stable structure forBa 2Fe2F9

    Jiˇrí Klimeš, David R Bowler, and Angelos Michaelides. Chemical accuracy for the van der waals density functional.Journal of Physics: Condensed Matter, 22(2):022201, 2009. 12 A Experimental Details A.1 Workflow Planning for Crystal Structure Prediction The workflow example in ...

    2009

Pith tools

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