• Format: All submissions must be in PDF format using the NeurIPS 2024 LaTeX style file . Please include the references and supplementary materials in the same PDF as the main paper. The maximum file size for submissions is 50MB. Submissions that violate the NeurIPS style (e.g., by decreasing margins or font sizes) or page limits may be rejected without further review. Filling the NeurIPS checklist with the paper is not compulsory for AI4Mat.

• Double-Blind Reviewing: The reviewing process will be double blind. As an author, you are responsible for anonymizing your submission. In particular, you should not include author names, author affiliations, or acknowledgements in your submission and you should avoid providing any other identifying information (even in the supplementary material).

• Dual-Submission Policy: We welcome ongoing and unpublished work. We will also accept papers that are under review at any venue at the time of submission. Submissions under review in venues for related fields (e.g. materials science, chemistry) are welcome. We will consider previously published work if we can think it can add to the discussion and knowledge sharing quality of the workshop. In such cases, author should make clear why they are submitting previously published work.

• Non-Archival: The workshop is a non-archival venue and will not have official proceedings. Workshop submissions can be subsequently or concurrently submitted to other venues. 

• Visibility: Submissions and reviews will not be public. Only accepted papers across the different tracks will be made public on the workshop website.

• AI4Mat-NeurIPS 2024 Themed Track: We especially encourage aligned to the themed track - "LLMs for Materials Science"

• Goals: The goal of the LLMs for Materials Science track is to encourage submissions of research that has strong relevance in enabling the application of LLMs to materials science and furthers community understanding of challenges, capabilities and limitations related to LLMs applied to accelerated materials discovery. Building on top of AI4Mat-2023’s LLM Fireside Chat, this track aims to highlight both the benefits and challenges of using LLMs in materials discovery to foster community discussion. Example research topics include, but are not limited to:

  • • Accessible use of LLMs (e.g., using laptops or workstations commonly found in materials science laboratories) to solve relevant material design and discovery challenges

    • Understanding limitations and failure cases of LLMs applied to materials discovery (e.g., hallucinations in a materials discovery context)

    • Datasets and benchmarks that help further the application of LLMs in materials discovery (e.g., knowledge extraction from diverse, multi-modal documents) 

    • Retrieval augmented generation (RAG), agentic systems, multi-modal LLMs (e.g. VLMs) applied to materials discovery challenges

    • Training and fine-tuning LLMs given common challenges in materials discovery (e.g., small and continuously changing experimental datasets)

Through this track, we aim to learn more about and share with the community the various technical considerations and best practices for using LLMs in automated materials design.

• Instructions: Similar to the Paper Track, we encourage submissions of short-form papers up to 4 pages in length with unlimited pages for references and supplementary materials. We will also consider full-length papers of up to 9 pages in length for works that have been submitted or works that authors intend to submit to other venues, such as ICLR 2025 or peer-reviewed journals. Submissions should be clearly identified as short-form or full-length and we discourage works that do not fall into either category (e.g. 6 page submissions that do not meet the standards for full-length papers). We will ask reviewers to apply higher standards to full-length submissions as those are assumed to be more finished work compared to short-form papers that are considered to be work in progress.

This year, AI4Mat is partnering with Machine Learning: Science and Technology (MLST), an IOP Publishing journal. Top-tier submissions on AI for materials design will be considered for publication in a focus collection in the journal. The submission process for the workshop will remain the same as for AI4Mat-24 with a single round of reviews through OpenReview. For the focus collection in MLST, manuscripts will undergo some additional steps:

• The program committee will identify high-quality manuscripts from the set of accepted contributions and recommend them for submission to the MLST focus collection.

• Authors will have the option to indicate if they would like their submission to be considered for the MLST focus collection for AI4Mat-24. 

  • Authors of recommended manuscripts interested in publication in MLST will have to prepare a full article (article types can be found here) for submission through the MLST portal. Author guidelines can be found here.

  • Authors should keep in mind that publication in MLST constitutes a peer-reviewed publication of the work and as such the manuscript may not be published in other venues depending on dual submission policies. This is different from regular workshop submissions which are non-archival.

  • Manuscripts submitted to the MLST focus collection will undergo peer review through MLST, in the same manner and to the same high standard as regular issue articles. Only accepted manuscripts will be included in the collection.

  • The focus collection will be finalized and shared with the community in 2025. 

The program committee would like to highlight some important features of the MLST focus collection:

• • IOP Publishing has transformative agreements with many different organizations that allow publishing open access without a fee. Check the full list of institutions on the MLST website.

  • Articles in the focus collection will be considered finished work and the manuscript will be reviewed according to that standard.

  • It is possible for authors to submit a 4-page extended abstract in the first round of the workshop reviews and later extend to a full manuscript for consideration for the collection.

  • Accepted articles will be published in regular MLST issues on acceptance without being delayed by other papers in the collection.

  • The program committee is open to including perspective and review articles in the collection. Both formats can be submitted through the OpenReview portal.

  • Only submissions to the workshop will be considered for the focus collection. It is not possible to submit manuscripts for consideration after the workshop submission deadline.

• Goals: The goal of the Paper Track is to highlight research work related to automated materials discovery that pushes the state-of-the-art and foster discussion among workshop participants. All submissions should explain why the proposed work helps accelerate material discovery and can be related thematically to the three distinct parts of self-driving laboratories (AI-Guided Design, Automated Synthesis, Automated Characterization) in cases where the relation may be ambiguous. If a submission does not fit into one of the aforementioned thematic tracks, we encourage the authors to provide a detailed explanation of why their work relates to automated materials discovery. Example research topics include, but are not limited to:

  • AI-Guided Materials Design:

    • Machine learning algorithms and deep learning architectures for accelerated materials simulations and property modelling 

    • Generative algorithms for materials discovery based on diverse machine learning techniques (e.g. diffusion models, flow matching networks, reinforcement learning, GFlowNets)

    • Datasets, benchmarks and analysis methods

  • Automated Synthesis

    • Optimization and discovery of materials synthesis and chemical synthesis procedures 

    • Datasets, benchmarks, knowledge extraction and data processing techniques for materials synthesis

    • Machine learning algorithms for small-data regimes (e.g. active learning with costly data acquisition)

  • Automated Characterization

    • Analysis of real-world characterization data, e.g., microscopy data (including multi-modal data like images, spectra and diffraction patterns), x-ray diffraction, optimal measurements, property measurements

    • Datasets, benchmarks and automation frameworks for data collection and analysis in characterization tools amenable to machine learning algorithms

    • Machine learning algorithms in defect and anomaly detection in settings relevant to automated materials design

• Instructions: We encourage submissions of short-form papers up to 4 pages in length with unlimited pages for references and supplementary materials. We will also consider full-length papers of up to 9 pages in length for works that have been submitted or works that authors intend to submit to other venues, such as peer-reviewed conferences and journals. Submissions should be clearly identified as short-form or full-length and we discourage works that do not fall into either category (e.g. 6 page submissions that do not meet the standards for full-length papers). We will ask reviewers to apply higher standards to full-length submissions as those are assumed to be more finished work compared to short-form papers that are considered to be work in progress.

• Goals: The goal of the Tools Track is to encourage submissions of useful tools for research in automated materials design that can be disseminated to the research community through the workshop. Tools can be conceptual (e.g. AI for new data modalities in materials science) or practical in nature (e.g software libraries) as long as they have a clear relation to advancing the research themes of the workshop. We encourage submissions that describe novel experimental equipment, tools and workflows that can facilitate the application of machine learning to materials discovery. 

• Instructions: We encourage submissions up to 4 pages in length along with supplementary materials outlining the primary content of the submissons. For software submissions, we encourage supplementary materials in the form of a notebook showcasing the presented tools. For conceptual or experimental tools, we encourage supplementary materials in the form of a presentation and video through the NeurIPS workshop site explaining the concepts in an accessible manner to the NeurIPS community.

• Goals: The goal of the Findings & Open Challenges Track is to encourage submissions of research that has strong relevance in enabling the application of AI to automated materials design, but might not fit perfectly with the other tracks. Submissions to the findings track may include, but are not limited to:

• • Open Challenges: Submissions introducing and discussing overlooked scientific questions and potential future directions for a given application area. We encourage submission that address open challenges and describe: 1. Why the current research and state-of-the-art fall short for a given challenges; 2. What directions the authors believe the community can focus on to help address the open challenge. We hope that submissions describing open challenges will enable the AI4Mat community to expand the range of interdisciplinary research the research community is working on.

  • Behind the Scenes: Essential engineering work that can often be lost in research discussions, such as dataset preparation, putting together simulations of complex systems or assembling intricate hardware systems for different types of robotic automation workflows.

  • Surveys: Survey papers centered around a relevant theme for the workshop that provides new conclusions based on the analysis of the existing literature or highlights new insights or ideas that have received less attention.

  • Responsible Use: Submissions discussing responsible use, in an inclusive sense, of data and methods related to AI for automated materials design.

• Through the findings track, we aim to learn more about and share with the community the various technical considerations and best practices needed to develop the complex systems required for state-of-the-art automated materials design.

• Instructions: We encourage submissions of short-form papers up to 5 pages in length with unlimited pages for references and supplementary materials. Submissions in this track should clearly explain how the proposed work helps accelerate material discovery in cases where the relation may be ambiguous. Reviewers will be asked to evaluate the thoroughness and quality of technical work described in the submission.