{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,14]],"date-time":"2026-09-14T14:05:39Z","timestamp":1789394739532,"version":"build-2803163510"},"reference-count":53,"publisher":"American Chemical Society (ACS)","issue":"17","license":[{"start":{"date-parts":[[2026,8,27]],"date-time":"2026-08-27T00:00:00Z","timestamp":1787788800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100006151","name":"Basic Energy Sciences","doi-asserted-by":"publisher","award":["DE-AC02-06CH11357"],"award-info":[{"award-number":["DE-AC02-06CH11357"]}],"id":[{"id":"10.13039\/100006151","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/na","name":"Engineering and Physical Sciences Research Council","doi-asserted-by":"publisher","award":["210153"],"award-info":[{"award-number":["210153"]}],"id":[{"id":"10.13039\/na","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,9,14]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Domain-specific language models are attractive because they can be tailored to suit their users. However, the high-end computational resources needed to pretrain large language models (LLMs) by conventional methods tend to prohibit their development. Furthermore, there is a dearth of large, domain-specific, question-answering (QA) data sets to finetune LLMs for prompt engineering, even when computational resources can be found to pretrain LLMs. Our paper presents a method for autogenerating a data set of 168,080 domain-specific QA pairs relating to magnetic materials. We show how this can be employed to finetune LLMs for the downstream task of Question-Answering (QA) for the magnetic materials domain. We use the bidirectional encoder representations from transformers (BERT) architecture as our benchmark LLM to compare the relative performance of 6 LLMs. These include 3 vanilla (BERT-base-cased) LLMs that have been finetuned on various QA data sets (i) our domain-specific QA data set (MagQA); (ii) the Stanford Question-Answering Data set (SQuAD v2); (iii) our QA data set mixed with SQuAD v2. The other 3 (BERT-base-cased) LLMs have been subjected to domain-adaptive pretraining on a corpus of 97,308 magnetic papers prior to their finetuning on the same combination of QA data sets (i)-(iii). As well as our QA data set autogeneration pipeline, we show how our work can be extrapolated to the larger architectures of RoBERTa-base and DeBERTa-base. We also undertake an ablation study of the finetuning process using four sizes of QA data sets to gain insights into the quantity of data that are needed for sufficient domain-specification. The highest performing LLM was MagBERT_MagQA_Mixed, a vanilla BERT-base-cased LLM that had been finetuned on the combination of our autogenerated MagQA data set and SQuAD v2, achieving an F1 score of 78.43% and an exact-match score of 72.84% when tested on a manually annotated data set on magnetic materials. These results demonstrate that domain-specific BERT models only need to be finetuned from vanilla BERT models (i.e. no domain-adaptive pretraining is needed), pending the availability of sufficiently large, high-quality, domain-specific QA data sets that this work shows how to autogenerate.<\/jats:p>","DOI":"10.1021\/acs.jcim.6c00136","type":"journal-article","created":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T20:18:27Z","timestamp":1788293907000},"page":"10896-10918","source":"Crossref","is-referenced-by-count":0,"title":["Autogenerating\na Domain-Specific Question-Answering\nData Set to Enable High-Performing Language Models for Magnetic Materials"],"prefix":"10.1021","volume":"66","author":[{"given":"Charles R.","family":"Kelly","sequence":"first","affiliation":[{"name":"University of Cambridge , , J. J. Thomson Avenue , ,","place":["Cambridge, U.K., CB3 0US"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1552-8743","authenticated-orcid":true,"given":"Jacqueline M.","family":"Cole","sequence":"additional","affiliation":[{"name":"University of Cambridge , , J. J. Thomson Avenue , ,","place":["Cambridge, U.K., CB3 0US"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"316","published-online":{"date-parts":[[2026,8,27]]},"reference":[{"issue":"5","key":"2026091409065599500_cit1","doi-asserted-by":"publisher","first-page":"2476","DOI":"10.1021\/acs.jcim.4c02029","article-title":"Cost-Efficient Domain-Adaptive Pretraining of Language\nModels for Optoelectronics Applications","volume":"65","author":"Huang","year":"2025","journal-title":"J.\nChem. Inf. Model."},{"issue":"8","key":"2026091409065599500_cit2","doi-asserted-by":"publisher","first-page":"3205","DOI":"10.1021\/acs.jcim.4c00063","article-title":"How Beneficial Is\nPretraining on a Narrow Domain-Specific\nCorpus for Information Extraction about Photocatalytic Water Splitting?","volume":"64","author":"Isazawa","year":"2024","journal-title":"J. Chem. Inf. Model."},{"issue":"4","key":"2026091409065599500_cit3","doi-asserted-by":"publisher","first-page":"1873","DOI":"10.1021\/acs.jcim.4c00857","article-title":"MechBERT:\nLanguage Models for Extracting\nChemical and Property Relationships about Mechanical Stress and Strain","volume":"65","author":"Kumar","year":"2025","journal-title":"J. Chem. Inf. Model."},{"issue":"24","key":"2026091409065599500_cit4","doi-asserted-by":"publisher","first-page":"6365","DOI":"10.1021\/acs.jcim.2c00035","article-title":"BatteryBERT: A Pretrained\nLanguage Model for Battery\nDatabase Enhancement","volume":"62","author":"Huang","year":"2022","journal-title":"J. Chem. Inf. Model."},{"issue":"7","key":"2026091409065599500_cit5","doi-asserted-by":"publisher","first-page":"1961","DOI":"10.1021\/acs.jcim.2c01259","article-title":"OpticalBERT and OpticalTable-SQA:\nText- and Table-Based Language Models for the Optical-Materials Domain","volume":"63","author":"Zhao","year":"2023","journal-title":"J. Chem. Inf. Model."},{"issue":"1","key":"2026091409065599500_cit6","doi-asserted-by":"publisher","first-page":"102","DOI":"10.1038\/s41524-022-00784-w","article-title":"MatSciBERT:\nA materials domain language\nmodel for text mining and information extraction","volume":"8","author":"Gupta","year":"2022","journal-title":"Npj Comput. Mater."},{"issue":"4","key":"2026091409065599500_cit7","doi-asserted-by":"publisher","first-page":"1234","DOI":"10.1093\/bioinformatics\/btz682","article-title":"BioBERT: a pre-trained biomedical language representation\nmodel for\nbiomedical text mining","volume":"36","author":"Lee","year":"2020","journal-title":"Bioinformatics"},{"key":"2026091409065599500_cit8","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1903.10676","article-title":"SciBERT: A Pretrained Language Model\nfor Scientific Text","author":"Beltagy","year":"2019","journal-title":"arXiv"},{"key":"2026091409065599500_cit9","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1905.05529","article-title":"Entity-Relation\nExtraction as Multi-Turn Question Answering","author":"Li","year":"2019","journal-title":"arxiv"},{"issue":"5","key":"2026091409065599500_cit10","doi-asserted-by":"publisher","first-page":"1233","DOI":"10.1039\/D3DD00113J","article-title":"14 examples of how LLMs\ncan transform materials science and chemistry:\na reflection on a large language model hackathon","volume":"2","author":"Jablonka","year":"2023","journal-title":"Digital Discovery"},{"issue":"1","key":"2026091409065599500_cit11","doi-asserted-by":"publisher","first-page":"1418","DOI":"10.1038\/s41467-024-45563-x","article-title":"Structured information\nextraction from scientific text with large\nlanguage models","volume":"15","author":"Dagdelen","year":"2024","journal-title":"Nat. Commun."},{"issue":"2","key":"2026091409065599500_cit12","doi-asserted-by":"publisher","first-page":"328","DOI":"10.1039\/D3DD00202K","article-title":"Harnessing GPT-3.5 for\ntext parsing in solid-state synthesis \u2013\ncase study of ternary chalcogenides","volume":"3","author":"Thway","year":"2024","journal-title":"Digital\nDiscovery"},{"issue":"1","key":"2026091409065599500_cit13","doi-asserted-by":"publisher","first-page":"1569","DOI":"10.1038\/s41467-024-45914-8","article-title":"Extracting accurate materials data from research papers\nwith conversational language models and prompt engineering","volume":"15","author":"Polak","year":"2024","journal-title":"Nat. Commun."},{"issue":"16","key":"2026091409065599500_cit14","doi-asserted-by":"publisher","first-page":"8579","DOI":"10.1021\/acs.jcim.5c00840","article-title":"Autogenerating a Domain-Specific\nQuestion-Answering\nData Set from a Thermoelectric Materials Database to Enable High-Performing\nBERT Models","volume":"65","author":"Sierepeklis","year":"2025","journal-title":"J. Chem. Inf. Model."},{"key":"2026091409065599500_cit15","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1606.05250","article-title":"SQuAD: 100,000+ Questions for Machine\nComprehension of Text","author":"Rajpurkar","year":"2016","journal-title":"arXiv"},{"key":"2026091409065599500_cit16","doi-asserted-by":"crossref","first-page":"8342","DOI":"10.18653\/v1\/2020.acl-main.740","article-title":"Don\u2019t Stop\nPretraining: Adapt Language Models to Domains and Tasks","volume-title":"Proceedings of the 58th annual meeting of the association\nfor computational linguistics","author":"Gururangan","year":"2020"},{"issue":"10","key":"2026091409065599500_cit17","doi-asserted-by":"publisher","first-page":"1894","DOI":"10.1021\/acs.jcim.6b00207","article-title":"ChemDataExtractor: A Toolkit for\nAutomated Extraction\nof Chemical Information from the Scientific Literature","volume":"56","author":"Swain","year":"2016","journal-title":"J. Chem. Inf. Model."},{"issue":"9","key":"2026091409065599500_cit18","doi-asserted-by":"publisher","first-page":"4280","DOI":"10.1021\/acs.jcim.1c00446","article-title":"ChemDataExtractor\n2.0: Autopopulated Ontologies for Materials Science","volume":"61","author":"Mavrac\u030cic\u0301","year":"2021","journal-title":"J. Chem. Inf. Model."},{"issue":"5","key":"2026091409065599500_cit19","doi-asserted-by":"publisher","first-page":"1207","DOI":"10.1021\/acs.jcim.1c01199","article-title":"Single Model for\nOrganic and Inorganic Chemical Named\nEntity Recognition in ChemDataExtractor","volume":"62","author":"Isazawa","year":"2022","journal-title":"J.\nChem. Inf. Model."},{"issue":"4","key":"2026091409065599500_cit20","doi-asserted-by":"publisher","first-page":"998","DOI":"10.1039\/D4DD00307A","article-title":"Auto-generating\nquestion-answering datasets with domain-specific\nknowledge for language models in scientific tasks","volume":"4","author":"Li","year":"2025","journal-title":"Digital Discovery"},{"key":"2026091409065599500_cit21","article-title":"BERT:\nPre-training of Deep Bidirectional\nTransformers for Language Understanding","volume-title":"Proceedings of the 2019 conference of the North American chapter\nof the association for computational linguistics: human language technologies","author":"Devlin","year":"2018"},{"key":"2026091409065599500_cit22","article-title":"HuggingFace\u2019s Transformers: State-of-the-art Natural Language\nProcessing","volume-title":"Proceedings of the 2020 conference\non empirical methods in natural language processing: system demonstrations","author":"Wolf","year":"2020"},{"key":"2026091409065599500_cit23","doi-asserted-by":"publisher","first-page":"3505","DOI":"10.1145\/3394486.3406703","article-title":"August. Deepspeed: System optimizations\nenable training deep learning models with over 100 billion parameters","volume-title":"Proceedings of the 26th ACM SIGKDD international conference\non knowledge discovery & data mining","author":"Rasley","year":"2020"},{"key":"2026091409065599500_cit24","first-page":"784","article-title":"Know\nWhat You Don\u2019t Know: Unanswerable\nQuestions for SQuAD","volume-title":"Proceedings of the\n56th Annual Meeting of the Association for Computational Linguistics","author":"Rajpurkar","year":"2018"},{"issue":"1","key":"2026091409065599500_cit25","doi-asserted-by":"publisher","first-page":"192","DOI":"10.1038\/s41597-022-01295-5","article-title":"A database of refractive indices and dielectric constants\nauto-generated using ChemDataExtractor","volume":"9","author":"Zhao","year":"2022","journal-title":"Sci.\nData"},{"issue":"1","key":"2026091409065599500_cit26","doi-asserted-by":"publisher","first-page":"648","DOI":"10.1038\/s41597-022-01752-1","article-title":"A thermoelectric\nmaterials database auto-generated\nfrom the scientific literature using ChemDataExtractor","volume":"9","author":"Sierepeklis","year":"2022","journal-title":"Sci. Data"},{"issue":"1","key":"2026091409065599500_cit27","doi-asserted-by":"publisher","first-page":"260","DOI":"10.1038\/s41597-020-00602-2","article-title":"A database of battery materials auto-generated\nusing\nChemDataExtractor","volume":"7","author":"Huang","year":"2020","journal-title":"Sci. Data"},{"issue":"1","key":"2026091409065599500_cit28","doi-asserted-by":"publisher","first-page":"651","DOI":"10.1038\/s41597-023-02511-6","article-title":"Automated Construction\nof a Photocatalysis Dataset\nfor Water-Splitting Applications","volume":"10","author":"Isazawa","year":"2023","journal-title":"Sci. Data"},{"issue":"1","key":"2026091409065599500_cit29","doi-asserted-by":"publisher","first-page":"80","DOI":"10.1038\/s41597-023-02897-3","article-title":"A database of thermally\nactivated delayed fluorescent\nmolecules auto-generated from scientific literature with ChemDataExtractor","volume":"11","author":"Huang","year":"2024","journal-title":"Sci. Data"},{"issue":"1","key":"2026091409065599500_cit30","doi-asserted-by":"publisher","first-page":"193","DOI":"10.1038\/s41597-022-01294-6","article-title":"Auto-generated database of semiconductor\nband gaps\nusing ChemDataExtractor","volume":"9","author":"Dong","year":"2022","journal-title":"Sci. Data"},{"issue":"1","key":"2026091409065599500_cit31","doi-asserted-by":"publisher","first-page":"1273","DOI":"10.1038\/s41597-024-03979-6","article-title":"A Database\nof Stress-Strain Properties\nAuto-generated from the Scientific Literature using ChemDataExtractor","volume":"11","author":"Kumar","year":"2024","journal-title":"Sci. Data"},{"issue":"1","key":"2026091409065599500_cit32","doi-asserted-by":"publisher","first-page":"2000","DOI":"10.1038\/s41597-025-06244-6","article-title":"A dataset of Curie and Ne\u0301el\ntemperatures auto-generated\nwith ChemDataExtractor and the Snowball algorithm","volume":"12","author":"Kelly","year":"2025","journal-title":"Sci. Data"},{"issue":"22","key":"2026091409065599500_cit33","doi-asserted-by":"publisher","first-page":"7045","DOI":"10.1021\/acs.jcim.3c01281","article-title":"Snowball 2.0: Generic\nMaterial Data Parser for ChemDataExtractor","volume":"63","author":"Dong","year":"2023","journal-title":"J. Chem. Inf. Model."},{"key":"2026091409065599500_cit34","doi-asserted-by":"crossref","first-page":"85","DOI":"10.1145\/336597.336644","article-title":"Snowball: Extracting\nrelations from large plain-text collections","volume-title":"Proceedings of the fifth ACM conference on Digital libraries","author":"Agichtein","year":"2000"},{"key":"2026091409065599500_cit35","unstructured":"Kumar, P.\n          \n          QA-Annotator. GitHub\nrepository. https:\/\/github.com\/gh-PankajKumar\/QA-Annotator."},{"issue":"7","key":"2026091409065599500_cit36","doi-asserted-by":"publisher","first-page":"1596","DOI":"10.1039\/D0CE01857K","article-title":"Performances variations\nof BiFeO3-based ceramics induced\nby additives with diverse phase structures","volume":"23","author":"Shi","year":"2021","journal-title":"CrystEngComm."},{"key":"2026091409065599500_cit37","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2005.14165","article-title":"Language Models\nare Few-Shot Learners","author":"Brown","year":"2020","journal-title":"arXiv"},{"key":"2026091409065599500_cit38","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2303.08774","article-title":"GPT-4 Technical Report","author":"Achiam","year":"2024","journal-title":"arXiv"},{"key":"2026091409065599500_cit39","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2601.03267","article-title":"OpenAI GPT-5\nSystem Card","author":"Singh","year":"2025","journal-title":"arXiv"},{"key":"2026091409065599500_cit40","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2312.11805","article-title":"Gemini: A Family of Highly Capable\nMultimodal Models","author":"Team","year":"2025","journal-title":"arXiv"},{"key":"2026091409065599500_cit41","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2403.05530","article-title":"Gemini 1.5: Unlocking multimodal understanding\nacross millions of tokens of context","author":"Team","year":"2024","journal-title":"arXiv"},{"key":"2026091409065599500_cit42","article-title":"Gemini 2.5: Pushing the Frontier with\nAdvanced Reasoning, Multimodality, Long Context, and Next Generation\nAgentic Capabilities","author":"Comanici","year":"2025","journal-title":"arxiv"},{"key":"2026091409065599500_cit43","unstructured":"Gemini 3.1 Pro. https:\/\/deepmind.google\/models\/model-cards\/gemini-3-1-pro\/."},{"key":"2026091409065599500_cit44","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2302.13971","article-title":"LLaMA: Open\nand Efficient\nFoundation Language Models","author":"Touvron","year":"2023","journal-title":"arXiv"},{"key":"2026091409065599500_cit45","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2307.09288","article-title":"Llama 2: Open Foundation and\nFine-Tuned Chat Models","author":"Touvron","year":"2023","journal-title":"arXiv"},{"key":"2026091409065599500_cit46","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2407.21783","article-title":"The Llama 3 Herd\nof Models","author":"Grattafiori","year":"2024","journal-title":"arXiv"},{"issue":"2","key":"2026091409065599500_cit47","doi-asserted-by":"publisher","first-page":"021001","DOI":"10.1115\/1.4063843","article-title":"MechGPT,\na Language-Based Strategy for Mechanics and Materials Modeling That\nConnects Knowledge Across Scales, Disciplines, and Modalities","volume":"76","author":"Buehler","year":"2024","journal-title":"Appl. Mech. Rev."},{"key":"2026091409065599500_cit48","doi-asserted-by":"publisher","DOI":"10.1007\/978-0-387-76426-9","volume-title":"Atomistic\nModeling of Materials Failure","author":"Buehler","year":"2008"},{"key":"2026091409065599500_cit49","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2309.08788","article-title":"BioinspiredLLM:\nConversational Large Language Model for the Mechanics of Biological\nand Bio-inspired Materials","author":"Luu","year":"2023","journal-title":"arXiv"},{"key":"2026091409065599500_cit50","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2604.00069","article-title":"Perspective: Towards\nsustainable exploration of chemical spaces with machine learning","author":"Sandonas","year":"2026","journal-title":"arXiv"},{"key":"2026091409065599500_cit51","unstructured":"UN General Assembly.\n          \n          Transforming our world: the\n2030 Agenda for Sustainable Development; UN General Assembly. https:\/\/www.refworld.org\/legal\/resolution\/unga\/2015\/en\/111816."},{"key":"2026091409065599500_cit52","unstructured":"Kelly, C. R.\n          \n          MagBERT_from_Curie_Neel_Dataset. https:\/\/github.com\/C-R-Kelly\/MagBERT_from_Curie_Neel_Dataset."},{"key":"2026091409065599500_cit53","unstructured":"MagBERT Models. https:\/\/huggingface.co\/collections\/CambridgeMolecularEngineering\/magbert-models."}],"container-title":["Journal of Chemical Information\nand Modeling"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pubs.acs.org\/doi\/pdf\/10.1021\/acs.jcim.6c00136","content-type":"application\/pdf","content-version":"vor","intended-application":"unspecified"},{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article-pdf\/66\/17\/10896\/67515692\/acs.jcim.6c00136.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article-pdf\/66\/17\/10896\/67515692\/acs.jcim.6c00136.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,14]],"date-time":"2026-09-14T13:10:31Z","timestamp":1789391431000},"score":1,"resource":{"primary":{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article\/66\/17\/10896\/5339775\/Autogenerating-a-Domain-Specific-Question"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,27]]},"references-count":53,"journal-issue":{"issue":"17","published-online":{"date-parts":[[2026,8,27]]},"published-print":{"date-parts":[[2026,9,14]]}},"URL":"https:\/\/doi.org\/10.1021\/acs.jcim.6c00136","relation":{},"ISSN":["1549-9596","1549-960X"],"issn-type":[{"value":"1549-9596","type":"print"},{"value":"1549-960X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,27]]}}}