CoChat
Sign Up
Cited by 79Artificial Intelligence in Healthcare and EducationMultimodal Machine Learning ApplicationsAI in cancer detection

Large language models and multimodal foundation models for precision oncology

Open access
Truhn, Daniel, Eckardt, Jan-Niklas, Ferber, Dyke, Kather, Jakob Nikolas·2024·npj Precision Oncology
DOI 10.1038/s41698-024-00573-2PMC PMC10959931
Open this paper in CoChat — free accountOpen PDF

Abstract

The technological progress in artificial intelligence (AI) has massively accelerated since 2022, with far-reaching implications for oncology and cancer research. Large language models (LLMs) now perform at human-level competency in text processing. Notably, both text and image processing networks are increasingly based on transformer neural networks. This convergence enables the development of multimodal AI models that take diverse types of data as an input simultaneously, marking a qualitative shift from specialized niche models which were prevalent in the 2010s. This editorial summarizes these developments, which are expected to impact precision oncology in the coming years.

Verified 2026-10-07 · 1 reference not checked yet

AI snapshot

Written from the paper’s full text.

TL;DR

This editorial summarizes how large language models (LLMs) and multimodal foundation models are expected to affect precision oncology, enabled by transformer architectures and expanding patient data. It highlights promising applications alongside challenges in data quality, clinical integration, regulation, evaluation, and interpretability.

Key findings

  • LLMs now perform at human-level competency in text processing, and transformer neural networks increasingly underlie both text and image processing, enabling multimodal AI models that take diverse data types simultaneously.
  • Medical adaptations include Bio-BERT for biomedical text mining, Med-PaLM, and Med-PaLM 2, which scored 86.5% on the US Medical Licensing Exam (USMLE).
  • Foundation models can reduce data requirements for specialized tasks; for example, linking chest X-rays to report text can alleviate manual annotation while preserving human-level accuracies and outperforming supervised methods.
  • The paper identifies challenges for oncology adoption: careful assessment of training data quality, quantity, and diversity; expert-guided system design; legal and regulatory approval as medical devices; ongoing evaluation; and limited interpretability, especially for text or multimodal tasks.

Methods

  • This is an editorial/narrative overview, not a primary research study.
  • It synthesizes cited literature and examples across oncology, medical imaging, pathology, EHRs, and foundation models.
  • No systematic search, dataset, or statistical analysis is described.

Limitations

  • The paper states that more scientific evidence is required to ensure LLMs and multimodal models provide quantifiable benefits in oncology.
  • It states that data quality, quantity, and diversity must be carefully assessed; integration should involve medical professionals, patient advocates, and the broader scientific community; regulatory approval is required; ongoing evaluation is needed; and interpretability remains a concern, with fewer explainability studies for text or multimodal medical tasks.
  • Evident from the text: as an editorial, it provides no original empirical validation or quantitative synthesis, and several cited examples are preprints.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41698-024-00573-2

Cites

29 of 30 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 28 cited sources checked · 0 supported · 0 flagged

  • Premisecitation

    Multimodal AI systems have been evaluated for various applications in precision oncology, such as outcome predictions.

    “Multimodal AI systems have been evaluated for various applications in precision oncology, such as outcome predictions14,15.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [14] Chen, R. J. et al. Pan-cancer integrative histology-genomic analysis via multimodal deep learning. Cancer Cell 40, 865–878.e6 (2022). source

      Not checked in the free run.

    • cites [15] Lipkova, J. et al. Artificial intelligence for multimodal data integration in oncology. Cancer Cell 40, 1095–1110 (2022). source

      Not checked in the free run.

  • Premisecitation

    Foundation models reduce the data requirements for specialized tasks, for example in predicting diseases from retinal photographs.

    “Foundation models reduce the data requirements for specialized tasks, for example in predicting diseases from retinal photographs17.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [17] Zhou, Y. et al. A foundation model for generalizable disease detection from retinal images. Nature https://doi.org/10.1038/s41586-023-06555-x (2023). source

      Not checked in the free run.

  • Premisecitation

    Fine-tuned LLMs have been shown to solve real-world problems such as predicting clinical outcomes just based on unstructured text data in EHRs.

    “However, fine-tuned LLMs have been shown to solve real-world problems such as predicting clinical outcomes just based on unstructured text data in EHRs13.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [13] Jiang, L. Y. et al. Health system-scale language models are all-purpose prediction engines. Nature 619, 357–362 (2023). source

      Not checked in the free run.

  • Premisecitation

    Between 2012 and 2022, neural networks were applied in numerous studies primarily focused on oncological imaging or text.

    “Between 2012 and 2022, neural networks were applied in numerous studies that primarily focused on the analysis of oncological imaging or text1.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [1] Shmatko, A., Ghaffari Laleh, N., Gerstung, M. & Kather, J. N. Artificial intelligence in histopathology: enhancing cancer research and clinical oncology. Nat. Cancer 3, 1026–1038 (2022). source

      Not checked in the free run.

  • Premisecitation

    The integration of foundation models in operable clinical software systems faces legal and regulatory challenges because these models require approval as medical devices.

    “Thirdly, the integration of such models in operable clinical software systems faces legal and regulatory challenges because these models require approval as medical devices26.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [26] Gilbert, S., Harvey, H., Melvin, T., Vollebregt, E. & Wicks, P. Large language model AI chatbots require approval as medical devices. Nat. Med. https://doi.org/10.1038/s41591-023-02412-6 (2023). source

      Not checked in the free run.

  • Classified as probably not a substantial claim. 22
  • May combine two claims. 9

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Model performance scales with size (number of parameters), and larger models show emergent behavior: they acquire an understanding of concepts underlying training data without explicit training.

    “Notably, model performance scales with size, i.e. number of parameters, and larger models show emergent behavior9: They acquire an understanding of concepts underlying the training data without having been explicitly trained on this.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [9] Wei, J. et al. Emergent abilities of large language models. Preprint at https://arxiv.org/abs/2206.07682 (2022).

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The technological shift to convolutional neural networks was subsequently integrated into medical research.

    “This technological shift was subsequently integrated into medical research1”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [1] Shmatko, A., Ghaffari Laleh, N., Gerstung, M. & Kather, J. N. Artificial intelligence in histopathology: enhancing cancer research and clinical oncology. Nat. Cancer 3, 1026–1038 (2022). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    The transformer neural network architecture enables multimodality.

    “This network architecture is well suited for almost any type of data, and enables multimodality.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Today’s LLMs are transformer neural networks.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    The transformer neural network architecture is well suited for almost any type of data.

    “This network architecture is well suited for almost any type of data, and enables multimodality.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Today’s LLMs are transformer neural networks.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    High-profile initiatives such as IBM Watson did not achieve their projected outcomes.

    “However, high-profile and ambitious initiatives, such as IBM Watson, did not achieve their projected outcomes4.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [4] Schmidt, C. M. D. Anderson breaks with IBM Watson, raising questions about artificial intelligence in oncology. J. Natl Cancer Inst. 109 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    A common application area under exploration for LLMs is healthcare.

    “A common application area under exploration for these LLMs is healthcare7.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [7] Bubeck, S. et al. Sparks of artificial general intelligence: early experiments with GPT-4. Preprint at https://arxiv.org/abs/2303.12712 (2023).

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Foundation models can alleviate the need for time-consuming manual annotation while preserving human-level accuracies and outperforming supervised methods.

    “For instance, linking images from chest X-rays to corresponding report text data, and foundation models can alleviate the need for time-consuming and laborious manual annotation while preserving human-level accuracies and outperforming supervised methods18.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [18] Tiu, E. et al. Expert-level detection of pathologies from unannotated chest X-ray images via self-supervised learning. Nat. Biomed. Eng. 6, 1399–1406 (2022). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Med-PaLM 2 scored 86.5% in the US Medical Licensing Exam (USMLE).

    “Recently, Google has introduced the next iteration, Med-PaLM 2, which scored a high 86.5% in the US Medical Licensing Exam (USMLE)12.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [12] Singhal, K. et al. Towards expert-level medical question answering with large language models. Preprint at https://arxiv.org/pdf/2305.09617.pdf (2023).

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Bio-BERT showed robust capabilities for biomedical text mining.

    “One of the first language models to be successfully transferred to the medical domain was Bio-BERT, which showed robust capabilities for biomedical text mining10.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [10] Lee, J. et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics 36, 1234–1240 (2020). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    More scientific evidence is required to ensure that LLMs and multimodal models provide quantifiable benefits in oncology.

    “However, more scientific evidence is required to ensure that LLMs and multimodal models provide quantifiable benefits in oncology.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The landscape of AI in oncology changed in 2022 and 2023 with the advent of large language models and multimodal AI models.

    “This landscape changed in 2022 and 2023 with the advent of two key innovations: large language models (LLMs) and multimodal AI models.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    GPT-4 displayed much-improved knowledge retrieval and logical reasoning capacities with far fewer hallucinations compared to GPT-3.5.

    “It was later succeeded by GPT-4, which displayed much-improved knowledge retrieval and logical reasoning capacities with far fewer hallucinations7.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [7] Bubeck, S. et al. Sparks of artificial general intelligence: early experiments with GPT-4. Preprint at https://arxiv.org/abs/2303.12712 (2023).

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The training data for LLMs is a large and diverse amount of text, usually sourced from the internet and commercial data providers, and can include diverse types of medical data.

    “The training data for these models is a large and diverse amount of text, usually sourced from the internet and commercial data providers, and can include diverse types of medical data7.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [7] Bubeck, S. et al. Sparks of artificial general intelligence: early experiments with GPT-4. Preprint at https://arxiv.org/abs/2303.12712 (2023).

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    The decade 2012-2022 can be viewed as a stabilization phase for AI applications in oncology, marked by incremental improvements and specialized uses, predominantly in image analysis.

    “This decade can be viewed as a stabilization phase for AI applications in oncology, marked by incremental improvements and specialized uses, predominantly in image analysis5.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5] He, J. et al. The practical implementation of artificial intelligence technologies in medicine. Nat. Med. 25, 30–36 (2019). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    A 2017 publication demonstrated neural network performance on par with human experts across large image datasets.

    “most notably evidenced by a 2017 publication demonstrating neural network performance on par with human experts across large image datasets2”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [2] Esteva, A. et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature 542, 115–118 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The most successful LLMs have recently relied on transformer-based architectures pertaining to their attention mechanisms.

    “While potentially this purpose can also be fulfilled by various model architectures, the most successful models have recently relied on transformer-based architectures pertaining to their attention mechanisms8.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [8] Vaswani, A. et al. Attention is all you need. Preprint at https://arxiv.org/abs/1706.03762 (2017).

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Med-PaLM outperformed its previous version in medical use cases.

    “Also, Google’s LLM “PALM” was fine-tuned on medical training data, resulting in Med-PaLM, which outperformed its previous version in medical use cases11.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [11] Singhal, K. et al. Large language models encode clinical knowledge. Nature 620, 172–180 (2023). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Hardware improvements have gradually lowered computational barriers for training and deploying resource-intensive models, broadening the user base capable of developing and refining AI algorithms.

    “In parallel, hardware improvements have gradually lowered the computational barriers for training and deploying resource-intensive models, thus broadening the user base capable of developing and refining AI algorithms3.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [3] Topol, E. J. High-performance medicine: the convergence of human and artificial intelligence. Nat. Med. 25, 44–56 (2019). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    The convergence of text and image processing networks on transformer architectures enables the development of multimodal AI models that take diverse data types as input simultaneously, marking a qualitative shift from specialized niche models prevalent in the 2010s.

    “This convergence enables the development of multimodal AI models that take diverse types of data as an input simultaneously, marking a qualitative shift from specialized niche models which were prevalent in the 2010s.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: LLMs now perform at human-level competency in text processing.
    • Rests on an assumption: Both text and image processing networks are increasingly based on transformer neural networks.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Overall, the advancements in LLMs and multimodal models have the potential to impact the practice of oncology through many different applications.

    “Overall, the advancements in LLMs and multimodal models have the potential to impact the practice of oncology through many different applications (Fig. 1C).”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The technological progress in AI has massively accelerated since 2022.
    • Rests on an assumption: LLMs now perform at human-level competency in text processing.
    • Rests on an assumption: Both text and image processing networks are increasingly based on transformer neural networks.
    • Rests on an assumption: The volume of patient-specific data in oncology is rapidly expanding due to the widespread introduction of electronic health records, advances in medical imaging, and the integration of large-scale genomic analyses into clinical routine.
    • Rests on an assumption: AI and machine learning have shown promise in helping healthcare professionals to process large oncology data.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    As foundation models diversify their capabilities, they open up new avenues for their potential application in oncology and cancer research, such as multimodal diagnostics and drug discovery.

    “As foundation models diversify their capabilities, they open up new avenues for their potential application in oncology and cancer research, such as multimodal diagnostics and drug discovery.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    A prominent concern of AI models based on neural network architecture is their often criticized lack of interpretability, earning them the term 'black boxes'.

    “Lastly, a prominent concern of AI models based on neural network architecture is their often criticized lack of interpretability which earned them the term ‘black boxes’27,28.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [27] Castelvecchi, D. Can we open the black box of AI? Nature 538, 20–23 (2016). source

      Not checked in the free run.

    • cites [28] Savage, N. Breaking into the black box of artificial intelligence. Nature https://doi.org/10.1038/d41586-022-00858-1 (2022). source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    While substantial progress has been made in model explainability for image-related tasks, fewer studies address explainability in text processing or multimodal tasks in medicine.

    “While substantial progress has been made in model explainability for image-related tasks, fewer studies address explainability in text processing or multimodal tasks in medicine29.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [29] Tjoa, E. & Guan, C. A survey on explainable artificial intelligence (XAI): toward medical XAI. IEEE Trans. Neural Netw. Learn Syst. 32, 4793–4813 (2021). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Early generalist models show consistently high performance across a variety of medical domains and tasks integrating knowledge from diverse domains.

    “Furthermore, early generalist models have been introduced which show consistently high performance across a variety of medical domains and tasks integrating knowledge from diverse domains22,23.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [22] Zhang, K. et al. BiomedGPT: a unified and generalist biomedical generative pre-trained transformer for vision, language, and multimodal tasks. Preprint at https://arxiv.org/abs/2305.17100 (2023).

      Not checked in the free run.

    • cites [23] Tu, T. et al. Towards generalist biomedical AI. Preprint at https://ai.nejm.org/doi/full/10.1056/AIoa2300138 (2023).

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    In pathology, large image datasets linked with contextual knowledge and case-specific information yield high performance in disease detection and biomarker prediction and can inform further diagnostic procedures such as additional stains.

    “Similar examples exist in pathology, where large image datasets are linked with contextual knowledge and case-specific information yielding high performance in disease detection as well as biomarker prediction and can also inform further diagnostic procedures such as additional stains20,21.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [20] Vorontsov, E. et al. Virchow: a million-slide digital pathology foundation model. Preprint at https://arxiv.org/abs/2309.07778 (2023).

      Not checked in the free run.

    • cites [21] Lu, M. Y. et al. A foundational multimodal vision language AI assistant for human pathology. Preprint at https://arxiv.org/abs/2312.07814 (2023).

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    There is a trend towards improving model efficiency by reducing model size while retaining model performance, given the resource-intensive nature of billion-parameter models.

    “Given the resource-intensive nature of training models with parameter sizes in the billions, there is a trend towards improving model efficiency by reducing model size while retaining model performance.”
    Quote in the paperNot a substantial claim
26 of 26

  • The paper relies on this without support.

    The technological progress in AI has massively accelerated since 2022.

    “The technological progress in artificial intelligence (AI) has massively accelerated since 2022, with far-reaching implications for oncology and cancer research.”

  • The paper relies on this without support.

    LLMs now perform at human-level competency in text processing.

    “Large language models (LLMs) now perform at human-level competency in text processing.”

  • The paper relies on this without support.

    Both text and image processing networks are increasingly based on transformer neural networks.

    “Notably, both text and image processing networks are increasingly based on transformer neural networks.”

  • The paper relies on this without support.

    The volume of patient-specific data in oncology is rapidly expanding due to the widespread introduction of electronic health records, advances in medical imaging, and the integration of large-scale genomic analyses into clinical routine.

    “The volume of patient-specific data in oncology is rapidly expanding. This is due to the widespread introduction of electronic health records (EHRs), advances in medical imaging, and the integration of large-scale genomic analyses into clinical routine.”

  • The paper relies on this without support.

    AI and machine learning have shown promise in helping healthcare professionals to process large oncology data.

    “Artificial intelligence (AI) and machine learning (ML) have shown promise in helping healthcare professionals to process such data.”

  • The paper relies on this without support.

    Today’s LLMs are transformer neural networks.

    “Today’s LLMs are transformer neural networks.”
6 of 6

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The predominance of pan-cancer clinical domains suggests that researchers leverage the versatility of LLMs to address broad oncological challenges across multiple cancer types, likely due to the generalizable nature of these models for various cancer data.

    “This suggests that researchers leverage the versatility of LLMs to address broad oncological challenges across multiple cancer types, likely due to the generalizable nature of these models for various cancer data [42].”

    From Large Language Model Applications for Health Information Extraction in Oncology: Scoping Review (2025) · cites this paper as [42]

Cites

Flagged first, then by citations

Research works printed but not linked: 3

  1. Vaswani, A. et al. Attention is all you need. Preprint at https://arxiv.org/abs/1706.03762 (2017). · arXiv
  2. Wei, J. et al. Emergent abilities of large language models. Preprint at https://arxiv.org/abs/2206.07682 (2022). · arXiv
  3. Zhang, K. et al. BiomedGPT: a unified and generalist biomedical generative pre-trained transformer for vision, language, and multimodal tasks. Preprint at https://arxiv.org/abs/2305.17100 (2023). · arXiv

Cited by

Sorted by citations

Related

Authors
Truhn, D., Eckardt, J.-N., Ferber, D., & Kather, J. N.
Affiliations
  • Universitätsklinikum Aachen (DE)
  • University Hospital Carl Gustav Carus (DE)
  • Heidelberg University (DE)
Corresponding author
Kather, Jakob Nikolas
ORCID
Venue
npj Precision Oncology
Publisher
Springer Science and Business Media LLC
Published
2024-03-22
Volume
8
Issue
1
Pages
72-72
ISSN
2397-768X (ISSN-L)
DOI
10.1038/s41698-024-00573-2
PMC
PMC10959931
OpenAlex
W4393078503
Semantic Scholar
7e9cd89173fa68a5bd8cc5cb7aa68df172f603a2
Cited by
79
References
30
Citations by year
2026: 37 · 2025: 25 · 2024: 17
Type
journal article
Language
en
Open access
Licence
Locations
  • npj Precision Oncology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PUBLISSO (German National Library of Medicine) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
  • RWTH Publications (RWTH Aachen) (submittedVersion, other-oa)
  • RWTH Publications (RWTH Aachen)
Funders
  • Amgen
  • AstraZeneca
Topics
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
  • Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
Keywords
large language models, multimodal foundation models, precision oncology, cancer research, transformer models, multimodal AI, artificial intelligence in oncology

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-07.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-07.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review