Assessing the robustness and responsiveness of large-scale frontier models in medical AI applications

Singhal, K. et al. Large-scale language models encode clinical knowledge. nature 620172–180 (2023).

Article CAS PubMed PubMed Central Google Scholar

Gu, Y. et al. Pre-training domain-specific language models for biomedical natural language processing. in ACM Transactions on Computing for Healthcare (HEALTH) (Editors Lee, I. & Stankovic, JA) 31−23 (Association for Computing Machinery, 2022).

Nori, H. Sequential diagnosis using language models. Preprint available at https://arxiv.org/abs/2506.22405 (2025).

Open AI. Introducing GPT-5. https://openai.com/index/introducing-gpt-5/ (2025).

Saab, K. et al. Functions of the Gemini model in medicine. Preprint available at https://arxiv.org/abs/2404.18416 (2024).

Tu, T. et al. Towards conversational diagnostic AI. Preprint available at https://arxiv.org/abs/2401.05654 (2024).

Wang, S. et al. LINS: A general medical Q&A framework to improve the quality and reliability of LLM-generated responses. nut. common. 169076 (2025).

Article CAS PubMed PubMed Central Google Scholar

Arora, RK et al. HealthBench: Evaluating large-scale language models for improving human health. Preprint available at https://arxiv.org/abs/2505.08775 (2025).

Handler, R., Sharma, S. & Hernandez-Boussard, T. The fragile intelligence of GPT-5 in medicine. nut. medicine. 313968–3970 (2025).

Article CAS PubMed Google Scholar

Farquhar, S., Kossen, J., Kuhn, L. & Gal, Y. Hallucination detection in large-scale language models using semantic entropy. nature 630625–630 (2024).

Article CAS PubMed PubMed Central Google Scholar

Jin, Q. et al. Hidden flaws behind the expert-level accuracy of multimodal GPT-4 vision in healthcare. NPJ digit. medicine. 7190 (2024).

Article PubMed PubMed Central Google Scholar

Pfau, J., Merrill, W. & Bowman, S.R. Thinking point by point: Hidden computation in transformer language models. in 1st Conference on Language Modeling (COLM) https://openreview.net/forum?id=NikbrdtYvG (2024).

Gayhos, R. et al. Shortcut learning in deep neural networks. nut. Mach. intelligence. 2665–673 (2020).

Article Google Scholar

Acosta, JN, Falcone, GJ, Rajpurkar, P. & Topol, EJ Multimodal biomedical AI. nut. medicine. 281773–1784 (2022).

Article CAS PubMed Google Scholar

Goodfellow, IJ, Shlens, J. & Szegedy, C. Explaining and leveraging adversarial examples. Preprint available at https://arxiv.org/abs/1412.6572 (2015).

Szegedi, C. et al. Interesting properties of neural networks. Preprint available at https://arxiv.org/abs/1312.6199 (2013).

New England Medical Journal: Image Challenge. https://www.nejm.org/image-challenge (2026).

JAMA Network Clinical Issues. https://jamanetwork.com/collections/44038/clinical-challenge (2026).

Comanici, G. et al. Gemini 2.5: Pushing the frontiers with advanced reasoning, multimodality, long context, and next-generation agent capabilities. Preprint available at https://arxiv.org/abs/2507.06261 (2025).

Human. Claude 3.5 Sonnet. https://www.anthropic.com/news/claude-3-5-sonnet (2024).

Open AI. GPT-4o system card. https://openai.com/index/gpt-4o-system-card/ (2024).

Open AI. OpenAI o3 and o4-mini system cards. https://openai.com/index/o3-o4-mini-system-card/ (2025).

Wei, J. et al. Thought chain prompts draw inferences in large-scale language models. in NIPS’22: Proceedings of the 36th International Conference on Neural Information Processing Systems 24824−24837 (Koyejo, S. et al., eds.) (Curran Associates, 2022).

Lau, JJ, Gayen, S., Ben Abacha, A., Demner-Fushman, D. A dataset of clinically generated visual questions and answers about radiology images. Science. data 5180251 (2018).

Article PubMed PubMed Central Google Scholar

Hu, Y. et al. OmniMedVQA: A new large-scale comprehensive evaluation benchmark for medical LVLM. in 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) https://doi.org/10.1109/CVPR52733.2024.02093 (IEEE, 2024).

Johnson, AE et al. MIMIC-CXR is an anonymized public chest radiograph database with free text reports. Science. data 6317 (2019).

Article PubMed PubMed Central Google Scholar

He, X., Zhang, Y., Mou, L., Xing, E., Xie, P. PathVQA: 30,000+ Questions for Medical Visual Question Answering. Preprint available at https://arxiv.org/abs/2003.10286 (2020).

Liu, B. et al. SLAKE: A semantically labeled knowledge enrichment dataset for medical visual question answering. Preprint at https://arxiv.org/abs/2102.09542 (2021).

Zhang, X. et al. PMC-VQA: Visual instruction tuning for medical visual question answering. Preprint available at https://arxiv.org/abs/2305.10415 (2023).

Yue, X. et al. MMMU: A large-scale, multidisciplinary, multimodal understanding and reasoning benchmark for expert AGI. in 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) https://doi.org/10.1109/CVPR52733.2024.00913 (IEEE, 2024).

Fleiss, JL Measuring the agreement of nominal scales between many raters. Psychol. Bull. 76378–382 (1971).

Article Google Scholar

Wu, Z. et al. DeepSeek-VL2: An expert mixed visual language model for advanced multimodal understanding. Preprint available at https://arxiv.org/abs/2412.10302 (2024).

Bai, S. et al. Qwen3-VL Technical Report. Preprint available at https://arxiv.org/abs/2511.21631 (2025).

Lee, C. et al. LLaVA-Med: Training large-scale language and visual assistants for biomedicine in one day. in NIPS ʼ23: Proceedings of the 37th International Conference on Neural Information Processing Systems (Oh, A. et al. eds.) 28541−28564 (Curran Associates, 2023).

Sellergren, A. et al. MedGemma Technical Report. Preprint available at https://arxiv.org/abs/2507.05201 (2025).