{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,22]],"date-time":"2026-10-22T18:39:27Z","timestamp":1792694367034,"version":"4.5.0"},"reference-count":29,"publisher":"American Association for the Advancement of Science (AAAS)","issue":"6822","content-domain":{"domain":["www.science.org"],"crossmark-restriction":true},"short-container-title":["Science"],"published-print":{"date-parts":[[2026,10,22]]},"abstract":"<jats:p>\n                    We explored the capabilities of frontier language models in sequential diagnosis, a diagnostic paradigm that captures the iterative nature of clinical reasoning. Given a brief clinical presentation, clinicians selectively acquired information to reach a diagnosis, enabling assessment of diagnostic accuracy and workup cost. We developed an interactive benchmark and an artificial intelligence (AI) system that simulates a panel of physicians, continually revising the differential diagnosis and selecting cost-effective tests. Using this benchmark, the system achieved 83.2% accuracy on\n                    <jats:italic toggle=\"yes\">New England Journal of Medicine<\/jats:italic>\n                    clinicopathological conference cases and 80.3% on MIMIC-IV-ED emergency department admissions, outperforming generalist physicians (36.1 and 46.8%, respectively). Although these findings were obtained in a controlled benchmark setting, they highlight the promise of language models for supporting more accurate, lower-cost diagnoses.\n                  <\/jats:p>","DOI":"10.1126\/science.aeh7930","type":"journal-article","created":{"date-parts":[[2026,10,22]],"date-time":"2026-10-22T18:00:48Z","timestamp":1792692048000},"page":"423-428","update-policy":"https:\/\/doi.org\/10.34133\/aaas_crossmark","source":"Crossref","is-referenced-by-count":0,"title":["Sequential diagnosis with language models"],"prefix":"10.1126","volume":"394","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-5442-1359","authenticated-orcid":true,"given":"Harsha","family":"Nori","sequence":"first","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-6768-8225","authenticated-orcid":true,"given":"Mayank","family":"Daswani","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1246-844X","authenticated-orcid":true,"given":"Christopher","family":"Kelly","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6280-0941","authenticated-orcid":true,"given":"Scott","family":"Lundberg","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3301-1297","authenticated-orcid":true,"given":"Marco Tulio","family":"Ribeiro","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Marc","family":"Wilson","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1286-0038","authenticated-orcid":true,"given":"Xiaoxuan","family":"Liu","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Viknesh","family":"Sounderajah","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6618-3653","authenticated-orcid":true,"given":"Jonathan M.","family":"Carlson","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8591-5861","authenticated-orcid":true,"given":"Matthew P.","family":"Lungren","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bay","family":"Gross","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0419-2199","authenticated-orcid":true,"given":"Peter","family":"Hames","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mustafa","family":"Suleyman","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1898-842X","authenticated-orcid":true,"given":"Dominic","family":"King","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8823-0614","authenticated-orcid":true,"given":"Eric","family":"Horvitz","sequence":"additional","affiliation":[{"name":"Microsoft, Redmond, WA, USA."}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"221","reference":[{"key":"e_1_3_2_2_2","unstructured":"H. Nori Y. T. Lee S. Zhang D. Carignan R. Edgar N. Fusi N. King J. Larson Y. Li W. Liu R. Luo S. M. McKinney R. O. Ness H. Poon T. Qin N. Usuyama C. White E. Horvitz Can generalist foundation models outcompete special-purpose tuning? Case study in medicine. arXiv:2311.16452 [cs.CL] (2023)."},{"key":"e_1_3_2_3_2","unstructured":"H. Nori N. King S. M. McKinney D. Carignan E. Horvitz Capabilities of GPT-4 on medical challenge problems. arXiv:2303.13375 [cs.CL] (2023)."},{"key":"e_1_3_2_4_2","unstructured":"H. Nori N. Usuyama N. King S. M. McKinney X. Fernandes S. Zhang E. Horvitz From Medprompt to o1: Exploration of run-time strategies for medical challenge problems and beyond. arXiv:2411.03590 [cs.CL] (2024)."},{"key":"e_1_3_2_5_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-025-08869-4"},{"key":"e_1_3_2_6_2","doi-asserted-by":"publisher","DOI":"10.1001\/jamainternmed.2024.0295"},{"key":"e_1_3_2_7_2","doi-asserted-by":"publisher","DOI":"10.1001\/jamanetworkopen.2024.40969"},{"key":"e_1_3_2_8_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.adz4433"},{"key":"e_1_3_2_9_2","unstructured":"New England Journal of Medicine Case Records of the Massachusetts General Hospital (2026); https:\/\/www.nejm.org\/medical-articles\/case-records-of-the-massachusetts-general-hospital."},{"key":"e_1_3_2_10_2","unstructured":"A. Johnson L. Bulgarelli T. Pollard L. A. Celi R. Mark S. Horng MIMIC-IV-ED (version 2.2). PhysioNet (2023); https:\/\/doi.org\/10.13026\/5ntk-km72."},{"key":"e_1_3_2_11_2","unstructured":"Kaiser Permanente Washington \u201cCentral Hospital standard charges\u201d (2023); https:\/\/healthy.kaiserpermanente.org\/content\/dam\/kporg\/final\/documents\/health-plan-documents\/coverage-information\/910511770-central-hospital-standard-charges-wa-en.xlsx."},{"key":"e_1_3_2_12_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-026-02545-1"},{"key":"e_1_3_2_13_2","unstructured":"H. Nori et\u00a0al. SDBench code and data. Zenodo (2026); https:\/\/doi.org\/10.5281\/zenodo.21286121."},{"key":"e_1_3_2_14_2","unstructured":"R. K. Arora J. Wei R. S. Hicks P. Bowman J. Qui\u00f1onero-Candela F. Tsimpourlas M. Sharman M. Shah A. Vallone A. Beutel J. Heidecke K. Singhal HealthBench: Evaluating large language models towards improved human health. arXiv:2505.08775 [cs.CL] (2025)."},{"key":"e_1_3_2_15_2","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMcpc2300897"},{"key":"e_1_3_2_16_2","doi-asserted-by":"publisher","DOI":"10.1111\/j.1749-6632.1964.tb00043.x"},{"key":"e_1_3_2_17_2","doi-asserted-by":"publisher","DOI":"10.1016\/0010-4809(68)90016-5"},{"key":"e_1_3_2_18_2","first-page":"630","volume-title":"Proceedings of the First IEEE Conference on Artificial Intelligence Applications","author":"Horvitz E. J.","year":"1984","unstructured":"E. J. Horvitz, D. E. Heckerman, B. N. Nathwani, L. M. Fagan, \u201c Diagnostic Strategies in the Hypothesis-Directed PATHFINDER System\u201d in Proceedings of the First IEEE Conference on Artificial Intelligence Applications (IEEE Computer Society Press, 1984), pp. 630\u2013636."},{"key":"e_1_3_2_19_2","doi-asserted-by":"publisher","DOI":"10.1055\/s-0038-1634867"},{"key":"e_1_3_2_20_2","first-page":"250","volume-title":"Proceedings of the Thirteenth Conference on Uncertainty in Artificial Intelligence","author":"Horvitz E.","year":"1997","unstructured":"E. Horvitz, A. Seiver, \u201c Time-Critical Action: Representations and Application\u201d, in Proceedings of the Thirteenth Conference on Uncertainty in Artificial Intelligence (Morgan Kaufmann Publishers, 1997), pp. 250\u2013257."},{"key":"e_1_3_2_21_2","doi-asserted-by":"publisher","DOI":"10.1126\/science.130.3366.9"},{"key":"e_1_3_2_22_2","doi-asserted-by":"publisher","DOI":"10.1001\/jama.1968.03140380053012"},{"key":"e_1_3_2_23_2","doi-asserted-by":"crossref","unstructured":"S. Chakraborty E. Bisong S. Bhatt T. Wagner R. Elliott F. Mosconi \u201cBioMedBERT: A Pre-trained Biomedical Language Model for QA and IR\u201d in Proceedings of the 28th International Conference on Computational Linguistics D. Scott N. Bel C. Zong Eds. (International Committee on Computational Linguistics 2020) pp. 669\u2013679; https:\/\/aclanthology.org\/2020.coling-main.59.","DOI":"10.18653\/v1\/2020.coling-main.59"},{"key":"e_1_3_2_24_2","doi-asserted-by":"publisher","DOI":"10.1145\/3458754"},{"key":"e_1_3_2_25_2","doi-asserted-by":"publisher","DOI":"10.2196\/45312"},{"key":"e_1_3_2_26_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-023-43436-9"},{"key":"e_1_3_2_27_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06291-2"},{"key":"e_1_3_2_28_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-025-04151-2"},{"key":"e_1_3_2_29_2","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0908"},{"key":"e_1_3_2_30_2","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMcpc2412526"}],"container-title":["Science"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.science.org\/doi\/pdf\/10.1126\/science.aeh7930","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,22]],"date-time":"2026-10-22T18:01:24Z","timestamp":1792692084000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.science.org\/doi\/10.1126\/science.aeh7930"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,22]]},"references-count":29,"journal-issue":{"issue":"6822","published-print":{"date-parts":[[2026,10,22]]}},"alternative-id":["10.1126\/science.aeh7930"],"URL":"https:\/\/doi.org\/10.1126\/science.aeh7930","relation":{},"ISSN":["0036-8075","1095-9203"],"issn-type":[{"value":"0036-8075","type":"print"},{"value":"1095-9203","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,22]]},"assertion":[{"value":"2026-04-04","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-08-10","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-10-22","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}