2026-10-10T16:15:40Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33938661 2026-09-18T17:42:31Z category_7 category_45 category_734 category_106 portal_5 item_type_3 month_year_09_2026

<p>Accuracy (%) on medical licensing examination questions. N = 655 questions, no supporting context provided. 95% confidence interval is 4.1 percentage points.</p> Kiran Bhattacharyya (24534373) Sreeram Kamabattula (24534370) Medicine Sociology Biological Sciences not elsewhere classified Science Policy recommend generalizable configurations new prompting method inference — offers continuous reflection ” call “ prompt 7 percentage points larger model ’ smaller model ’ test baseline prompts practicing clinicians available meet accuracy requirements div >< p increased reasoning budgets offs across prompting time scaling </ time scaling model scales p </ generation budgets favor smaller medical reasoning technical skill statistically supported specific effects specific accuracy premise models overlapping 95 increasing accuracy improve domain general framework free alternative example use engineering llm empirical results efficiency objectives direct answer confidence interval clinical settings cases relevant based systems balance cost augmented generation assess accuracy also identify allowing us achieve accuracy accessibility concerns 3 27b <p>Accuracy (%) on medical licensing examination questions. N = 655 questions, no supporting context provided. 95% confidence interval is 4.1 percentage points.</p> 2026-09-18T18:00:05Z Dataset Dataset 10.1371/journal.pdig.0001182.t007 https://figshare.com/articles/dataset/_p_Accuracy_on_medical_licensing_examination_questions_N_655_questions_no_supporting_context_provided_95_confidence_interval_is_4_1_percentage_points_p_/33938661 CC BY 4.0