新智元 (微信公众号)T2·Models
72 pts
Future Doctor conducted a multicenter study with top Chinese thoracic surgery departments, testing its medical LLM MedGPT against 250 tertiary-hospital physicians on 40 unpublished complex lung cancer cases. MedGPT averaged 85.29 versus 53.26 for unaided doctors, winning 88 to 12 in head-to-head comparisons while producing answers over 40% shorter. The evaluation used 240 expert-defined scoring rules, including 40 safety red lines, echoing Microsoft's earlier Science diagnostic study.