ChatGPT-4.0 Achieves 85% in Clinical Neurology Exam
The latest iteration of OpenAI‘s language model, ChatGPT-4.0, has achieved a milestone by scoring 85% in a clinical neurology exam, marking a significant leap in its capabilities. The experiment conducted by researchers from University Hospital Heidelberg and the German Cancer Research Center revealed promising potential for ChatGPT models in clinical neurology applications.
The test, based on questions from the American Board of Psychiatry and Neurology along with a subset from the European Board for Neurology, showcased the performance of both ChatGPT-3.5 and the newer ChatGPT-4.0. While the former scored 66.8%, the latter impressively answered 85% of the questions correctly.
Published on December 7, this proof-of-concept study underscores the evolving capabilities of language models in addressing clinical queries. Notably, ChatGPT-4.0 surpassed the average human score of 73.8%, particularly excelling in behavioral, cognitive, and psychological topics.
However, the study highlighted a limitation as both versions exhibited weaker performance in tasks requiring “higher-order thinking” than those demanding “lower-order thinking.” Yet, this achievement sets a precedent, suggesting future applications of language models in complex medical assessments, following refinements and enhancements.
