ChatGPT achieved excellent results in the neurology professional level test

In a recent cross-sectional study, researchers examined the performance of large language models (LLMs) on neurology board-type tests.

The study, which used a question bank approved by the American Board of Psychiatry and Neurology, revealed insights into these advanced language models.

ChatGPT administers the neurological exam

The study included two LL.M. ChatGPT includes versions — 3.5 and version 4. The findings show that LLM 2 is far superior to its predecessor. In addition, he even surpassed the average human score on the Neurology Board Exam.

According to the findings, LLM 2 answered 85.0% of questions correctly. Meanwhile, the average result of the person is 73.8%.

This data suggests that, with further improvements, large-scale language models may find significant applications in clinical neurology and healthcare.

Read more: 9 ChatGPT questions and tips for crafting the perfect job description

ChatGPT works best on low-order test questions

However, even the old model LLM 1 performed well enough, albeit slightly below the human average, scoring 66.8%.

Both models use the language of confidence regardless of the accuracy of their answers, indicating areas for improvement for future iterations.

Based on the study, questions were classified into low-level and high-level based on Bloom's Taxonomy.

Both models performed better on low-order queries. However, LLM 2 excelled on both low- and high-order questions, demonstrating its versatility and cognitive abilities.

Read more: ChatGPT vs. Google Bard: Comparison of AI Chatbots