Performance of large language models on neuroanatomy-based medical riddles: a comparative study

Key Points

• Large language models demonstrate significant performance variability in solving neuroanatomy riddles, highlighting their potential utility.
• The average correct response rate of the models was assessed, revealing a wide range in accuracy across different riddles.
• Comparative analysis across various models was conducted to identify strengths and weaknesses in solving medical riddles.
• Results may inform future developments in using language models for clinical applications in neuroanatomy.