Clinical reasoning study finds gaps beneath high average AI scores
An evaluation of 21 language models found weaker performance when listing possible diagnoses, even when overall accuracy appeared high.
Tag: Clinical AI
Breast imaging pilot finds GPT-4 closer to guideline recommendations
A breast imaging pilot compared GPT-4 and GPT-3.5 with ACR recommendations, revealing differences by clinical scenario and question format.
ChatGPT’s accuracy varied across the steps of a clinical case
A 36-case study found that ChatGPT performed differently across diagnosis, testing and management, with implications for clinical AI evaluation.