1 Hr
Source:
المشهد
AI records bring caution
A study by researchers at the Carlos III University of Madrid evaluated AI models for simplifying medical reports, amid evidence that 46% of European citizens struggle to understand medical documents. The test covered 50 medical reports and 150 simplified versions. Llama-3.2-3B delivered balanced performance, while RigoChat-v2-7B stood out for readability. However, the models made errors, including omitting clinical details, inventing or changing information, and producing poorly connected text—making human oversight essential when using them in healthcare.
Size: 120%Text Size
Notice: This Is an AI-Generated Summary
Comments (0)