September 10, 2025PLoS ONEOpen Access

Evaluation of large language models as a diagnostic tool for medical learners and clinicians using advanced prompting techniques

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Observational analysis shows diagnostic accuracy of LLMs using advanced prompting techniques, highlighting challenges with proprietary models.

Llama-3.1-70b-versatile achieved 79% correct responses, indicating its efficacy in clinical diagnostics.
Using advanced prompting strategies resulted in better accuracy metrics compared to basic prompting methods.
Analysis included sensitivity and specificity, surpassing the performance of proprietary models like GPT-3.5.
The findings suggest open-source models may offer more reliable and accessible alternatives for medical diagnostics.

Gaebe et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Observational analysis shows diagnostic accuracy of LLMs using advanced prompting techniques, highlighting challenges with proprietary models.