Scott McGrath @smcgrath.phd · Jun 17 Human clinicians often lean on unstructured text for care options, but the AI achieved significantly higher preciseness in drug dosing and test selection. It backed choices with explicit guideline citations by utilizing long-context lookups across hundreds of manual pages. #MedSky
0 likes 1 replies
?
Replies
Long-term disease management requires highly discriminant medication logic. To test this, the authors built RxQA, a 600-question drug formulary benchmark. While open-book tools helped both sides, the AI outperformed human doctors on the most complex, higher-difficulty cases. #MedSky