Researchers explore the potential of mid-sized language models for clinical QA tasks. Large language models have performance and sustainability problems, but on-device AI offers a promising solution. Two types of models are applicable in a biomedical context, and their performance in clinical QA applications is still uncertain. A team of researchers conducted an evaluation of four models using popular tasks in the clinical QA domain. The top-performing model showed potential for clinical question-answering tasks, but the question of whether larger biomedical specialty models would outperform it remains open.
32 Impressions