Multi-model ensembles raised factual accuracy by 8% — and product marketing still has to learn how to check claims
https://wiki-cafe.win/index.php/23_Document_Formats_from_One_AI_Conversation
Multi-model ensembles lifted factual accuracy by 8% in internal benchmarks The data suggests a clear but modest win: in a series of internal evaluations, combining outputs from several models produced an average 8% absolute increase in