BeClaude
Research2026-05-06

MedMosaic: A Challenging Large Scale Benchmark of Diverse Medical Audio

Source: Arxiv CS.AI

arXiv:2605.00969v1 Announce Type: cross Abstract: We present MedMosaic, a medical audio question-answering dataset designed to benchmark language and audio reasoning models under realistic clinical constraints. Medical audio data is difficult to collect due to privacy regulations and high...

arxivpapersbenchmark