Current Large Audio Language Models largely transcribe rather than listen

(arxiv.org) by earcar | Feb 25, 2026 | 0 comments on HN