Current Large Audio Language Models largely transcribe rather than listen

(arxiv.org) by PaulHoule | Mar 9, 2026 | 0 comments on HN