2 ms·Current Large Audio Language Models largely transcribe rather than listen1 points by PaulHoule 8mo ago