STEMMA: Song-to-Stem Multi-Audio Reasoning for Large Audio Language Models
Paper • 2610.11884 • Published
STEMMA
Song-to-Stem Multi-Audio Reasoning for Large Audio Language Models: models, training data, and benchmark.
Note Paper
hoyso48/STEMMA-MF2601
Audio-Text-to-Text • 8B • Updated • 43Note Music Flamingo 2601 fine-tuned on STEMMA-Instruct
hoyso48/STEMMA-MOSS-Music
Audio-Text-to-Text • 9B • Updated • 49Note MOSS-Music fine-tuned on STEMMA-Instruct
hoyso48/STEMMA-Instruct
Viewer • Updated • 326k • 69Note Multi-audio instruction data (public corpora)
hoyso48/STEMMA-Bench
Viewer • Updated • 300 • 38Note Multi-audio song/section/stem relation benchmark