Reproduce “Local Multimodal Music Alignment from Global Supervision” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Irmak Bukey, Zachary Novack, Jongmin Jung, Dasaem Jeong, Chris Donahue
FuSiLi aligns score patches with audio frames before pooling, using pairwise similarities and a Sinkhorn-derived soft alignment to form its contrastive score. It is trained on globally paired score-and-audio segments, without local alignment labels in the training loss. On MSMD, its best batch configuration reaches 30% top-1 frame-alignment accuracy, versus 16% for a global-only baseline, measured against note-derived patch labels.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.