Do Language Models Need Music Supervision? Verifiable Rewards for Multi-Constraint Symbolic Music Generation.

arXiv:2609.23665v3 Announce Type: replace-cross Abstract: Language models can now generate symbolic music from natural-language instructions. However, existing models still struggle to satisfy multiple explicit musical constraints simultaneously. To systematic