AI RESEARCH

DuoTok: Source-Aware Dual-Track Tokenization for Multi-Track Music Language Modeling

arXiv CS.AI

ArXi:2511.20224v2 Announce Type: replace-cross Audio tokenization bridges continuous waveforms and multi-track music language models. In dual-track modeling, tokens should preserve three properties at once: high-fidelity reconstruction, strong predictability under a language model, and cross-track correspondence. We