YuE2: Unifying Symbolic and Audio Music Generation at Frontier Quality

The paper introduces YuE2, a model that unifies symbolic and audio music generation through symbolic planning using a single AR-NAR Mixture-of-Transformers (MoT). It writes a readable score, expands it into semantic music tokens, and realizes full-song audio. To learn from unaligned recordings, the authors introduce MERT2 and SheetSage2 for semantic and symbolic supervision. YuE2 achieves top scores on WildSongBench, matches proprietary song generators like Suno, and enables agentic music editing via external language models.

Experts prefer symbolic planning for overall quality and musicality, with 49.3% of overall preferences versus 34.6% without planning.

On WildSongBench, YuE2 scores 6.73 on SongBench Global Avg, exceeding all evaluated public baselines.

Selecting from eight candidates (best-of-8), YuE2 reaches 6.96, the highest observed mean among all evaluated systems.