Jianwei Yu

4 papers Microsoft

• CoVoMix2: Advancing Zero-Shot Dialogue Generation with Fully Non-Autoregressive Flow Matching
• MMAR: A Challenging Benchmark for Deep Reasoning in Speech, Audio, Music, and Their Mix
• MoonCast: High-Quality Zero-Shot Podcast Generation
• SongBloom: Coherent Song Generation via Interleaved Autoregressive Sketching and Diffusion Refinement