📄 Multimodal Latent Language Modeling with Next-Token Diffusion ✅ 7.2/10 | 前50% | arxiv ← 返回 2026-05-23 语音/音乐/音频论文速递