arXiv cs.AI· Zhihan Yang, Wei Guo, Shuibai Zhang, Subham Sekhar Sahoo, Yongxin Chen, Arash Vahdat, Morteza Mardani, John Thickstun·· 4 小时前AI 评分47
连续扩散在语言建模上可媲美离散扩散：RePlaid 首次给出连续 DLM scaling law
Continuous Diffusion Scales Competitively with Discrete Diffusion for Language
AI 导读
研究者对 Plaid 进行架构对齐构建出连续扩散语言模型 RePlaid，并建立了首个可媲美离散扩散模型的连续 DLM scaling law：RePlaid 与自回归模型的算力差距仅 20×，参数更少却优于 Duo，并在过训练区间超过 MDLM。
正文 · 原文
Abstract:While diffusion has drawn considerable recent attention from the language modeling community, continuous diffusion has appeared less scalable than discrete approaches. To challenge this belief we revisit Plaid, a likelihood-based continuous diffusion language model (DLM), and construct RePlaid by aligning the architecture of Plaid with modern discrete DLMs. In this unified setting, we establish the first scaling law for continuous DLMs that rivals discrete DLMs: RePlaid exhibits a compute gap of only $20\times$ compared to autoregressive models, outperforms Duo while using fewer parameters, and outperforms MDLM in the over-trained regime. We benchmark RePlaid against recent continuous DLMs: on OpenWebText, RePlaid achieves a new state-of-the-art PPL bound of $22.1$ among continuous DLMs and superior generation quality. These results suggest that continuous diffusion, when trained via likelihood, is a highly competitive and scalable alternative to discrete DLMs. Moreover, we offer theoretical insights to understand the advantage of likelihood-based training. We show that optimizing the noise schedule to minimize the ELBO's variance naturally yields linear cross-entropy (information loss) over time. This evenly distributes denoising difficulty without any case-specific time reparameterization. In addition, we find that optimizing embeddings via likelihood creates structured geometries and drives the most significant likelihood gain.
| Comments: | NeurIPS 2026 |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Machine Learning (stat.ML) |
| Cite as: | arXiv:2605.18530 [cs.CL] |
| (or arXiv:2605.18530v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.18530 arXiv-issued DOI via DataCite |
Submission history
From: Zhihan Yang [view email]
[v1]
Mon, 18 May 2026 15:15:24 UTC (657 KB)
[v2]
Wed, 9 Sep 2026 21:20:41 UTC (658 KB)
[v3]
Sat, 24 Oct 2026 20:05:07 UTC (1,670 KB)
来源：arXiv cs.AI · arxiv.org