arXiv cs.CL· Guangfu Guo, Xiaoqian Lu, Linsey Pang, Weiran Yao, Haolin Chen, Long Cheng, Kunpeng Liu·· 11 小时前AI 评分31
RangeSWA：面向连续扩散语言模型的无训练长度外推方法
RangeSWA: Training-Free Length Extrapolation for Continuous Diffusion Language Models
AI 导读
RangeSWA 是一种无需训练的长度外推方法，用于解决连续扩散语言模型在生成长度超过训练长度时出现的文本重复、语义漂移和质量下降问题。该方法结合局部滑动窗口注意力、远端块摘要与动态内容检索，并通过位置重映射、条件注意力修正和去噪阶段自适应窗口调度缓解外推带来的生成退化。在多个连续扩散语言模型上的实验显示，RangeSWA 能有效降低长文本生成的生成困惑度和文本重复率。
正文 · 原文
Abstract:Continuous diffusion language models show promising potential in text generation tasks, but are prone to problems such as text repetition, semantic drift, and generation quality degradation when the generation length exceeds the training length. To address these challenges, we propose RangeSWA, a training-free length extrapolation method. The method combines local sliding-window attention, remote block summaries, and dynamic content retrieval to preserve local details and support long-range context access. In addition, generation degradation caused by length extrapolation is mitigated through positional remapping, conditioning attention correction, and denoising-stage-adaptive window scheduling. We conducted experiments on multiple continuous diffusion language models, and the results show that RangeSWA can effectively reduce generative perplexity and text repetition rates in long-text generation.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.42084 [cs.CL] |
| (or arXiv:2610.42084v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.42084 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Guangfu Guo [view email]
[v1]
Fri, 30 Oct 2026 17:35:48 UTC (236 KB)
来源：arXiv cs.CL · arxiv.org