PulseAugur
中
实时 01:30:58
English(EN) Who Should Lead Decoding Now? Tracking Reliable Trajectories for Ensembling Masked Diffusion Language Models

新方法通过改进填充和知识集成来增强MDLM · 追踪6个来源

作者 PulseAugur 编辑部 · [5 个来源] ·

研究人员为掩码扩散语言模型（MDLM）引入了两种新颖的方法。第一种，VoidPadding，将序列结束（[EOS]）标记的语义终止和填充作用解耦，使用新的[VOID]标记进行填充。据报道，该方法在Dream-7B-Instruct等基准测试中提高了性能，并降低了解码效率低下问题。第二种方法，TIE（基于轨迹的迭代集成），侧重于整合来自多个MDLM的知识。TIE通过追踪置信度动态来识别可靠的解码轨迹，并选择性地在模型之间转移中间状态，以在生成过程中利用它们互补的优势。 AI

影响 引入了提高MDLM生成效率和知识融合能力的方法。

排序理由 两篇研究论文介绍了掩码扩散语言模型的新颖技术。

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →

新方法通过改进填充和知识集成来增强MDLM · 追踪6个来源

报道来源 [5]

  1. arXiv cs.CL TIER_1 English(EN) · Chunyu Liu, Zhengyang Fan, Kaisen Yang, Alex Lamb ·

    VoidPadding：让 [VOID] 处理 Masked Diffusion 语言模型中的填充，以便 [EOS] 能够专注于语义终止

    arXiv:2606.17999v1 Announce Type: new Abstract: MDLMs generate text by denoising a preallocated masked response canvas, making response-length modeling central to instruction tuning. Existing MDLMs often inherit the autoregressive convention of using repeated \texttt{[EOS]} token…

  2. arXiv cs.CL TIER_1 English(EN) · Alex Lamb ·

    VoidPadding：让 [VOID] 处理 Masked Diffusion Language Models 中的 Padding，以便 [EOS] 能够专注于语义终止

    MDLMs generate text by denoising a preallocated masked response canvas, making response-length modeling central to instruction tuning. Existing MDLMs often inherit the autoregressive convention of using repeated \texttt{[EOS]} tokens for padding during instruction tuning, giving …

  3. arXiv cs.AI TIER_1 English(EN) · Heecheol Yun, Joonhyung Park, Joowon Kim, Eunho Yang ·

    谁应领导当前的解码工作？追踪集成掩码扩散语言模型的可靠轨迹

    arXiv:2606.16281v1 Announce Type: cross Abstract: Masked Diffusion Language Models (MDLMs) have emerged as a distinct paradigm for sequence generation. As MDLMs become diverse in capabilities and knowledge coverage, an important question is how to combine their knowledge. Toward …

  4. arXiv cs.CL TIER_1 English(EN) · Eunho Yang ·

    谁应领导解码？追踪集成掩码扩散语言模型的可靠轨迹

    Masked Diffusion Language Models (MDLMs) have emerged as a distinct paradigm for sequence generation. As MDLMs become diverse in capabilities and knowledge coverage, an important question is how to combine their knowledge. Toward this, we first investigate the unique decoding dyn…

  5. Hugging Face Daily Papers TIER_1 English(EN) ·

    谁应领导解码？追踪集成掩码扩散语言模型的可靠轨迹

    Masked diffusion language models exhibit unique decoding dynamics where reliable trajectories show stable confidence patterns, enabling iterative ensemble methods that transfer partially denoised sequences between models based on confidence evolution.

相关实体

相关话题