用轻量编辑修复扩散语言模型并行生成的不一致问题
Edit-Based Refinement for Parallel Masked Diffusion Language Models
- 通过替换、删除、插入等编辑操作优化完整生成结果
- 在LLaDA上实现HumanEval+11.6、GSM8K+33.6得分提升
- 仅需1/8扩散步数，兼顾效率与生成质量
掩码扩散语言模型支持并行词元生成，相比自回归模型具有更高的解码效率。然而，其在同时生成多个词元时性能显著下降，原因在于词元级训练目标与整体序列一致性之间存在不匹配。本文提出ME-DLM，一种基于编辑的精炼框架，在扩散生成后引入轻量级后编辑步骤。生成完整响应后，模型根据全序列条件进行最小编辑操作（替换、删除、插入），训练监督来自编辑距离，在固定规范化方案下提供确定性信号以学习最小修正。该方法通过全局条件编辑促进序列一致性，同时保持并行扩散解码的效率优势。大量实验表明，ME-DLM显著提升了多词元并行生成的质量与鲁棒性。特别地，在LLaDA基础上，本方法在HumanEval上获得11.6分提升，在GSM8K上获得33.6分提升，且仅使用八分之一的总扩散步数。代码已开源。
原文摘要 · Abstract (English)
Masked diffusion language models enable parallel token generation and offer improved decoding efficiency over autoregressive models. However, their performance degrades significantly when generating multiple tokens simultaneously, due to a mismatch between token-level training objectives and joint sequence consistency. In this paper, we propose ME-DLM, an edit-based refinement framework that augments diffusion generation with lightweight post-editing steps. After producing an initial complete response, the model refines it through minimal edit operations, including replacement, deletion, and insertion, conditioned on the full sequence. Training supervision is derived from edit distance, providing a deterministic signal under a fixed canonicalization scheme for learning minimal corrections. This approach encourages sequence-level consistency through globally conditioned edits while preserving the efficiency benefits of parallel diffusion decoding. Extensive experiments demonstrate that ME-DLM improves the quality and robustness of multi-token parallel generation. In particular, when built upon LLaDA, our method achieves consistent gains of 11.6 points on HumanEval and 33.6 points on GSM8K while using one-eighth of the total diffusion steps. Code is available at https://github.com/renhouxing/ME-DLM.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。