提出无需训练的并行解码方法，显著加速扩散语言模型推理。
Accelerating Diffusion LLM Inference via Local Determinism Propagation
- 基于高置信锚点的局部确定性传播，实现精准并行解码
- 减少90%以上解码步骤，吞吐提升6.94倍，质量几乎不变
- 适合追求高效推理的部署者和研究者使用
扩散大语言模型（dLLMs）在文本生成中实现了并行词元解码，但现有开源实现存在质量与速度的权衡问题。保守采样策略通常每步仅解码最置信的词元以保证质量（即贪婪解码），导致重复冗余的细化迭代——我们称之为延迟解码。通过对dLLM解码动态的系统分析，我们刻画了这一现象，并提出一种无需训练的自适应并行解码策略LocalLeap，以解决效率瓶颈。LocalLeap基于两个经验原则：围绕高置信锚点的局部确定性传播，以及渐进的空间一致性衰减。通过应用这些原则，LocalLeap识别锚点，并在限定邻域内执行局部松弛并行解码，通过提前确定已明确的词元大幅减少解码步数，且不损害输出质量。在多个基准上的全面评估表明，LocalLeap实现6.94倍吞吐提升，解码步数降至原始需求的14.2%，性能影响可忽略。源代码见：https://github.com/friedrichor/LocalLeap。
原文摘要 · Abstract (English)
Diffusion large language models (dLLMs) represent a significant advancement in text generation, offering parallel token decoding capabilities. However, existing open-source implementations suffer from quality-speed trade-offs that impede their practical deployment. Conservative sampling strategies typically decode only the most confident token per step to ensure quality (i.e., greedy decoding), at the cost of inference efficiency due to repeated redundant refinement iterations--a phenomenon we term delayed decoding. Through systematic analysis of dLLM decoding dynamics, we characterize this delayed decoding behavior and propose a training-free adaptive parallel decoding strategy, named LocalLeap, to address these inefficiencies. LocalLeap is built on two fundamental empirical principles: local determinism propagation centered on high-confidence anchors and progressive spatial consistency decay. By applying these principles, LocalLeap identifies anchors and performs localized relaxed parallel decoding within bounded neighborhoods, achieving substantial inference step reduction through early commitment of already-determined tokens without compromising output quality. Comprehensive evaluation on various benchmarks demonstrates that LocalLeap achieves 6.94$\times$ throughput improvements and reduces decoding steps to just 14.2\% of the original requirement, achieving these gains with negligible performance impact. The source codes are available at: https://github.com/friedrichor/LocalLeap.
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。