解析扩散语言模型推理加速的三类关键技术，助力高效部署。
Accelerating Masked Diffusion Large Language Models: A Survey of Efficient Inference Techniques
- 提出统一延迟分解框架，拆解算法、架构与系统因素影响。
- 归纳三类加速技术：算法创新、架构优化与推理扩展。
- 提供可复现基准建议，揭示并行生成的潜力与挑战。
扩散大语言模型（dLLMs）在理论上具有并行生成的优势，超越标准自回归模型。然而，仅靠并行生成无法保证实际速度提升，需依赖特定推理机制如扩散感知缓存与重用。随着推理效率成为实际部署的前提，近期研究在算法、架构与系统层面积极探索加速技术。但严格对比仍困难，因端到端延迟涉及算法、架构与系统级因素的复杂权衡，常被现有基准混淆。本文提出一种统一的延迟分解框架，解耦各因素对真实部署中推理速度的影响。基于此框架，我们从算法创新、架构与系统优化、推理时扩展三个维度分类加速技术。最后，提供可复现基准的指导，并指出实现并行生成全部潜力的开放挑战。
原文摘要 · Abstract (English)
Diffusion large language models (dLLMs) offer a theoretical advantage in parallel generation over standard autoregressive models. However, parallel generation alone does not guarantee practical speedups. Realizing this efficiency requires specialized inference mechanisms, such as diffusion-aware caching and reuse. Consequently, as inference efficiency becomes a prerequisite for practical deployment, recent research has actively explored acceleration techniques across algorithms, architectures, and systems. However, rigorous comparisons remain difficult, as end-to-end latency stems from intricate trade-offs between algorithmic, architectural, and system-level factors that are often conflated in existing benchmarks. In this survey, we introduce a unified latency decomposition framework for dLLMs to disentangle these factors and analyze their impact on inference speed in real deployments. Guided by this framework, we categorize acceleration techniques along three axes covering algorithmic innovations, architectural and system optimizations, and inference-time scaling. Finally, we provide guidelines for reproducible benchmarking and highlight open challenges for realizing the full potential of parallel generation.
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。