一致性扩散语言模型：推理速度提升14倍且不损质量
原标题：Consistency diffusion language models: Up to 14x faster inference without sacrificing quality
AI 摘要
Together AI 的研究团队提出了一致性扩散语言模型（CDLM），通过结合基于一致性的多令牌定稿和块级 KV 缓存，将扩散语言模型的推理速度提升高达 14.5 倍，同时保持质量。该方法通过后训练配方，使用块因果注意力实现精确的 KV 缓存，并联合优化蒸馏损失、一致性损失和辅助去噪损失。实验表明，CDLM 在数学和编程任务上显著减少了推理步骤，并提升了吞吐量。
正文节选
We introduce consistency diffusion language models (CDLM), which accelerates diffusion language model inference by combining consistency-based multi-token finalization with block-wise KV caching, achieving up to 14.5x latency speedups on math and coding tasks. Diffusion Language Models (DLMs) are emerging as a promising alternative to autoregressive (AR) LMs. Instead of generating one token at a time, DLMs iteratively refine a partially masked sequence over multiple sampling steps, gradually tra