用树形结构并行生成候选文本，显著提升扩散模型的解码速度。
DARTree: Speculative Diffusion Decoding with Autoregressive Draft Trees
- 构建多分支候选树，批量扩展与评分节点，实现并行生成
- 单轮验证可接受最多12.97个词元，比现有方法高98.6%
- 无需训练，适配多种模型与温度设置，加速效果突出
推测解码通过并行验证多个候选词元无损加速自回归语言模型。基于扩散的生成器可并行预测一整块词元，但其位置分布为边缘分布，未考虑沿每条路径选择的条件信息。现有递归修正仅沿单一路径传递因果信息，而扩散树结构虽扩大候选覆盖范围，却未能将修正传播至各分支。我们提出 DARTree，一种无需训练的推测解码方法，将预训练自回归修正头从链式结构扩展至树形结构。DARTree 首先在单批次内展开并评分每一层的所有节点，构建固定宽度的候选树；随后仅采用最佳优先剪枝选出验证树，从而将自回归头推理与串行堆操作解耦。在七个数学、代码和聊天基准上，DARTree 在所有四种模型-温度配置下均达到最高平均接受长度与加速比，单轮验证最多可接受12.97个词元，比DFlash高出98.6%，比Domino高出27.9%；最高速度提升达9.73×，且为无损加速。
原文摘要 · Abstract (English)
Speculative decoding losslessly accelerates autoregressive language models by verifying multiple draft tokens in parallel. Diffusion-based drafters further reduce proposal latency by predicting an entire token block in parallel, but their position-wise distributions are marginal rather than conditioned on tokens selected along each draft path. Existing recurrent correction incorporates causal information along a single draft chain, whereas diffusion-based tree construction broadens candidate coverage without carrying this correction along individual branches. We introduce DARTree, a training-free speculative decoding method that extends a pretrained AR correction head from chains to trees. DARTree first constructs a fixed-width candidate tree by expanding and scoring all nodes at each depth in a single batch, and then only applies best-first pruning to select the verification tree, decoupling AR-head inference from sequential heap operations. Across seven math, code, and chat benchmarks, DARTree achieves the highest average acceptance length and speedup in all four model--temperature configurations, accepting up to 12.97 tokens per verification round, 98.6\% more than DFlash and 27.9\% more than Domino in the same setting, and reaching up to 9.73$\times$ lossless speedup over locally measured autoregressive decoding.
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。