FlowBlock: Wavefront-Parallel Decoding for Self-Correcting Diffusion Language Models
2026-07-20 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors study how to speed up block-wise diffusion large language models, which normally decode text one block at a time in a strict order. They introduce FlowBlock, a method that allows blocks to be decoded in parallel without needing retraining, by using a token-to-token editing process to fix drafts made with slightly outdated context. This approach improves decoding speed significantly and also boosts accuracy compared to previous methods. Their system manages multiple requests efficiently by grouping blocks into wavefronts and ensuring consistent reuse of cached information.
diffusion large language modelsblock-wise decodingKV-cachetoken-to-token editingparallel decodingwavefront decodingbatched processinglatencyaccuracy
Authors
Bing Tian, Haikun Liu, Xiaocheng Zhong, Zhuohui Duan, Zhaokai Luo, Huayi Jin, Zhiyong Wang, Xiaofei Liao
Abstract
Block-wise diffusion large language models (dLLMs) decode sequentially at the block level, enabling effective KV-cache reuse across blocks but making inter-block decoding strictly serial. Prior work has attempted to unlock inter-block parallelism through post-training methods, but achieves only modest speedups and often degrades accuracy. We observe that self-correcting dLLMs offer a training-free alternative: token-to-token (T2T) editing can repair tokens drafted with a slightly stale upstream context, so a downstream block requires only an informative draft rather than a finalized predecessor. This turns block finality from a hard dependency into a scheduling resource. We propose \textbf{\flowblock{}}, a training-free parallel decoding framework built on two mechanisms. (i) \emph{Gated Wavefront Decoding} admits blocks into a bounded wavefront only when a readiness gate is satisfied, jointly refines active blocks via T2T editing, and commits blocks in order under a windowed block-causal mask that preserves exact frozen-prefix KV caches reuse. (ii) \emph{Heterogeneous Wavefront Packing} assigns each request an independent wavefront while packing asynchronous windows into dense, shape-stable batched forwards. Across different benchmarks, \flowblock{} improves tokens per second (TPS) over LLaDA-2.1 and LLaDA-2.0, two serial block-wise dLLMs, by up to 2.95$\times$ and 4.01$\times$, while reducing latency by up to 53.6\% and 77.1\%, respectively. It also improves average accuracy by 1.3 points. Compared with D2F, a training-based inter-block-parallel baseline, \flowblock{} achieves higher accuracy and up to 16$\times$ higher batched serving throughput.