Masked diffusion language models (MDLMs) promise fast, non-autoregressive text generation, yet existing samplers, which pick tokens to unmask based on model confidence, ignore interactions when unmasking multiple positions in parallel and effectively reduce to slow, autoregressive behavior. We propose the Dilated Unmasking Scheduler (DUS), an inference-only, planner-model-free method that partitions sequence positions into non-adjacent dilated groups and unmasks them in parallel so as to minimize an upper bound on joint entropy gain at each denoising step. By explicitly trading off the number of network calls against generation quality, DUS recovers most of the performance lost under traditional parallel unmasking strategies. Across math (GSM8K, MATH500), code (HumanEval, MBPP), general-knowledge (BBH, MMLU-Pro), and instruction following (IFEval) benchmarks, DUS outperforms confidence-based planners and turns the diffusion-specific quality-speed trade-off into a deterministic, predictable speedup set by the block size $B$, yielding up to $5.8\times$ wall-clock speedup over token-by-token MDLM decoding without modifying the underlying denoiser. Applied as a drop-in post-filter, dilated spacing also improves adaptive samplers. Code is available at https://github.com/omerlux/DUS.

翻译：掩码扩散语言模型（MDLM）承诺实现快速、非自回归的文本生成，然而现有采样器根据模型置信度选择要解掩码的令牌，忽视了在并行解掩码多个位置时的交互作用，实际上退化为缓慢的自回归行为。我们提出膨胀解掩码调度器（DUS），这是一种仅推理、无需规划器（planner-model-free）的方法，它将序列位置划分为非相邻的膨胀组并并行解掩码，从而在每个去噪步骤中最小化联合熵增益的上界。通过明确地在网络调用次数与生成质量之间进行权衡，DUS 恢复了在传统并行解掩码策略下损失的大部分性能。在数学（GSM8K、MATH500）、代码（HumanEval、MBPP）、通用知识（BBH、MMLU-Pro）和指令遵循（IFEval）基准测试中，DUS 优于基于置信度的规划器，并将扩散特化的质量-速度权衡转化为由块大小 $B$ 确定的确定性、可预测的加速，在无需修改底层去噪器的情况下，相比逐令牌 MDLM 解码实现了高达 $5.8\times$ 的挂钟时间加速。作为即插即用的后滤波器，膨胀间距还能改进自适应采样器。代码可在 https://github.com/omerlux/DUS 获取。