拒绝混合：面向高效 DLLM 推理的掩码词元快速语义传播
Rejection Mixing: Fast Semantic Propagation of Mask Tokens for Efficient DLLM Inference
ReMix 的连续混合状态与拒绝回退机制可迁移到视觉语言模型的掩码扩散解码中，用于缓解并行解码时的组合矛盾，实现免训练加速且不损失质量。
原文依据 ReMix thus mitigates combinatorial contradictions by enabling continuous-space refinement during discrete diffusion decoding.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论扩散大语言模型（DLLM）有望实现快速非自回归推理，但在并行解码中面临严重的质量与速度权衡。这源于“组合矛盾”现象，即并行词元形成语义不一致的组合。我们通过将连续表示集成到离散解码过程中来解决这一问题，因为连续表示保留了丰富的跨位置依赖关系。我们提出了 ReMix（拒绝混合）框架，引入一种新颖的连续混合状态，作为初始掩码状态与最终解码词元状态之间的中间状态。该中间状态允许词元的表示在连续空间中迭代细化，在坍缩为最终离散样本之前解决与其他词元的相互冲突。此外，拒绝规则将不确定的表示从连续状态回退到掩码状态以重新处理，确保稳定性并防止错误传播。因此，ReMix 通过在离散扩散解码过程中实现连续空间细化来缓解组合矛盾。大量实验表明，ReMix 作为一种免训练方法，在没有任何质量下降的情况下实现了 2-8 倍的推理加速。
英文原文摘要 1254 字符
Diffusion Large Language Models (DLLMs) promise fast non-autoregressive inference but suffer a severe quality and speed tradeoff in parallel decoding. This stems from the "combinatorial contradiction" phenomenon, where parallel tokens form semantically inconsistent combinations. We address this by integrating continuous representations into the discrete decoding process, as they preserve rich inter-position dependency. We propose ReMix (Rejection Mixing), a framework that introduces a novel Continuous Mixing State as an intermediate between the initial masked state and the final decoded token state. This intermediate state allows a token's representation to be iteratively refined in a continuous space, resolving mutual conflicts with other tokens before collapsing into a final discrete sample. Furthermore, a rejection rule reverts uncertain representations from the continuous state back to the masked state for reprocessing, ensuring stability and preventing error propagation.ReMix thus mitigates combinatorial contradictions by enabling continuous-space refinement during discrete diffusion decoding. Extensive experiments demonstrate that ReMix, as a training-free method, achieves a 2-8xinference speedup without any quality degradation.