提出新方法提升扩散语言模型的生成准确性。
Lookahead Path Likelihood Optimization for Diffusion LLMs
- 设计路径似然目标，引导全局最优解码路径选择。
- 在6个推理任务中平均提升2%~3%准确率，计算开销可控。
- 适合追求高精度生成且关注推理效率的研究者。
扩散大语言模型（dLLMs）支持任意顺序生成，但其推理性能高度依赖去掩码顺序。现有策略依赖启发式方法，仅局部优化置信度，难以找到全局一致且准确的去掩码路径。为此，我们提出路径似然（Path LL），一种与下游准确率强相关的轨迹条件目标，可指导最优路径选择。为在推理时优化路径似然，我们设计了POKE，一种高效的值估计器，用于预测部分解码轨迹的未来路径似然期望。进一步将该前瞻信号集成到基于序列蒙特卡洛（POKE-SMC）的搜索框架中，动态识别最优去掩码路径。在6个推理任务上的大量实验表明，POKE-SMC在保持相近推理开销的前提下，相比强基准方法平均提升2%~3%准确率，并推进了准确率-计算权衡前沿。
原文摘要 · Abstract (English)
Diffusion Large Language Models (dLLMs) support arbitrary-order generation, yet their inference performance critically depends on the unmasking order. Existing strategies rely on heuristics that greedily optimize local confidence, offering limited guidance for identifying unmasking paths that are globally consistent and accurate. To bridge this gap, we introduce path log-likelihood (Path LL), a trajectory-conditioned objective that strongly correlates with downstream accuracy and enables principled selection of unmasking paths. To optimize Path LL at inference time, we propose POKE, an efficient value estimator that predicts the expected future Path LL of a partial decoding trajectory. We then integrate this lookahead signal into POKE-SMC, a Sequential Monte Carlo-based search framework for dynamically identifying optimal unmasking paths. Extensive experiments across 6 reasoning tasks show that POKE-SMC consistently improves accuracy, achieving 2%--3% average gains over strong decoding-time scaling baselines at comparable inference overhead on LLaDA models and advancing the accuracy--compute Pareto frontier.
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。