通过温度调控提升扩散语言模型采样多样性，兼顾效率与质量。
A Tale of Two Temperatures: Simple, Efficient, and Diverse Sampling from Diffusion Language Models
- 用温度调整信心重掩码策略，简单提升采样多样性。
- 在相同计算成本下，通过率（pass@NFE）超越传统方法。
- 适合需要多样化输出的下游任务和测试时计算扩展场景。
针对扩散语言模型（dLLMs）的采样速度与质量权衡已有大量研究，但样本间的多样性仍不明确。本文提出使用软化后的温度版本重掩码启发式方法，在保持原有计算优势的同时显著提升多样性。通过构建理想化的分叉标记形式模型，分析了重掩码对分叉点熵的影响。实验表明，该方法在控制计算成本（pass@NFE）的前提下，使探索差距（pass@k）接近自回归采样表现，优于现有信心驱动方法。进一步研究表明，多样性的提升有效增强了后训练及测试时计算扩展的下游效果。结果证明，实现高效、简单且多样化的扩散语言模型采样是可行的。
原文摘要 · Abstract (English)
Much work has been done on designing fast and accurate sampling for diffusion language models (dLLMs). However, these efforts have largely focused on the tradeoff between speed and quality of individual samples; how to additionally ensure diversity across samples remains less well understood. In this work, we show that diversity can be increased by using softened, tempered versions of familiar confidence-based remasking heuristics, retaining their computational benefits and offering simple implementations. We motivate this approach by introducing an idealized formal model of fork tokens and studying the impact of remasking on the expected entropy at the forks. Empirically, the proposed tempered heuristics close the exploration gap (pass@k) between existing confidence-based and autoregressive sampling, hence outperforming both when controlling for cost (pass@NFE). We further study how the increase in diversity translates to downstream post-training and test-time compute scaling. Overall, our findings demonstrate that simple, efficient, and diverse sampling from dLLMs is possible.
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。