Recently, diffusion models have garnered significant interest in the field of text processing due to their many potential advantages compared to conventional autoregressive models. In this work, we propose Diffusion-of-Thought (DoT), a novel approach that integrates diffusion models with Chain-of-Thought, a well-established technique for improving the reasoning ability of autoregressive language models. In contrast to autoregressive language models that make decisions in a left-to-right, token-by-token manner, DoT allows reasoning steps to diffuse over time through a diffusion language model and offers greater flexibility in trading-off computation for reasoning performance. Our experimental results demonstrate the effectiveness of DoT in multi-digit multiplication, boolean logic, and grade school math problems, with a small diffusion model outperforming a much larger autoregressive model in both efficiency and accuracy. In addition to that, DoT showcases promising self-correction abilities and benefits from existing reasoning-enhancing techniques like self-consistency decoding. Our findings contribute to the understanding and development of reasoning with diffusion language models.

翻译：近年来，扩散模型因其相较于传统自回归模型的诸多潜在优势，在文本处理领域引起了广泛关注。本文提出思维扩散（DoT），这是一种将扩散模型与思维链（一种用于提升自回归语言模型推理能力的成熟技术）相结合的新方法。与以从左到右、逐词方式决策的自回归语言模型不同，DoT允许推理步骤通过扩散语言模型随时间扩散，并在计算量与推理性能的权衡上提供了更大的灵活性。我们的实验结果表明，DoT在多位数乘法、布尔逻辑和小学数学问题上具有有效性，其中一个小型扩散模型在效率和准确性上均优于一个规模大得多的自回归模型。此外，DoT展现出有前景的自我纠正能力，并能受益于自洽解码等现有的推理增强技术。我们的发现有助于理解和开发基于扩散语言模型的推理。