用自回归模型生成的计划引导扩散模型，显著提升多步推理能力。
Think First, Diffuse Fast: Improving Diffusion Language Model Reasoning via Autoregressive Plan Conditioning
- 先用自回归模型生成简短计划，再作为全局上下文输入扩散模型。
- GSM8K上准确率从75.6%提升至87.2%，代码任务提升12.8个百分点。
- 对扩散模型效果远超传统模型，适合需要稳定推理的场景。
扩散大语言模型通过迭代去噪生成文本，但在多步推理任务中表现持续不佳。我们假设这一差距源于协调问题：自回归模型逐词构建连贯性，而扩散模型需同时协调所有位置。为此提出计划条件化方法——在扩散模型提示前添加一个来自自回归模型的短（约100词）自然语言计划。该计划作为固定支架，使每个位置从首个去噪步骤起即可全局可见并关注。在GSM8K上，计划条件化将LLaDA-8B-Instruct准确率从75.6%提升至87.2%（+11.6个百分点），接近同规模自回归模型（LLaMA 3.1 8B，87.7%），尽管基线弱6.4个百分点。在HumanEval上提升12.8个百分点（37.2%→50.0%），表明计划具有泛化能力。相同计划在传统扩散模型上的增益仅为+5.7pp（GSM8K）和+1.3pp（HumanEval），扩散模型受益2–10倍，支持协调问题假说。五次随机种子下，计划条件化在GSM8K上的准确率标准差为零，推理极稳定。消融实验显示模型遵循计划策略（错误策略导致-16.3pp），但对计划数值鲁棒（扰动后仅-1.1pp）；规划器质量存在陡峭阈值：小规模模型计划损害（-1.6至-6.8pp），前沿计划才能带来全部提升。注意力分析确认机制：早期去噪阶段，计划词获得1.8倍超额注意力，随完成词固化逐渐均匀。计划条件化每次推理成本约$0.002，延迟增加约2秒。
原文摘要 · Abstract (English)
Diffusion large language models (dLLMs) generate text via iterative denoising but consistently underperform on multi-step reasoning. We hypothesize this gap stems from a coordination problem: AR models build coherence token-by-token, while diffusion models must coordinate all positions simultaneously. We propose plan conditioning, a training-free method that prepends a short (~100-token) natural-language plan from an AR model to the diffusion model's prompt. The plan serves as a frozen scaffold -- globally visible context that every token position can attend to from the first denoising step. On GSM8K, plan conditioning improves LLaDA-8B-Instruct from 75.6% to 87.2% (+11.6 percentage points), matching a same-size AR model (LLaMA 3.1 8B, 87.7%) despite a 6.4pp weaker baseline. On HumanEval, the gain is +12.8pp (37.2% to 50.0%), showing plans generalize to code. The same plans improve LLaMA by only +5.7pp on GSM8K and +1.3pp on HumanEval -- diffusion models benefit 2-10x more, supporting the coordination-problem hypothesis. Across 5 random seeds, plan-conditioned GSM8K accuracy has zero standard deviation, making diffusion inference highly stable. Ablations reveal the model follows plan strategy (wrong-strategy plans cause -16.3pp) but is robust to plan values (perturbed numbers: -1.1pp), and that planner quality has a sharp threshold: smaller Llama-class plans hurt (-1.6 to -6.8pp) while frontier plans provide the full lift. Attention analysis confirms the mechanism: plan tokens receive 1.8x excess attention during early denoising, declining to uniform as completion tokens solidify. Plan conditioning costs ~$0.002 per problem and adds ~2s of latency.
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。