不训练模型，用推测解码加速数学推理，省算力还提准确率。
SSR: Speculative Parallel Scaling Reasoning in Test-time
- 在推理步骤级引入推测解码，用内部评分选优策略。
- MATH-500上算力降为30%仍保准确率，LiveMathBench提升13.84%。
- 无需训练，适合高效部署多步数学推理系统。
大型语言模型在多步数学推理任务中表现优异，但计算开销大。尤其对于测试时扩展方法如并行解码，虽提高答案多样性，却效率低下。为此，我们提出SSR（Speculative Parallel Scaling Reasoning），一种免训练框架，核心思路是在步骤级别引入推测解码，实现加速而不牺牲正确性。SSR包含两个组件：选择性并行模块（SPM），通过模型内评分识别少数有前景的推理策略；以及步骤级推测解码（SSD），支持细粒度推理中的草稿-目标协作。在三个数学基准测试——AIME 2024、MATH-500 和 LiveMathBench——上的实验表明，SSR显著优于基线。例如，在LiveMathBench上，pass@1准确率提升13.84%，计算量降至基线的80.5% FLOPs；在MATH-500上，计算量仅需30%且无准确率损失。
原文摘要 · Abstract (English)
Large language models (LLMs) have achieved impressive results on multi-step mathematical reasoning, yet at the cost of high computational overhead. This challenge is particularly acute for test-time scaling methods such as parallel decoding, which increase answer diversity but scale poorly in efficiency. To address this efficiency-accuracy trade-off, we propose SSR (Speculative Parallel Scaling Reasoning), a training-free framework that leverages a key insight: by introducing speculative decoding at the step level, we can accelerate reasoning without sacrificing correctness. SSR integrates two components: a Selective Parallel Module (SPM) that identifies a small set of promising reasoning strategies via model-internal scoring, and Step-level Speculative Decoding (SSD), which enables efficient draft-target collaboration for fine-grained reasoning acceleration. Experiments on three mathematical benchmarks-AIME 2024, MATH-500, and LiveMathBench - demonstrate that SSR achieves strong gains over baselines. For instance, on LiveMathBench, SSR improves pass@1 accuracy by 13.84% while reducing computation to 80.5% of the baseline FLOPs. On MATH-500, SSR reduces compute to only 30% with no loss in accuracy.
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。