通过双粒度压缩让大模型推理更省资源且更准确
CtrlCoT: Dual-Granularity Chain-of-Thought Compression for Controllable Reasoning
- 分层抽象与逻辑保真剪枝结合，兼顾语义压缩与关键信息保留
- 在MATH-500上减少30.7%令牌数，准确率提升7.6个百分点
- 适合需要高效可控推理的场景，如教育、金融问答
链式思维（CoT）提示能提升大模型推理能力，但冗长的推理过程导致高延迟和高内存开销。现有方法或仅在语义层面简化，过于保守；或激进地删除标记，可能遗漏关键线索并降低准确率。此外，两者结合面临序列依赖、任务无关剪枝和分布不匹配等挑战。本文提出CtrlCoT，一种双粒度压缩框架，包含三个组件：层次化推理抽象生成多粒度语义层级的推理路径；逻辑保真蒸馏训练一个关注逻辑的剪枝器，确保在不同剪枝率下保留关键推理线索（如数字、运算符）；分布对齐生成使压缩后的推理痕迹与推理时流畅风格一致，避免碎片化。在MATH-500数据集上使用Qwen2.5-7B-Instruct模型，CtrlCoT以30.7%更少的令牌数，实现比最强基线高出7.6个百分点的准确率，证明其推理更高效可靠。代码将开源。
原文摘要 · Abstract (English)
Chain-of-thought (CoT) prompting improves LLM reasoning but incurs high latency and memory cost due to verbose traces, motivating CoT compression with preserved correctness. Existing methods either shorten CoTs at the semantic level, which is often conservative, or prune tokens aggressively, which can miss task-critical cues and degrade accuracy. Moreover, combining the two is non-trivial due to sequential dependency, task-agnostic pruning, and distribution mismatch. We propose \textbf{CtrlCoT}, a dual-granularity CoT compression framework that harmonizes semantic abstraction and token-level pruning through three components: Hierarchical Reasoning Abstraction produces CoTs at multiple semantic granularities; Logic-Preserving Distillation trains a logic-aware pruner to retain indispensable reasoning cues (e.g., numbers and operators) across pruning ratios; and Distribution-Alignment Generation aligns compressed traces with fluent inference-time reasoning styles to avoid fragmentation. On MATH-500 with Qwen2.5-7B-Instruct, CtrlCoT uses 30.7\% fewer tokens while achieving 7.6 percentage points higher than the strongest baseline, demonstrating more efficient and reliable reasoning. Our code will be publicly available at https://github.com/fanzhenxuan/Ctrl-CoT.
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。