用多教师协作生成更优长链推理，让小模型逼近大模型效果。
Distilling Long-CoT Reasoning through Collaborative Step-wise Multi-Teacher Decoding
- 多教师分步协作，按预测困惑度动态筛选推理步骤。
- 用少且结构化的监督信号，学生模型接近教师水平。
- 适合需要高效高质量推理的场景，尤其开放域任务。
大模型长链推理（Long-CoT）虽能力强，但全规模推理计算开销大，难以实用。现有基于筛选的方法在事后选择完整推理路径，忽视异构教师间的协作与动态探索，导致冗余采样并遗漏互补推理。本文提出CoRD框架，通过预测困惑度评分与束搜索引导的分步推理合成，使异构大模型协同构建连贯推理轨迹，高效保留多样且高潜力的假设。实验表明，CoRD生成的推理数据质量更高，仅需少量结构化监督即可实现接近教师水平的学生模型表现，且无明显效率损失。该方法在跨领域和开放性任务中也具备良好泛化能力。代码与模型已开源：https://github.com/DISL-Lab/CoRD。
原文摘要 · Abstract (English)
Distilling large reasoning models is essential for making Long-CoT reasoning practical, as full-scale inference remains computationally prohibitive. Existing curation-based approaches select complete reasoning traces post-hoc, overlooking collaboration among heterogeneous teachers and lacking dynamic exploration, which leads to redundant sampling and missed complementary reasoning. We introduce CoRD, a collaborative multi-teacher decoding framework that performs step-wise reasoning synthesis guided by predictive perplexity-based scoring and beam search. This enables heterogeneous LRMs to jointly construct coherent reasoning trajectories while efficiently preserving diverse, high-potential hypotheses. Experiments show that CoRD produces higher-quality reasoning data and achieves near teacher-level student performance with fewer, structured supervision signals, without substantial efficiency overhead. CoRD further generalizes well to out-of-domain and open-ended settings. The dataset and model are available at \href{https://github.com/DISL-Lab/CoRD}{https://github.com/DISL-Lab/CoRD}.
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。