推理的迁移：解析思维链如何在模型间传递
Reasoning that Travels: Dissecting How Chain-of-Thought Transfers Across Models
AI总结：
本文通过提供者-接收者框架，研究大型推理模型生成的思维链如何影响其他模型的回答，发现完整思维链的迁移效果因基准测试而异，且部分思维链前缀能指导接收者继续推理，答案一致性可作为提前停止提供者推理的无标签信号。
摘要：
大型推理模型（LRMs）在生成最终答案之前，通常会生成大量的思维链（CoT）痕迹。作为显式的文本产物，这些痕迹可以传递给其他模型以解决相同任务，从而实现跨模型的推理迁移。然而，成功的迁移本身并不能揭示提供的CoT如何贡献于另一个模型的答案。我们通过一个受控的提供者-接收者框架研究这一问题，其中提供者生成推理痕迹，接收者从逐渐增长的痕迹前缀中解决相同问题。我们比较了强制回答（接收者直接从前缀中回答）和自由生成（接收者在回答前可以继续推理）两种模式。跨模型和基准测试，完整痕迹通常能成功迁移，但前缀轨迹揭示了不同的机制。在强制回答模式下，AIME的迁移主要由显式答案的可用性驱动。MMLU-Pro则反映了接收者能力的更大作用，而ZebraLogic依赖于部分结构化答案信息而非完整答案泄露。在自由生成模式下，部分CoT提高了跨基准测试的性能，表明前缀可以指导继续推理。最后，接收者之间的答案一致性为提前停止提供者推理提供了无标签信号。总体而言，跨模型CoT迁移并非单一现象：它可以反映答案提取、推理支架或接收者依赖的能力。
英文摘要：
Large reasoning models (LRMs) often generate extensive chain-of-thought (CoT) traces before producing a final answer. As explicit textual artifacts, these traces can be passed to other models to solve the same task, enabling cross-model reasoning transfer. Yet successful transfer alone does not reveal how the provided CoT contributes to another model's answer. We study this question with a controlled provider--receiver framework, where a provider generates a reasoning trace and a receiver solves the same problem from increasingly longer trace prefixes. We compare force-answer, where the receiver answers directly from the prefix, with free-generation, where it may continue reasoning before answering. Across models and benchmarks, full traces often transfer successfully, but prefix trajectories reveal distinct mechanisms. In force-answer mode, AIME transfer is largely driven by explicit answer availability. MMLU-Pro instead reflects a larger role for receiver competence, while ZebraLogic depends on partial structured-answer information rather than complete-answer leakage alone. In free-generation mode, partial CoTs improve performance across benchmarks, indicating that prefixes can guide continued reasoning. Finally, answer agreement among receivers provides a gold-free signal for stopping provider reasoning early. Overall, cross-model CoT transfer is not a single phenomenon: it can reflect answer extraction, reasoning scaffolding, or receiver-dependent competence.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
