用代理上下文训练模型，让大模型更好处理长文本推理。
Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning
- 用小片段推理生成高质量思维链，再迁移到完整长文本
- 在多个数据集上超越强基线，计算开销更低
- 训练后能泛化到未见过的任务，适合长文本场景
当前大语言模型支持长达1000万词元的输入，但在需要复杂推理的长上下文任务中表现不佳。这类任务只需输入的一部分（代理上下文）即可解决，而非整个序列。尽管推理过程相同，模型在代理与全上下文上的表现差异显著。为此，我们提出ProxyCoT训练框架，将短代理上下文中的推理能力迁移至完整长上下文。具体而言，先通过强化学习或从更大教师模型中蒸馏，获得代理上下文上的高质量思维链推理轨迹，再通过监督微调将这些轨迹锚定在完整长上下文中。跨不同数据集的实验表明，ProxyCoT持续优于强基线，且计算开销更低。此外，经由ProxyCoT训练的模型能将其长上下文推理能力泛化至域外任务。
原文摘要 · Abstract (English)
Recent large language models support inputs of up to 10 million tokens, yet they perform poorly on long-context tasks that require complex reasoning. Such tasks can be solved using only a subset of the input -- a proxy context -- rather than the full sequence. Despite sharing the same underlying reasoning process, models exhibit a significant performance disparity between proxy and full contexts. To improve long-context reasoning, we propose ProxyCoT, a novel training framework that transfers reasoning capabilities from short proxy contexts to full long contexts. Specifically, we first obtain high-quality chain-of-thought reasoning traces on proxy contexts through reinforcement learning or distillation from a larger teacher model, and then ground the generated traces in full long contexts with supervised fine-tuning. Experiments across different datasets demonstrate that ProxyCoT consistently outperforms strong baselines with reduced computational overhead. Furthermore, models trained with ProxyCoT generalize their long-context reasoning capabilities to out-of-domain tasks.
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。