分而治之：弱模型如何超越 GPT-4o 处理长上下文
原标题：Plan, divide, and conquer: How weak models excel at long context tasks
AI 摘要
Together AI 的研究（ICLR 2026）提出了一种分而治之的框架，让较弱的模型通过策略性拆分长上下文任务，在长上下文任务上匹配或超越 GPT-4o 的单次处理性能。该框架识别了模型噪声、任务噪声和聚合器噪声三种噪声来源，并通过调整提示词来减少聚合器噪声。实验表明，Llama-3-70B 或 Qwen-72B 等较弱模型在长上下文任务上优于单次读取的 GPT-4o，同时成本更低、速度更快。但该方法不适用于跨块依赖强的任务。
正文节选
The Intuition: Don't ask one genius to read a library in an hour. Ask ten interns to read one book each. Our research, "When Does Divide and Conquer Work for Long Context LLM?" (ICLR 2026), introduces a framework to study this. We found that smaller models using a strategic "Divide & Conquer" design can match or beat GPT-4o single-shot on long context tasks. [paper, ICLR 2026] [code] Modern LLMs increasingly support massive context windows like 128K, 200K, even 1M+ tokens. This theoretically unl