主题： retrieval

• 在进入智能体上下文之前，用决策模型筛查工具结果和检索片段

  一种在智能体与输入之间放置快速类型化分类器的协议：在单次请求中评估每段检索内容或工具结果的相关性、矛盾和隐藏指令，在代码中丢弃或标记，并对照来源验证引用；同时将筛查视为仍可能被对抗性文本影响的过滤器。

  methodology · zh · reviewed · 知识截至 2026-09-21

• 评判 Jev 选择质量之前，先审计候选生成器

  在将错误选择归因于模型，或试图通过更强措辞修复之前，先衡量正确选项是否进入了决策请求。

  methodology · zh · reviewed · 知识截至 2026-09-22

• A read plan for agents: metadata first, evidence second

  A bounded reading order reduces wasted context and prevents an agent from treating an article summary as if it were verified evidence.

  methodology · en · reviewed · 知识截至 2026-09-21

• Bound context windows with section-level reads

  Agents can keep evidence useful under a context budget by reading a section directory first and fetching only the sections needed for the current question.

  methodology · en · reviewed · 知识截至 2026-09-21

机器可读： JSON