长视野搜索代理必须管理快速增长的工作环境，因为它们推理,调用工具,并观察信息。天真地积累所有中间内容可能会压垮代理,，从而增加成本和错误风险。我们建议，有效的上下文管理应该是自适应的:代理的轨迹的一部分根据其当前与任务的相关性保持在不同的细节级别。为了实施这一原则,，我们引入了 Context-ReAct,，一种用于弹性上下文编排的通用代理范例，它将推理, 上下文管理, 和工具使用集成在一个统一的循环中。 Context-ReAct 提供五个原子操作: 跳过, 压缩, 回滚, 片段和删除,，允许代理动态重塑其工作上下文, 保留重要证据, 总结已解决的信息, 丢弃无用的分支, 并控制上下文大小。我们证明压缩算子在表达上是完整的,，而其他专业算子提供效率和保真度保证，降低生成成本和幻觉风险。在此范例, 的基础上，我们开发了 LongSeeker,，这是一种长范围搜索代理，在 10k 条合成轨迹上根据 Qwen3-30B-A3B 进行了微调。在四个代表性搜索基准测试中，, LongSeeker 在 BrowseComp 上取得了 61.5% 的成绩，在 BrowseComp-ZH, 上取得了 62.5% 的成绩，大大优于 Tongyi DeepResearch (43.2% 和 46.7%) 以及 AgentFold (36.2% 和 47.3%)。这些结果凸显了自适应上下文管理, 的潜力，表明智能体可以通过主动塑造工作记忆来实现更可靠、更高效的长视野推理。

Long-horizon search agents must manage a rapidly growing working context as they reason, call tools, and observe information. Naively accumulating all intermediate content can overwhelm the agent, increasing costs and the risk of errors. We propose that effective context management should be adaptive: parts of the agent的 trajectory are maintained at different levels of detail depending on their current relevance to the task. To operationalize this principle, we introduce Context-ReAct, a general agentic paradigm for elastic context orchestration that integrates reasoning, context management, and tool use in a unified loop. Context-ReAct provides five atomic operations: Skip, Compress, Rollback, Snippet and Delete, which allow the agent to dynamically reshape its working context, preserving important evidence, summarizing resolved information, discarding unhelpful branches, and controlling context size. We prove that the Compress operator is expressively complete, while the other specialized operators provide efficiency and fidelity guarantees that reduce generation cost and hallucination risk. Building on this paradigm, we develop LongSeeker, a long-horizon search agent fine-tuned from Qwen3-30B-A3B on 10k synthesized trajectories. Across four representative search benchmarks, LongSeeker achieves 61.5% on BrowseComp and 62.5% on BrowseComp-ZH, substantially outperforming Tongyi DeepResearch (43.2% and 46.7%) and AgentFold (36.2% and 47.3%). These results highlight the potential of adaptive context management, showing that agents can achieve more reliable and efficient long-horizon reasoning by actively shaping their working memory.

科目:人工智能(cs.AI)

Subjects: Artificial Intelligence (cs.AI)