arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2606.09730cs.AI
收藏

SearchSwarm:面向长周期深度研究的代理LLM委托智能

SearchSwarm: Towards Delegation Intelligence in Agentic LLMs for Long-Horizon Deep Research

Tsinghua University(清华大学);Peking University(北京大学);Ant Group(蚂蚁集团);Gaoling School of Artificial Intelligence, Renmin University of China(中国人民大学高瓴人工智能学院)

Xiaochong Lan, Quan Chen, Kun Tao, Xinyu Tang, Tianshu Wang, Qianggang Cao, Xinyu Kong, Zujie Wen, Zhiqiang Zhang, Jun Zhou

AI总结:

提出SearchSwarm框架,通过监督微调将任务分解与委托决策内化到模型权重中,在BrowseComp和BrowseComp-ZH上取得同规模最佳性能。

摘要:

大型语言模型越来越需要处理复杂的、长周期的真实世界任务,这些任务的上下文需求可能无限增长,但模型上下文窗口本质上是有限的。最近的研究探索了一种范式,其中主代理分解任务并将子任务分派给子代理,子代理执行并仅返回汇总结果,从而节省主代理的上下文预算。然而,要很好地执行这一任务需要委托智能:分解复杂任务、确定何时委托以及委托什么、并将返回结果整合到持续工作流中的能力。这种能力的训练数据在自然文本中很少见,据我们所知,如何合成此类数据并训练模型获得这种能力在开源社区中仍基本未被探索。为填补这一空白,我们针对深度研究这一代表性的长周期代理任务进行了初步探索。具体来说,我们设计了一个引导工具,引导模型进行高质量的任务分解和委托,同时约束子代理正确返回结果以支持主代理的工作流。引导工具生成的轨迹自然地编码了正确的委托决策,我们将其作为监督微调数据,将委托智能内化到模型权重中。我们的模型SearchSwarm-30B-A3B在BrowseComp上达到68.1,在BrowseComp-ZH上达到73.3,在所有同规模模型中取得最佳结果。我们将发布我们的引导工具、模型权重和训练数据,以促进未来研究。

英文摘要:

Large language models are increasingly expected to handle complex, long-horizon real-world tasks whose context demands can grow without bound, yet model context windows remain inherently finite. Recent work explores a paradigm where a main agent decomposes tasks and dispatches subtasks to subagents, which execute and return only summarized results, conserving the main agent's context budget. However, performing this well requires delegation intelligence: the ability to decompose complex tasks, determine when and what to delegate, and integrate returned results into the ongoing workflow. Training data for this capability is scarce in naturally occurring text, and to our knowledge, how to synthesize such data and train models to acquire this capability remains largely unexplored in the open-source community. To bridge this gap, we present a preliminary exploration targeting deep research, a representative long-horizon agent task. Specifically, we design a harness that guides the model toward high-quality task decomposition and delegation, while constraining subagents to return results properly to support the main agent's workflow. The harness-guided trajectories naturally encode correct delegation decisions, which we use as supervised fine-tuning data to internalize delegation intelligence into model weights. Our resulting model, SearchSwarm-30B-A3B, achieves 68.1 on BrowseComp and 73.3 on BrowseComp-ZH, the best results among all models of comparable scale. We will release our harness, model weights, and training data to facilitate future research.

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。

↑