SearchSwarm: Towards Delegation Intelligence in Agentic LLMs for Long-Horizon Deep Research
2026-06-08 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors study how large language models can handle very long and complex tasks by breaking them into smaller pieces and letting helper subagents work on those pieces. They focus on teaching the main model when and how to delegate tasks smartly, by creating a system that guides this process and produces training data. Using this, they fine-tuned a model that performs well on benchmarks testing task delegation and browsing. They plan to share their tools and data to help others build on their work.
large language modeltask decompositiondelegation intelligencesubagentfine-tuningcontext windowlong-horizon tasksBrowseComp benchmarksupervised learningmodel weights
Authors
Pu Ning, Quan Chen, Kun Tao, Xinyu Tang, Tianshu Wang, Qianggang Cao, Xinyu Kong, Zujie Wen, Zhiqiang Zhang, Jun Zhou
Abstract
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.