arXiv cs.MA· Ruhao Liu, Qi Li, Shizun Wang, Xinchao Wang·· 12 小时前AI 评分38
DynaFlow：学习自由形式的多智能体编排
DynaFlow: Learning Free-Form Multi-Agent Orchestration
AI 导读
DynaFlow 将自由形式的可执行工作流合成纳入主智能体的动作空间，由其自行决定何时协作并生成任务专属程序来协调子智能体。仅通过蒸馏轨迹的监督微调训练主智能体、子智能体保持固定，在数学推理、科学问答和多跳检索六个基准上以 Qwen3.5-27B 为骨干取得 88.08% 的 Forced 模式总体准确率，超出最强基线 10.90 个百分点。
正文
Abstract:Multi-agent systems often rely on predefined collaboration structures, limiting their ability to adapt to diverse task requirements. We introduce DynaFlow, a framework that integrates free-form executable workflow synthesis into the action space of the main agent. The agent decides when collaboration is needed and generates task-specific programs that coordinate sub-agents through sequential or parallel execution, structured information exchange, and verification. Execution results and traces inform subsequent actions, enabling the agent to adapt collaboration as reasoning progresses. We train only the main agent through supervised fine-tuning on distilled trajectories, followed by refinement on examples filtered for unsupported premises, biased verification, and information-flow defects. All sub-agents remain fixed, separating orchestration learning from executor capability improvement. Across six benchmarks spanning mathematical reasoning, scientific question answering, and multi-hop retrieval, DynaFlow achieves 88.08% overall accuracy in Forced mode, outperforming the strongest evaluated baseline by 10.90 percentage points with the same Qwen3.5-27B backbone. Workflow analysis finds more explicit use, verification, and refinement of upstream outputs after training, highlighting cross-stage information flow as a key dimension of multi-agent orchestration.
| Subjects: | Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2610.34467 [cs.MA] |
| (or arXiv:2610.34467v1 [cs.MA] for this version) | |
| https://doi.org/10.48550/arXiv.2610.34467 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ruhao Liu [view email]
[v1]
Sun, 25 Oct 2026 17:38:55 UTC (661 KB)
来源：arXiv cs.MA · arxiv.org