arXiv：cs.AI· Qi Cheng, Shengyu Chen, Wei Cheng, Yiqun Xie, Xiaowei Jia, Haoyu Wang, Haifeng Chen·· 10 小时前AI 评分43
OOPMAS：面向查询级工作流生成的面向对象多智能体系统
OOPMAS: Object-Oriented Multi-Agent Systems for Query-Level Workflow Generation
AI 导读
OOPMAS 是一个免训练框架，能在单个查询粒度上生成智能体集合与协调工作流，智能体以面向对象类定义表示，工作流则写成可执行的主函数。在覆盖代码、数学和问答的混合任务基准上，OOPMAS 达到 89.6% 准确率，比最强基线高出 18.1 个百分点；跨四个 LLM 主干替换实验显示一致性扩展，最强模型下达到 92.4%。
正文
Abstract:Multi-agent systems (MAS) powered by large language models have shown strong performance across code generation, mathematical reasoning, and question answering. However, existing methods for automating MAS design mostly operate at the task level, producing a single fixed workflow per benchmark that is applied uniformly to all queries. This assumption fails under realistic conditions. Query difficulty varies widely within a task, and real-world workloads mix heterogeneous task types. We introduce OOPMAS, a training-free framework that generates both the agent set and the coordination workflow at the granularity of individual queries. Agents are represented as object-oriented class definitions with dedicated roles, tools, and persistent state, and workflows are expressed as executable main functions over these agent objects. A dynamic skill library accumulates structured lessons from execution feedback across optimization rounds, enabling in-context improvement without any gradient updates or fine-tuning. On a mixed-task benchmark of queries spanning code, math, and QA, OOPMAS achieves 89.6% accuracy, outperforming the strongest baseline by 18.1 percentage points. A model-swap study across four LLM backbones shows consistent scaling, reaching 92.4% with the strongest model.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.07787 [cs.AI] |
| (or arXiv:2610.07787v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.07787 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Qi Cheng [view email]
[v1]
Tue, 6 Oct 2026 05:31:29 UTC (211 KB)
来源：arXiv：cs.AI · arxiv.org