Reproduce “Agent-as-a-Service based on Agent Network” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuhan Zhu, Haojie Liu, Jian Wang, Bing Li, Zikang Yin, Yefei Liao
Researchers from Wuhan University developed the Agent-as-a-Service based on Agent Network (AaaS-AN) framework, a service-oriented paradigm that treats AI agents as modular services capable of dynamic discovery and coordination. The framework achieved an average accuracy improvement of 5.77% on mathematical reasoning tasks and consistently outperformed other multi-agent frameworks in application-level code generation.
Key insights
- Treating AI agents as modular, standardized services allows for dynamic discovery, deployment, and coordination within collaborative networks, akin to traditional software services.
- A dynamic Agent Network structure, composed of vertices (agents/groups) and edges (routing), enables self-collaboration and adaptation for scalable multi-agent systems.
- A structured interoperability protocol (Plan, Multi-round Service Call, Summary) and a Service Scheduler with an Execution Graph facilitate efficient context management and task coordination across agents.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.