Single-Agent Systems
Single-Agent Systems is an Agentic Reasoning pattern in which a single model performs the typical agentic patterns of reflection, planning, and tool use.
Single-agent systems tend to thrive when there's a well-defined problem and feedback from other agent personas or the user isn't needed (Masterman et al., 2024).
Agents are given system prompts and the necessary tools; no other models provide feedback. However, there could be options for humans to guide the output.
See also Multi-Agent Systems.
Papers:
- The Landscape of Emerging AI Agent Architectures for Reasoning, Planning, and Tool Calling: A Survey
- ReAct (Agent)
References
Tula Masterman, Sandi Besen, Mason Sawtell, and Alex Chao. The Landscape of Emerging AI Agent Architectures for Reasoning, Planning, and Tool Calling: A Survey. April 2024. arXiv:2404.11584, doi:10.48550/arXiv.2404.11584. ↩