CAIRN: Dynamic Fact-Intent DAGs for Multi-Agent Exploration

2026-09-26

AI Summary

This study addresses the challenges of unstructured organization and inefficient exploration in LLM-based multi-agent systems tackling complex tasks by proposing a collaborative paradigm grounded in a fact-intent dynamic directed acyclic graph (DAG). In this framework, reasoner agents plan intents while executor agents generate facts. A persistent DAG structure is innovatively introduced to manage task dependencies, enabling knowledge reuse, trajectory auditability, and human intervention support. Experimental evaluations demonstrate that in compute-intensive scenarios such as cybersecurity and mathematical reasoning, the proposed framework accelerates execution in 76.5% of cases, achieving up to a 3.08× speedup while significantly reducing token consumption.

Abstract

LLM-powered autonomous systems have demonstrated promising capabilities in mathematical reasoning, engineering, and cybersecurity. Yet how to organize these systems for effective, reliable, and sustained performance remains an open question. In this paper, we present CAIRN, a fact-intent-driven multi-agent paradigm for goal-directed exploration. CAIRN represents observations and planned investigations as a dynamic directed acyclic graph (DAG). A reasoner interprets facts to propose intents, which workers execute to produce new facts. Each intent references its supporting facts and defines a potential exploration branch. The persistent graph preserves goals, dependencies and findings across workers, supporting knowledge reuse and parallel exploration. The graph also makes execution trajectories traceable and auditable, providing a basis for human verification and intervention. We evaluate CAIRN across cybersecurity and mathematical reasoning tasks, examining task success, time to solution, and token consumption. DAG-based coordination can incur higher token costs with no observable performance gains on tasks that require little effort. However, on high-effort tasks (at least 1M tokens), we observe faster solutions in 76.5% of cases, with speedups of up to 3.08x. Moreover, as task effort increases, these time gains become more pronounced while relative token overhead declines, highlighting the potential of DAG-guided parallel exploration.

Problem

• multi-agent systems
• large language models
• autonomous systems
• task coordination

Innovation

• Multi-Agent System
• Directed Acyclic Graph (DAG)
• Fact-Intent Paradigm
• Parallel Exploration
• Traceability