No Action Without a NOD: A Heterogeneous Multi-Agent Architecture for Reliable Service Agents

TL;DR

This paper introduces NOD, a multi-agent architecture for LLM service agents that enhances reliability and safety in long-horizon tasks by explicit state tracking and oversight.

Contribution

It proposes a novel heterogeneous multi-agent system with explicit global state and oversight mechanisms to improve service agent reliability.

Findings

NOD achieves higher task success rates on $ au^2$-Bench.

It reduces policy violations and tool hallucinations.

NOD improves long-horizon task reliability.

Abstract

Large language model (LLM) agents have increasingly advanced service applications, such as booking flight tickets. However, these service agents suffer from unreliability in long-horizon tasks, as they often produce policy violations, tool hallucinations, and misaligned actions, which greatly impedes their real-world deployment. To address these challenges, we propose NOD (Navigator-Operator-Director), a heterogeneous multi-agent architecture for service agents. Instead of maintaining task state implicitly in dialogue context as in prior work, we externalize a structured Global State to enable explicit task state tracking and consistent decision-making by the Navigator. Besides, we introduce selective external oversight before critical actions, allowing an independent Director agent to verify execution and intervene when necessary. As such, NOD effectively mitigates error propagation and unsafe behavior in long-horizon tasks. Experiments on $\tau^2$-Bench demonstrate that NOD achieves higher task success rates and critical action precision over baselines. More importantly, NOD improves the reliability of service agents by reducing policy violations, tool hallucinations, and user-intent misalignment.