AgentArk: Distilling Multi-Agent Intelligence into a Single LLM Agent

TL;DR

AgentArk introduces a framework to distill multi-agent reasoning capabilities into a single large language model, enhancing efficiency and robustness while maintaining strong reasoning performance.

Contribution

The paper presents novel hierarchical distillation strategies that encode multi-agent dynamics into a single model, reducing computational costs and error propagation.

Findings

Distilled models retain multi-agent reasoning abilities.

Enhanced robustness and generalization across tasks.

Significant reduction in inference computational cost.

Abstract

While large language model (LLM) multi-agent systems achieve superior reasoning performance through iterative debate, practical deployment is limited by their high computational cost and error propagation. This paper proposes AgentArk, a novel framework to distill multi-agent dynamics into the weights of a single model, effectively transforming explicit test-time interactions into implicit model capabilities. This equips a single agent with the intelligence of multi-agent systems while remaining computationally efficient. Specifically, we investigate three hierarchical distillation strategies across various models, tasks, scaling, and scenarios: reasoning-enhanced fine-tuning; trajectory-based augmentation; and process-aware distillation. By shifting the burden of computation from inference to training, the distilled models preserve the efficiency of one agent while exhibiting strong reasoning and self-correction performance of multiple agents. They further demonstrate enhanced robustness and generalization across diverse reasoning tasks. We hope this work can shed light on future research on efficient and robust multi-agent development. Our code is at https://github.com/AIFrontierLab/AgentArk.