AgentArk: Distilling Multi-Agent Intelligence into a Single LLM Agent
This addresses the problem of deploying multi-agent systems efficiently for AI researchers and practitioners, though it is incremental as it builds on existing distillation techniques.
The paper tackles the high computational cost and error propagation of LLM multi-agent systems by proposing AgentArk, a framework that distills multi-agent dynamics into a single model, achieving strong reasoning and self-correction performance while preserving efficiency.
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.