Recursive Agent Optimization

TL;DR

Recursive Agent Optimization (RAO) is a reinforcement learning method that trains recursive agents capable of spawning sub-agents, improving scalability, efficiency, and generalization to complex tasks.

Contribution

The paper introduces RAO, a novel training approach enabling recursive agents to effectively delegate tasks and scale beyond their context window.

Findings

Recursive agents trained with RAO outperform single agents in efficiency.

RAO-trained agents can handle tasks beyond their initial context window.

Recursive agents generalize to more difficult tasks than those seen during training.

Abstract

We introduce Recursive Agent Optimization (RAO), a reinforcement learning approach for training recursive agents: agents that can spawn and delegate sub-tasks to new instantiations of themselves recursively. Recursive agents implement an inference-time scaling algorithm that naturally allows agents to scale to longer contexts and generalize to more difficult problems via divide-and-conquer. RAO provides a method to train models to best take advantage of such recursive inference, teaching agents when and how to delegate and communicate. We find that recursive agents trained in this way enjoy better training efficiency, can scale to tasks that go beyond the model's context window, generalize to tasks much harder than the ones the agent was trained on, and can enjoy reduced wall-clock time compared to single-agent systems.