Recursive Agent Optimization
For reinforcement learning and agentic AI, RAO provides a training method for recursive agents that improves scalability and generalization, though the specific gains over existing recursive or hierarchical methods are not quantified.
RAO trains recursive agents that delegate sub-tasks to new instances of themselves, enabling scaling to longer contexts and harder problems via divide-and-conquer. The method improves training efficiency, generalizes to tasks beyond the training distribution, and reduces wall-clock time compared to single-agent systems.
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.