EASy Framework Optimizes Agentic System Efficiency via Reinforcement Learning

EASy uses reinforcement learning to train an LLM orchestrator that jointly optimizes for task success and computational cost. The framework utilizes a milestone-plan-act workflow and explicit cost-capability profiles for heterogeneous executors to improve routing beyond simple performance metrics.