FloWright: Co-Evolving Multi-Agent Workflows via Hierarchical Rewards
October 2, 2026
FloWright enables the simultaneous optimization of both workflow generators and executing agents using a structure-aware reward paradigm. This method allows multiple roles to co-evolve within a workflow without requiring additional models, external labels, or extra executions.
HOW THIS AFFECTS YOU
●
builderThis provides a path to building more efficient multi-agent systems that improve themselves through execution.
●
researcherYou can now optimize entire agentic ecosystems rather than just the top-level orchestrator.