AI Summary: This paper introduces AOrchestra, a groundbreaking framework that moves multi-agent systems away from static, pre-defined workflows toward dynamic, on-the-fly agent generation. Instead of relying on a fixed pool of specialized bots, the central orchestrator evaluates the current task...
Language agents have shown strong promise for task automation, but realizing this promise for increasingly complex, long-horizon tasks has driven the rise of a sub-agent-as-tools paradigm. However, existing designs lack a dynamic abstraction view of sub-agents, thereby hurting adaptability. We address this challenge with a unified, framework-agnostic agent abstraction that models any agent as a compositional tuple: Instruction, Context, Tools, Model. Building on this abstraction, we introduce an agentic system called AOrchestra, where the central orchestrator concretizes the tuple at each step by curating task-relevant context, selecting tools, and delegating execution via on-the-fly automatic agent creation.
Share your opinion to help other learners triage faster.
Write a reviewInvite someone by email to share an invited review for AOrchestra: Automating Sub-Agent Creation for Agentic Orchestration.