AOrchestra: Automating Sub-Agent Creation for Agentic Orchestration

2026-02-03

AI Summary

This work addresses the limited adaptability of existing language agents in handling complex, long-horizon tasks due to their inability to dynamically abstract sub-agents. To overcome this, we propose a unified, framework-agnostic agent quadruple abstraction—comprising instruction, context, tools, and model—that enables on-demand composition and generation of specialized sub-agents for the first time. Building upon this abstraction, we develop a context-aware task decomposition mechanism coupled with dynamic tool and model selection, facilitating plug-and-play agent orchestration and controllable trade-offs between performance and cost. Evaluated on GAIA, SWE-Bench, and Terminal-Bench using Gemini-3-Flash as the base model, our system achieves an average improvement of 16.28% over the strongest baseline.

Technology Category

• Multiagent Systems: Adversarial Agents
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Language agents have shown strong promise for task automation. Realizing this promise for increasingly complex, long-horizon tasks has driven the rise of a sub-agent-as-tools paradigm for multi-turn task solving. However, existing designs still 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 tuple Instruction, Context, Tools, Model. This tuple acts as a compositional recipe for capabilities, enabling the system to spawn specialized executors for each task on demand. Building on this abstraction, we introduce an agentic system AOrchestra, where the central orchestrator concretizes the tuple at each step: it curates task-relevant context, selects tools and models, and delegates execution via on-the-fly automatic agent creation. Such designs enable reducing human engineering efforts, and remain framework-agnostic with plug-and-play support for diverse agents as task executors. It also enables a controllable performance-cost trade-off, allowing the system to approach Pareto-efficient. Across three challenging benchmarks (GAIA, SWE-Bench, Terminal-Bench), AOrchestra achieves 16.28% relative improvement against the strongest baseline when paired with Gemini-3-Flash. The code is available at: https://github.com/FoundationAgents/AOrchestra

Problem

• language agents
• sub-agent abstraction
• task automation
• agentic orchestration
• adaptability

Innovation

• agent abstraction
• dynamic sub-agent creation
• agentic orchestration
• framework-agnostic
• on-the-fly agent generation