Paper 2602.03786
AOrchestra: Automating Sub-Agent Creation for Agentic Orchestration
- Published
- Feb 2026
- Research lab
- Independent
- Citations
- 18
- GitHub
- 153 stars
01 In brief
Summary
AORCHESTRA is an agentic framework that automates sub-agent creation for complex, long-horizon tasks.
It introduces a unified four-tuple abstraction (Instruction, Context, Tools, Model) to model any agent, enabling on-demand specialization.
A central orchestrator decomposes tasks, curates context, selects tools and models, and delegates execution to dynamically created sub-agents.
This design decouples orchestration from execution, making sub-agents plug-and-play and the orchestrator learnable.
The orchestrator can be improved via supervised fine-tuning (SFT) for task orchestration and iterative in-context learning for cost-aware model routing.
Evaluated on GAIA, Terminal-Bench 2.0, and SWE-Bench-Verified, AORCHESTRA consistently outperforms baselines like ReAct, OpenHands, Mini-SWE, and Claude Code.
With Gemini-3-Flash, it achieves 80.00 pass@1 on GAIA, 52.86 on Terminal-Bench, and 82.00 on SWE-Bench, a 16.28% relative improvement over the strongest baseline.
SFT improves GAIA pass@1 by +11.51%, and in-context learning reduces average cost by 18.5% while improving accuracy.
The framework also demonstrates plug-and-play sub-agent flexibility and Pareto-efficient cost-performance trade-offs.
02 From the paper
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