aiqbench analysis
This research introduces DeOrch, a novel framework that decouples task decomposition from worker selection in multi-agent systems to improve transferability and credit assignment. By separating planning from concrete agent choice, the method enables dynamic adaptation to unseen worker pools without retraining, addressing key limitations in current learned orchestration approaches.
Decoupled Multi-Agent Orchestration
arXiv:2610.07556v2 Announce Type: replace Abstract: Learned orchestration can automatically construct effective language-model multi-agent systems, but existing approaches couple planning to fixed worker pools and train decomposition and collaboration from the same terminal outcome, limiting transfer and obscuring credit assignment. We introduce DeOrch, which separates worker-agnostic planning from concrete worker selection. Its two-stage planner first decomposes the task without worker information, then chooses collaboration operations using compact, worker-identity-free matchability feedback from the pool, enabling conditional credit assignment to decomposition and collaboration decisions. A lightweight matcher estimates worker suitability from behavior on a fixed probe set and adapts online with a contextual bandit, allowing new workers to be incorporated without retraining the planner or matcher. Across diverse in- and out-of-distribution tasks, DeOrch outperforms prior automatic MAS orchestration methods with fewer worker calls than competing learned orchestrators, remains effective when transferred to an entirely unseen worker pool without retraining, and shows consistent gains from both components.
Abstract:Learned orchestration can automatically construct effective language-model multi-agent systems, but existing approaches couple planning to fixed worker pools and train decomposition and collaboration from the same terminal outcome, limiting transfer and obscuring credit assignment. We introduce DeOrch, which separates worker-agnostic planning from concrete worker selection. Its two-stage planner first decomposes the task without worker information, then chooses collaboration operations using compact, worker-identity-free matchability feedback from the pool, enabling conditional credit assignment to decomposition and collaboration decisions. A lightweight matcher estimates worker suitability from behavior on a fixed probe set and adapts online with a contextual bandit, allowing new workers to be incorporated without retraining the planner or matcher. Across diverse in- and out-of-distribution tasks, DeOrch outperforms prior automatic MAS orchestration methods with fewer worker calls than competing learned orchestrators, remains effective when transferred to an entirely unseen worker pool without retraining, and shows consistent gains from both components.
Artificial Intelligence (cs.AI)
arXiv:2610.07556 [cs.AI]
(or arXiv:2610.07556v2 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.07556 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 240 words
Related coverage
AI-driven development lifecycle using Amazon Bedrock AgentCore
AWS · Sep 3, 2026
PANDA: A Decentralized Architecture with Flexible Orchestration for Scalable, Fault-Tolerant Multi-Agent Systems
arXiv · Oct 1, 2026
Agentic ERP: Multi-Agent Large Language Model Architecture for Autonomous Enterprise Resource Planning
arXiv · Jul 21, 2026
A Production-Grade, High-Efficiency Agentic RL Framework for Frontier LLMs with Validated Recipes
arXiv · Oct 14, 2026