Orchestrating Intelligence: A Framework for Dynamic Task Decomposition and Role Specialization in Multi-Agent Generative AI Systems

Mayank .; Rajender Nath

Orchestrating Intelligence: A Framework for Dynamic Task Decomposition and Role Specialization in Multi-Agent Generative AI Systems

Author(s)	Mayank ., Prof. Rajender Nath
Country	India
Abstract	The emergence of large language model (LLM)-based agents has catalyzed a paradigm shift in artificial intelligence, transitioning from monolithic single-model pipelines to collaborative multi-agent generative AI (MAGAI) systems capable of solving complex, open-ended tasks. Despite considerable progress in individual agent capabilities, the principled design of multi-agent architectures — encompassing dynamic task decomposition, inter-agent communication protocols, role specialization, and conflict resolution — remains an open research challenge. This paper proposes OrchestRAG, a novel framework for orchestrating multi-agent generative AI systems through a hierarchical planning layer combined with a retrieval-augmented generation (RAG) memory substrate. OrchestRAG introduces four core contributions: (1) a semantic task decomposer that partitions complex queries into subtask graphs, (2) a role-allocation engine that dynamically assigns specialized sub-agents based on competency embeddings, (3) a shared episodic memory module enabling asynchronous inter-agent knowledge transfer, and (4) a consensus arbitration mechanism that resolves conflicting agent outputs through evidence-weighted voting. Comprehensive evaluation across six benchmark datasets — spanning multi-hop reasoning, software engineering, biomedical question answering, and autonomous web navigation — demonstrates that OrchestRAG achieves a 23.4% improvement in task completion accuracy and a 31.7% reduction in token consumption relative to single-agent baselines, and outperforms three competing multi-agent frameworks. Ablation studies confirm the individual contribution of each architectural component. It is argued that structured orchestration is a necessary condition for reliable deployment of multi-agent AI in high-stakes enterprise environments.
Keywords	Multi-Agent Systems, Generative AI, Large Language Models, Task Decomposition, Role Specialization, Retrieval-Augmented Generation, Orchestration Frameworks
Field	Computer > Artificial Intelligence / Simulation / Virtual Reality
Published In	Volume 7, Issue 2, March-April 2026
Published On	2026-03-19

View / Download PDF File

E-ISSN 3048-7641

doi

CrossRef DOI is assigned to each research paper published in our journal.

AIJFR DOI prefix is
10.63363/aijfr

Downloads

Research Paper Format Copyright Permission Form and Undertaking Form Cover Page Vol 7 Isu 1 Cover Page Vol 6 Isu 6 Cover Page Vol 6 Isu 5

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA

About AIJFR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9898-79-39-59	editor@aijfr.com

Advanced International Journal for Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Orchestrating Intelligence: A Framework for Dynamic Task Decomposition and Role Specialization in Multi-Agent Generative AI Systems

Share this