Adaptive Self–Mutual–Human Orchestration in Large Language Models: A Universal Framework for Function-Independent Selection, Dynamic Switching, Combination, and Reconfiguration of Artificial and Human Operational Mechanisms

Large language models (LLMs) are increasingly capable of performing complex functions through autonomous processing, interactions with other artificial intelligence systems, and collaboration with human participants. However, existing approaches frequently emphasize particular operational configurations, such as self-refinement, multi-agent collaboration, or human-in-the-loop interaction, without establishing a sufficiently general mechanism for dynamically determining which configuration should perform an arbitrary function at a particular time. This paper proposes a universal adaptive operational framework that enables LLM-based systems to move beyond dependence on any single execution pathway. The framework establishes three fundamental operational modes: Self-X, in which an individual model performs a function through its own internal capabilities; Mutual-X, in which multiple models or artificial agents interact, exchange information, and jointly perform the function; and Human-X, in which human individuals or groups participate in performing, directing, evaluating, or supporting the function. The variable X represents an arbitrary function, operation, capability, process, or mechanism rather than a predefined task category. The central contribution is the integration of adaptive functional selection, dynamic operational switching, simultaneous mechanism combination, and structural reconfiguration within a unified architecture. Instead of permanently assigning a function to self-execution, multi-model collaboration, or human participation, the proposed framework determines the most appropriate operational configuration according to task complexity, capability availability, performance, uncertainty, contextual conditions, resource requirements, risk, and feedback. It further supports transitions between operational modes during task execution, including Self-X to Mutual-X, Mutual-X to Human-X, Human-X to Self-X, and other possible transitions, as well as hybrid configurations involving two or three modes simultaneously. The framework is formalized through a generalized function space, an operational configuration space, an adaptive selection policy, a switching model, a combination mechanism, and a dynamic reconfiguration architecture. Its applicability extends beyond reasoning and generation to arbitrary LLM-related functions, including learning, reinforcement, supervision, evaluation, verification, correction, planning, memory management, optimization, communication, and decision-making. The proposed framework builds upon the author's previously introduced Universal Self-Mutual-Human Model (USMHM), extending its theoretical foundation toward an explicitly function-independent, dynamically orchestrated operational paradigm. To the best of our knowledge, this paper is the first to formally propose this particular unified architecture in which arbitrary LLM functions can adaptively alternate among self-directed, multi-model mutual, and human-participatory execution pathways, rather than remaining constrained to a predetermined operational mode. This work establishes a general theoretical foundation for the development of dynamically configurable, context-sensitive, and heterogeneous human–AI operational systems.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-19
DOI
https://doi.org/10.5281/zenodo.24358607
Primary Topic
Artificial Intelligence Applications
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Adaptive Self–Mutual–Human Orchestration in Large Language Models: A Universal Framework for Function-Independent Selection, Dynamic Switching, Combination, and Reconfiguration of Artificial and Human Operational Mechanisms

Jincheng Zhang
Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence Applications
preprint

Adaptive Self–Mutual–Human Orchestration in Large Language Models: A Universal Framework for Function-Independent Selection, Dynamic Switching, Combination, and Reconfiguration of Artificial and Human Operational Mechanisms

Jincheng Zhang
preprint en

Abstract

Large language models (LLMs) are increasingly capable of performing complex functions through autonomous processing, interactions with other artificial intelligence systems, and collaboration with human participants. However, existing approaches frequently emphasize particular operational configurations, such as self-refinement, multi-agent collaboration, or human-in-the-loop interaction, without establishing a sufficiently general mechanism for dynamically determining which configuration should perform an arbitrary function at a particular time. This paper proposes a universal adaptive operational framework that enables LLM-based systems to move beyond dependence on any single execution pathway. The framework establishes three fundamental operational modes: Self-X, in which an individual model performs a function through its own internal capabilities; Mutual-X, in which multiple models or artificial agents interact, exchange information, and jointly perform the function; and Human-X, in which human individuals or groups participate in performing, directing, evaluating, or supporting the function. The variable X represents an arbitrary function, operation, capability, process, or mechanism rather than a predefined task category. The central contribution is the integration of adaptive functional selection, dynamic operational switching, simultaneous mechanism combination, and structural reconfiguration within a unified architecture. Instead of permanently assigning a function to self-execution, multi-model collaboration, or human participation, the proposed framework determines the most appropriate operational configuration according to task complexity, capability availability, performance, uncertainty, contextual conditions, resource requirements, risk, and feedback. It further supports transitions between operational modes during task execution, including Self-X to Mutual-X, Mutual-X to Human-X, Human-X to Self-X, and other possible transitions, as well as hybrid configurations involving two or three modes simultaneously. The framework is formalized through a generalized function space, an operational configuration space, an adaptive selection policy, a switching model, a combination mechanism, and a dynamic reconfiguration architecture. Its applicability extends beyond reasoning and generation to arbitrary LLM-related functions, including learning, reinforcement, supervision, evaluation, verification, correction, planning, memory management, optimization, communication, and decision-making. The proposed framework builds upon the author's previously introduced Universal Self-Mutual-Human Model (USMHM), extending its theoretical foundation toward an explicitly function-independent, dynamically orchestrated operational paradigm. To the best of our knowledge, this paper is the first to formally propose this particular unified architecture in which arbitrary LLM functions can adaptively alternate among self-directed, multi-model mutual, and human-participatory execution pathways, rather than remaining constrained to a predetermined operational mode. This work establishes a general theoretical foundation for the development of dynamically configurable, context-sensitive, and heterogeneous human–AI operational systems.

Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon