Optimization method of multi-agent system, data processing method and electronic equipment

By constructing and updating the graph structure of the multi-agent system, the problem of poor optimization effect in the existing technology is solved, and dual optimization of the system architecture in terms of structure and semantics is achieved, thereby improving the system's performance and success rate.

CN122433833APending Publication Date: 2026-07-21ALIBABA (CHINA) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ALIBABA (CHINA) CO LTD
Filing Date
2026-03-06
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing technologies for optimizing multi-agent systems mainly focus on surface-level content tuning of text configurations (such as fine-tuning wording or format), resulting in poor optimization effects and a lack of structural understanding and optimization of the system architecture.

Method used

By constructing an initial graph structure to represent the system architecture of the multi-agent system, and using the target execution feedback information to update the initial graph structure and optimize it into a target graph structure, the system structure and semantics are optimized simultaneously, including adding and deleting tools and reconstructing the interaction logic of the agents.

Benefits of technology

It improves the optimization effect of multi-agent systems, enhances the system's performance and success rate in complex tasks, and achieves dual optimization of the system architecture in terms of structure and semantics.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122433833A_ABST
    Figure CN122433833A_ABST
Patent Text Reader

Abstract

The application discloses an optimization method of a multi-agent system, a data processing method and an electronic device. It relates to the field of artificial intelligence, and the method comprises the following steps: obtaining text configuration information of an agent in a multi-agent system; constructing an initial graph structure based on the text configuration information of the agent, wherein the initial graph structure is used for structurally representing the system architecture of the multi-agent system; updating the initial graph structure based on target execution feedback information of the multi-agent system to obtain a target graph structure, and optimizing the multi-agent system based on the target graph structure. The method solves the problem that related technologies optimize the surface content of text configuration (such as fine-tuning the diction or format) to optimize the multi-agent system, and the optimization effect is poor.
Need to check novelty before this filing date? Find Prior Art