Data processing method and device based on RAG architecture and electronic equipment

By implementing a tagging and reuse strategy for RAG architecture documents, the problem of excessive token consumption in the RAG architecture was solved, enabling token reuse across requests and rounds, thereby improving the system's throughput and the stability of generated content.

CN122451065APending Publication Date: 2026-07-24BEIJING BAIDU NETCOM SCI & TECH CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202610288942.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-10
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

In the RAG architecture, excessive token consumption occurs during multi-round queries, especially in high-concurrency or multi-round interaction scenarios. Existing optimization solutions have limited effectiveness, resulting in resource waste and limited throughput.

Method used

By tagging documents in the RAG architecture, tokens with duplicate or similar document content can be identified and reused. By employing document slicing, filtering, and concatenation strategies, duplicate token input can be reduced, enabling token reuse across requests and rounds.

Benefits of technology

It effectively reduces token consumption, improves system throughput and the stability of generated content, and reduces resource consumption, especially in high-concurrency and long-session scenarios.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a data processing method and device based on an RAG architecture and electronic equipment, relates to the technical field of artificial intelligence large models, and further relates to an RAG architecture optimization technology. According to the specific implementation scheme, the method comprises the steps of marking a document obtained through an RAG architecture after a generative system receives a query queried by a user; and after the RAG architecture initiates the retrieval recall process again, if at least a first document in the retrieved and recalled documents is repeated or similar to the marked document in content, reusing the lexical token of the marked document for the at least a first document. By utilizing the method and the device, the token consumption in the generative system can be reduced.
Need to check novelty before this filing date? Find Prior Art