An intelligent test recording method for a test bank, a storage medium and an equipment

By combining heterogeneous large language models and logical dependency graphs, the problem of inconsistent answer parsing in single-model question recording methods is solved, achieving high accuracy and teaching applicability of the question bank, and optimizing the question recording process.

CN122114108APending Publication Date: 2026-05-29读书郎教育科技有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
读书郎教育科技有限公司
Filing Date
2026-02-06
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing intelligent question recording methods mostly use a single model to generate answers, lacking multimodal verification and dynamic error correction mechanisms. This makes it difficult to meet the logical consistency requirements of complex questions, resulting in a decline in the accuracy of the question bank and its applicability to teaching.

Method used

At least two heterogeneous large language models are used to compare the consistency of answers. A confidence assessment and manual intervention process are combined. The parsed text is corrected through a logical dependency graph. A dual-person review and arbitration model is used to ensure the consistency between the answer and the parsing. Editing permissions are dynamically adjusted.

Benefits of technology

It improved the accuracy of answers and the consistency of explanations, optimized the question recording process, and enhanced the accuracy of the question bank and its ability to handle complex questions.

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

Abstract

The application discloses a question bank intelligent recording method, a storage medium and equipment, utilizes a heterogeneous large language model to generate multiple answers for cross verification, and combines confidence evaluation to effectively filter high-risk answers; through logical dependency graph and branch rewriting technology, accurate synchronization of analysis is realized, and the cost of manual review is reduced; through double-person review and arbitration mechanism, the authority of dispute processing is guaranteed, and through behavior portrait, dynamic management and control of editing authority are realized; through the RAG strategy, the processing capacity of complex questions is enhanced, and multiple scene teaching needs are adapted.
Need to check novelty before this filing date? Find Prior Art