Conflict detection method and apparatus, electronic device, and storage medium

By extracting multimodal information features through a self-attention mechanism, modal conflicts in generative artificial intelligence models are detected, which solves the security threats between multimodal input data, improves the accuracy and security of the system, and prevents erroneous decisions.

CN122364698APending Publication Date: 2026-07-10CHINA UNITED NETWORK COMM GRP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHINA UNITED NETWORK COMM GRP CO LTD
Filing Date
2026-03-17
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing technologies struggle to effectively detect and prevent conflicts between multimodal input data, leading to erroneous decisions by generative AI models and posing a security threat.

Method used

A self-attention mechanism is used to extract features from multimodal information, including text, image, and audio features. Modal conflicts are detected by using consistency and difference features. The self-attention mechanism is used to determine the consistency and difference features between any two pieces of modal information, thereby detecting whether there is a modal conflict.

Benefits of technology

It improves the accuracy, robustness, and security of the conflict detection system, avoids input mismatches in multimodal generative artificial intelligence models, prevents security threats, and ensures stable operation in security-critical scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122364698A_ABST
    Figure CN122364698A_ABST
Patent Text Reader

Abstract

This application provides a conflict detection method, apparatus, electronic device, and storage medium, relating to the field of artificial intelligence. The method includes: acquiring multimodal information; extracting features based on the multimodal information to obtain text features, image features, and audio features; determining consistency and difference features between any two modal information segments using a self-attention mechanism based on the text features, image features, and audio features; and detecting the existence of modal conflicts based on the consistency and difference features. This application can fully utilize information from multiple modalities, providing rich semantic information for conflict detection. The self-attention mechanism helps the conflict detection system discover whether contradictory content exists between different modal information segments, thereby improving the accuracy, robustness, security, and decision credibility of the conflict detection system.
Need to check novelty before this filing date? Find Prior Art