A large model construction method for intelligent determination of traffic accident liability

By constructing a mapping dictionary and efficiently fine-tuning the parameters of a large language model, the problem of data format differences in traffic accident liability determination was solved, and the model achieved high accuracy and practicality under multiple data collection sources.

CN122173932APending Publication Date: 2026-06-09TRAFFIC MANAGEMENT RES INST OF THE MIN OF PUBLIC SECURITY
View PDF 4 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
TRAFFIC MANAGEMENT RES INST OF THE MIN OF PUBLIC SECURITY
Filing Date
2026-04-28
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing technologies for determining liability in traffic accidents suffer from low accuracy in practical applications due to the diversity of data collection sources and the resulting differences in data formats. This makes it difficult to apply the models to various scenarios.

Method used

By constructing a mapping dictionary to convert structured data into natural language descriptions and using a large language model for efficient parameter fine-tuning, an intelligent traffic accident liability determination model that can be applied to multiple data collection sources is generated.

Benefits of technology

This improves the applicability and accuracy of the model under different data acquisition sources, reduces the performance requirements of computing devices, and enhances the practicality of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122173932A_ABST
    Figure CN122173932A_ABST
Patent Text Reader

Abstract

The application provides a large model construction method for intelligent determination of traffic accident liability, which converts structured historical data into structured training sample data based on natural language expression, compared with a training sample construction method in the prior art for converting natural language into formatted data, the conversion result is more accurate, that is, the accuracy of the sample data can be improved, and the accuracy of the model training result is further improved; the method constructs training sample data for a large language model based on complete traffic accident handling records, obtains structured training sample data in the format of guide instructions-input text-output text, and performs parameter efficient supervision fine-tuning on the pre-trained large language model based on the sample data, so that the general large model obtains professional traffic accident liability determination capability while maintaining the general capability of the model, and the practicability of the large language model is effectively improved.
Need to check novelty before this filing date? Find Prior Art