Information display method and apparatus
By receiving user input and historical interruption information in a large language model to determine prompt words, the bias problem of large language models in intent understanding and multi-turn interaction is solved, and more accurate and coherent responses are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- VIVO MOBILE COMM CO LTD
- Filing Date
- 2026-03-20
- Publication Date
- 2026-06-02
AI Technical Summary
Large language models suffer from semantic ambiguity and interpretation in intent understanding and multi-turn interaction processing, causing responses to deviate from user expectations, fail to effectively convey the intent of previous interactions, and are susceptible to interference from invalid historical information, thus reducing the accuracy of responses.
By receiving user input, the output of the large language model is stopped. By combining historical interruption information and current query information, targeted prompt words are determined to guide the large language model to generate an accurate response, retain historical interaction intent, and avoid interference in multi-turn interactions.
It improves the accuracy of responses from large language models and the coherence of interactions, reduces redundancy and deviation, and ensures that responses align with user needs.
Smart Images
Figure CN122133670A_ABST