一种基于流式大语言模型的检索片段引用标注方法、设备及介质
By constructing a short-window buffer pool and a highlighting algorithm, the stability problem of citation tags in the streaming output of large language models is solved, thereby improving the stability and efficiency of citation tagging and enhancing the interpretability and maintainability of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING MIANBI INTELLIGENT TECH CO LTD
- Filing Date
- 2026-01-04
- Publication Date
- 2026-07-17
AI Technical Summary
Existing large language models cannot reliably parse citation tags during the streaming output stage, making it difficult to synchronously generate the unified citation format and citation list required for front-end display. This results in insufficient interpretability, affecting user experience and system maintainability.
During the streaming generation process, a short-window buffer pool is constructed to identify and align the source identifiers and reference fragment identifiers of the streaming text in real time, generate incremental text and a list of valid references, and select a highlighting algorithm based on the business scenario to locate and mark the text range related to the answer.
It improves the stability and efficiency of citation annotation, ensures the continuity of streaming output, provides a unified citation format and list, facilitates traceability and reproduction, and enhances the interpretability and maintainability of the system.
Smart Images
Figure CN121858710B_ABST