Model combination method of big language model for personalized long-period information

• Main Authors: KANG ZHIXI, ZHAN ZHENJIANG
• Format: Patent
• Language: Chinese; English
• Published: 01.12.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a model combination method of a large language model for personalized long-period information, which consists of a main parameter part, a bypass parameter part and a user parameter part, and comprises the following steps of: after an external text is input, embedding a word vector; and the user parameter part completes the combination of information related to the current user in the user parameter part, after the combination is formed, the calculation process of the main parameter part and the branch parameter part is started respectively, vector superposition is completed on the basis of the calculation results respectively, and final output is formed. According to the model combination method of the large language model for the personalized long-period information, the personalized information can be used in the model at low cost; the problem that the use of other users is influenced when local information updating is carried out on the LLM model at present can be solved; the reason