基于大模型的记忆信息确定方法、任务处理方法和装置
By extracting and filtering information from interactions between agents and users using large models, and identifying memory information with high confidence and importance, the problem of poor user experience during agent-user interactions is solved, and more accurate responses that match user needs are achieved.
CN121581201BActive Publication Date: 2026-07-17BEIJING BAIDU NETCOM SCI & TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING BAIDU NETCOM SCI & TECH CO LTD
- Filing Date
- 2025-11-12
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
The problem of poor user experience when intelligent agents interact with users.
Method used
By extracting information from the interaction between the target account and the agent using a large model, the confidence and importance of candidate information are determined, and appropriate memory information is selected to perform the target task.
Benefits of technology
It improves the accuracy of intelligent agent responses and the degree to which they match user needs, thus optimizing the user experience.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN121581201B_ABST
Abstract
本公开提供了基于大模型的记忆信息确定方法、任务处理方法和装置，涉及人工智能技术领域，尤其涉及自然语言处理、深度学习、大模型技术领域。方法包括：通过第一大模型对目标账号与智能体之间的交互信息进行信息提取，以确定各第一候选信息；基于所述第一候选信息的置信度和所述第一候选信息的重要度中的至少一项，从各所述第一候选信息中确定所述目标账号关联的第一记忆信息；其中，所述目标账号关联的各记忆信息用于所述智能体执行所述目标账号关联的目标任务。
Need to check novelty before this filing date? Find Prior Art