Empathic training agent system and method based on multiple memory perception

The empathic training agent system, controlled by the plot tree and plot nodes, solves the problems of difficult context organization and unified integration of multimodal information in empathic training, realizes the continuity of the training process and the continuous feedback of results, and improves the effectiveness and personalized guidance of empathic training.

CN122363518APending Publication Date: 2026-07-10TSINGHUA SHENZHEN INTERNATIONAL GRADUATE SCHOOL
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
TSINGHUA SHENZHEN INTERNATIONAL GRADUATE SCHOOL
Filing Date
2026-04-20
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing empathy training methods suffer from problems such as short duration, unbalanced training dimensions, lack of continuous recording during the training process, insufficient individual difference responses, and difficulty in uniformly organizing multimodal information, making it difficult to feed back training results to subsequent training.

Method used

An empathic training agent system based on multiple memory perception is adopted. The training phase is controlled by the plot tree and plot nodes. Combined with dynamic information summarization and multiple memory management, a runtime context is formed. The execution results are continuously fed back to the memory system through business writing and empathic evaluation, forming a training closed loop.

Benefits of technology

It achieves consistency during the training phase, continuity of context in long-term interactions, continuous recording of training results, and personalized guidance, thereby improving the effectiveness and observability of empathic training.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122363518A_ABST
    Figure CN122363518A_ABST
Patent Text Reader

Abstract

A system and method for training an empathic intelligent agent based on multiple memory perception includes: an input acquisition module that receives voice, text, operations, forum browsing, encyclopedia viewing, and optional image input from players in a virtual reality scene; a plot control module that manages the plot tree and nodes, defining training phase boundaries, character states, and task constraints; a dynamic information summarization module that summarizes the first forum post associated with the current plot tree, player encyclopedia event descriptions, scene operations within a preset time window, and node backgrounds to generate a dynamic summary; a multiple memory management module that maintains working / short-term / long-term memories and a knowledge base, combining retrieval results with the dynamic summary to form a runtime context; an interaction judgment module that, based on the runtime context, first performs rule judgment and then semantic judgment to determine whether it is a normal reply, forum post, encyclopedia addition / modification, plot transfer, or empathy assessment; a business writing module that executes forum posting, encyclopedia writing, and plot state updates; and a feedback generation module that performs empathy evaluation and generates feedback, feeding the results back to the memory system.
Need to check novelty before this filing date? Find Prior Art