Scene-Aware Memory Discrimination: Deciding Which Personal Knowledge Stays

2026-02-01

Knowledge-Based Systems

AI Summary

This work proposes a Scene-Aware Memory Discrimination (SAMD) approach to address the challenges of irrelevant information interference and high computational overhead in massive user interactions. SAMD introduces a selective attention mechanism into personal knowledge management for the first time, employing a Gated Unit Module (GUM) to filter non-essential interactions and a Clustering Prompt Module (CPM) to dynamically establish scene-adaptive criteria for memory retention. This framework guides large language models to efficiently identify and preserve high-value memories. Experimental results demonstrate that SAMD significantly enhances both the efficiency and quality of memory construction in direct and indirect evaluations, while maintaining robustness in dynamic environments and effectively recalling critical information.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Large language models for search

Problem

• personal knowledge
• memory discrimination
• large language models
• selective attention
• user memory

Innovation

• Scene-Aware Memory Discrimination
• Gating Unit Module
• Cluster Prompting Module
• Selective Attention
• Personal Knowledge Management