Know thyself: LLM schema for personal memory (github.com)

A newly proposed framework aims to enhance how large language models (LLMs) manage personal memory by transforming flat memory lists into structured graphs. This approach categorizes information into typed nodes—such as facts, episodes, patterns, and open questions—where each claim is underpinned by provenance. This structure allows for the accumulation of confidence from independent derivations rather than treating repeated assertions as multiple pieces of evidence, thereby improving the reliability of the model's knowledge representation. This innovation is significant for the AI/ML community as it addresses limitations in traditional memory systems used by LLMs, promoting a more nuanced understanding of personal knowledge. By implementing a validity window and separating observations from interpretations, the schema fosters adaptive learning over time, crucial for applications that require ongoing context awareness. Overall, the personal-graph schema builds on established W3C standards while incorporating unique extensions tailored for individual memory, paving the way for more sophisticated interactions and personalized AI experiences.