AI RESEARCH

A Heterogeneous Temporal Memory Governance Framework for Long-Term LLM Persona Consistency

arXiv CS.AI

ArXi:2605.14802v1 Announce Type: new Large language models often suffer from fact loss, timeline confusion, persona drift, and reduced stability during long-range interaction, especially under high-noise knowledge bases, context clearing, and cross-model transfer. To address these issues, we