A survey of graph-based agent memory 2023 to 2026, and the case that its retrieval problems are geometric.
The research
Agents forget between sessions. We study graph-structured, temporal memory for agents: what to keep, how to retrieve it, and why the geometry of the memory matters.
What it changes in practice
AI native means the memory that holds a case together belongs to the system, not to one person. This field is how that memory gets built and kept honest.