AI RESEARCH
Mitigating Provenance-Role Collapse in Long-Term Agents via Typed Memory Representation
arXiv CS.CL
•
ArXi:2605.25869v1 Announce Type: new Long-term memory is essential for persistent LLM agents, yet prevailing architectures historical interactions as unstructured, flat text. This unconstrained storage induces provenance-role collapse, a critical failure mode where agents suffer from source-monitoring errors. To resolve this cognitive vulnerability at the architectural level, we propose MemIR, a typed Memory Intermediate Representation that operationalizes source monitoring as a structural constraint.