Mitigating Provenance-Role Collapse in Long-Term Agents via Typed Memory Representation
For developers of persistent LLM agents, MemIR addresses a critical architectural failure mode (source-monitoring errors) that undermines reliability in long-term interactions.
MemIR introduces a typed memory representation to prevent provenance-role collapse in long-term LLM agents, achieving consistent improvements over baselines on LoCoMo and BEAM-100K, particularly in source tracking and temporal grounding tasks.
Long-term memory is essential for persistent LLM agents, yet prevailing architectures store 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. MemIR writes long-term memory into grounded atoms that separate raw evidence, retrieval cues, and truth-bearing claims, with factual authorization restricted to supported claim atoms. It then applies multi-route atomic projection and provenance-scoped utilization to transform heterogeneous retrieval hits into claim-centered candidate bundles and a normalized fact interface for answer generation. Experiments on LoCoMo and BEAM-100K demonstrate that MemIR consistently outperforms existing memory baselines, especially on tasks requiring source tracking, temporal grounding, and aggregation of fragmented evidence.