Saved in:
Bibliographic Details
Main Authors: Jin, Zhengda, Wang, Bingbing, Li, Jing, Xu, Ruifeng, Zhang, Min
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.25869
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911716220076032
author Jin, Zhengda
Wang, Bingbing
Li, Jing
Xu, Ruifeng
Zhang, Min
author_facet Jin, Zhengda
Wang, Bingbing
Li, Jing
Xu, Ruifeng
Zhang, Min
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2605_25869
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Mitigating Provenance-Role Collapse in Long-Term Agents via Typed Memory Representation
Jin, Zhengda
Wang, Bingbing
Li, Jing
Xu, Ruifeng
Zhang, Min
Computation and Language
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.
title Mitigating Provenance-Role Collapse in Long-Term Agents via Typed Memory Representation
topic Computation and Language
url https://arxiv.org/abs/2605.25869