ROAM improves memory management accuracy in AI language agents
ROAM: Robust Organization of Atomic Memories for Agents through Semantic Relations
Computation and Language
Summary
Language AI agents need to remember information over time, but keeping lots of small memory pieces can cause confusion or overlap. The authors created ROAM, a method that organizes these small memory pieces by understanding their relationships—such as whether one piece overlaps or conflicts with another. ROAM also combines related memories into main and supporting groups, helping the AI find the best information when answering questions. This approach makes the AI’s answers more accurate and avoids redundant or outdated information.
What this means in practice
- •For software engineers: Build AI systems with improved long-term memory management for consistent and accurate responses across many interactions.
- •For virtual assistant developers: Enhance virtual assistants’ ability to maintain precise and up-to-date knowledge without confusion from overlapping or conflicting memories.$Commercial implications: Improves virtual assistant products by enabling smarter memory handling, which drives better user experience and engagement.
Authors
Jianjie Zheng, Peng Lai, Sijie Cheng, Jiehui Zhao, Lei Yang, Guanhua Chen
Abstract
Long-term language-model agents rely on external memory across interactions. Atomic memories are particularly useful: their fine-grained semantic boundaries enable precise retrieval and direct comparison between observations. Yet accumulating atoms inevitably become redundant, overlapping, or conflicting. Existing methods often ask an LLM manager to add, update, delete, or rewrite memories directly, coupling semantic interpretation, storage decisions, and content generation in one error-prone operation. We introduce ROAM, a relation-guided framework that uses atomicity for management while allowing richer answer-time representations. ROAM classifies incoming--stored atom pairs as independent, equivalent, directionally subsuming, or conflicting, then organizes observations into active Primary and supporting Evidence roles. Fusion subsequently combines complementary details and temporal changes into compact, potentially non-atomic views. Only Primary views are retrieved for answering, preventing redundant or outdated atoms from competing independently. Across models and evaluation settings, ROAM improves answer accuracy by up to 29.8 percentage points. Ablations show complementary benefits from different relations and consistent gains from fusion beyond role organization. Mechanism analysis further finds 15.6-point higher answer-critical source recall and an 11.5-point lower confounder-token share. ROAM remains robust across manager scales.