Continuum Memory Architectures for Long-Horizon LLM Agents
This addresses the structural inability of RAG to accumulate or mutate memory for long-horizon agents, though it is incremental as it builds on existing memory augmentation strategies.
The paper tackles the problem of LLM agents lacking temporal continuity and stateful memory in retrieval-augmented generation (RAG) by proposing Continuum Memory Architecture (CMA), which shows consistent behavioral advantages on tasks like knowledge updates and temporal association.
Retrieval-augmented generation (RAG) has become the default strategy for providing large language model (LLM) agents with contextual knowledge. Yet RAG treats memory as a stateless lookup table: information persists indefinitely, retrieval is read-only, and temporal continuity is absent. We define the \textit{Continuum Memory Architecture} (CMA), a class of systems that maintain and update internal state across interactions through persistent storage, selective retention, associative routing, temporal chaining, and consolidation into higher-order abstractions. Rather than disclosing implementation specifics, we specify the architectural requirements CMA imposes and show consistent behavioral advantages on tasks that expose RAG's structural inability to accumulate, mutate, or disambiguate memory. The empirical probes (knowledge updates, temporal association, associative recall, contextual disambiguation) demonstrate that CMA is a necessary architectural primitive for long-horizon agents while highlighting open challenges around latency, drift, and interpretability.