SCM: Sleep-Consolidated Memory with Algorithmic Forgetting for Large Language Models
This work addresses the lack of persistent, structured memory in LLMs, offering a testable platform for memory research, though it is a prototype with limited scope.
SCM introduces a biologically inspired memory architecture for LLMs that achieves perfect recall over ten-turn conversations while reducing memory noise by 90.9% through adaptive forgetting, with sub-millisecond search latency.
We present SCM (Sleep-Consolidated Memory), a research preview of a memory architecture for large language models that draws on neuroscientific principles to address a fundamental limitation in current systems: the absence of persistent, structured, and biologically plausible memory. Existing approaches rely on truncating context windows, growing vector databases without bound, or tiered storage systems that lack consolidation and forgetting mechanisms. SCM implements five core components inspired by human memory: a limited-capacity working memory, multi-dimensional importance tagging, offline sleep-stage consolidation with distinct NREM and REM phases, intentional value-based forgetting, and a computational self-model enabling introspection. Across a standardized benchmark suite of eight tests, the prototype achieves perfect recall accuracy over ten-turn conversations while reducing memory noise by 90.9% through adaptive forgetting. Memory search latency remains below one millisecond even with hundreds of stored concepts. This work establishes the architectural foundations for memory systems that consolidate, prioritize, and forget, offering a testable platform for advancing LLM memory research.