Abstract
ZenBrain is an open-source memory architecture for autonomous AI systems implementing seven distinct memory layers inspired by cognitive neuroscience: working, short-term, episodic, long-term, procedural, core, and predictive memory. A unified MemoryCoordinator integrates fifteen validated neuroscience mechanisms — nine foundational algorithms (vmPFC-coupled FSRS spaced repetition with prediction-error gating, two-factor synaptic Hebbian dynamics for knowledge graphs, simulation-selection sleep consolidation modeling hippocampal replay, Bayesian confidence propagation, compositional context embeddings, spectral knowledge-graph health monitoring, context-adaptive information-bottleneck budgeting, dual-process chain-of-thought consolidation, and iMAD selective debate) and six Predictive Memory Architecture (PMA) components: NeuromodulatorEngine (dopamine/norepinephrine/serotonin/acetylcholine tonic levels with phasic prediction-error bursts and opponent-process dynamics), ReconsolidationEngine (memory destabilization on retrieval with prediction-error-gated updates and rollback support), TripleCopyMemory (fast/medium/deep copy divergence with exponential per-tier decay and rediscovery detection), PriorityMap (composite scoring across frequency, recency, emotional valence, and social signal with neuromodulator-boosted amygdala floor), StabilityProtector (NogoA/HDAC3-analog rigidity locks with age-proportional update thresholds), and MetacognitiveMonitor (confirmation/recency/asymmetry bias detection, urgency encoding, novelty windows, efficiency tracking). On LongMemEval-500, ZenBrain matches a long-context oracle's binary-judge accuracy within 4.5 percentage points at 1/106th of the per-query token cost, and wins all twelve head-to-head answer-quality cells against Letta, Mem0, and A-Mem under Bonferroni-corrected significance. The system is published as npm packages under the @zensation scope (zensation/algorithms, zensation/core) with 11,589 CI tests, and validated end-to-end on the LongMemEval and LoCoMo benchmarks.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Bering, Alexander, "ZenBrain: A Neuroscience-Inspired 7-Layer Memory Architecture for Autonomous AI Systems (v7)", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10021