RDLNN-DCRAS: Non-Destructive Continual Learning via Hebbian Lateral Voting in a Biologically-Grounded Liquid Neural Cognitive Architecture

We present DCRAS (Distributed Coupled Reservoir Attractor System), the cross-columnarvoting and continual learning subsystem of the Recurrent Dendritic Liquid Neural Network(RDLNN) — a 10-module biologically-grounded cognitive architecture integrating LiquidState Machines, Amygdala fear conditioning, Basal Ganglia action selection, and Hippocampal episodic memory. DCRAS eliminates catastrophic forgetting for sequential associativelearning by avoiding the mechanism that causes it: unlike gradient-based approaches thatoverwrite shared weight configurations, DCRAS stores associative mappings as attractorsin fixed-weight LSM reservoirs coupled via Hebbian lateral voting matrices. New knowledgeaccumulates additively in high-dimensional orthogonal subspaces rather than competingdestructively for shared parameters.We validate this claim at three reservoir scales using a fully vectorized batched CUDAengine (B=8 seeds in parallel, Tesla T4). In an 8-task sequential aversive conditioningparadigm (CS-A through CS-H, 25 trials each), RDLNN-DCRAS achieves mean retention0.984 (±0.044) at N=512, and perfect 1.000 (±0.000) at both N=1024 and N=2048,with Backward Transfer BWT=+0.000 at the two larger scales. All 8 seeds pass the ≥0.875threshold at every scale. A blank-input saturation probe (CS-null) scores 0.000 avoidanceacross all 24 probe measurements (8 seeds × 3 scales), ruling out non-specific excitation.An LSTM baseline exhibits BWT=−0.458; LSTM+EWC achieves BWT=−0.048.We further demonstrate that RDLNN-DCRAS maintains latent representations of absentsensory modalities through continuous cross-modal constraint propagation, achieving 0.817cosine similarity to the reference attractor under complete sensory deprivation — withoutgenerative training, diffusion models, or autoregressive loss. A control agent with votingdisabled collapses to 0.287. These results establish RDLNN-DCRAS as a non-destructivecontinual learning system operating in a fundamentally different regime from gradient-basedsequential learners, with scale-invariant zero-forgetting confirmed across a 16× range ofreservoir size.