NEST: A Graph-Theoretic Architecture Unifies Cognitive Models into One Framework

Ishant's NEST paper, submitted to arXiv in July 2026, proposes a graph-theoretic representational ontology that models cognition as structured state formation rather than a finished empirical model. At its core are typed, weighted graphs where nodes may carry subgraph payloads and edges fall into six relation classes: causal, containment, temporal, associative, evidential, and spatial. The architecture separates durable belief graphs from transient, capacity-limited working-memory graphs, and defines operators for activation, graph properties, working-memory transitions, awareness, and belief revision. Diagnostics like fragmentation, conflict, and coherence are derived within the same ontology.

Crucially, NEST provides compatibility mappings that embed existing cognitive architectures—including ACT-R, Soar, Sigma, the Common Model of Cognition, Global Workspace Theory, and chunking theories—as constrained regions of one unified language. This makes NEST a transparent substrate for future empirical, computational, and domain-specific work. The paper's contribution is foundational: it offers a canonical way to represent and compare diverse cognitive models, potentially enabling cross-architecture toolkits and benchmarks. While still theoretical, NEST could accelerate cognitive AI research by giving researchers a common ontological ground.