Litcius/Paper detail

Fractal Semantic Architecture: Scale-Parameterized Relational Training Across Semantic Granularities (v2.2)

Nobel Glas, Talos Morrow, Johannes Sigil

2026Zenodo (CERN European Organization for Nuclear Research)8 citationsDOIOpen Access PDF

Abstract

A formal white paper proposing a complementary training paradigm for neural language models that adds multi-scale relational learning and version-differential training to existing generative architectures. FSA defines a parameterized family of relational classifiers operating on typed relationships between semantic units at variable granularity (sentence through version-sequence), with automated extraction pipelines, bidirectional cross-scale consistency constraints, and inference-time integration via skeleton planning, coherence gating, revision scoring, and reranking. Includes a falsifiable experimental design for testing the hypothesis that discrete relational structures improve collapse resistance under recursive synthetic retraining.

Topics & Concepts

Computer scienceArtificial intelligenceGranularityNatural language processingFalsifiabilityGenerative grammarConsistency (knowledge bases)Parameterized complexityRelational databaseCoherence (philosophical gambling strategy)JoinsGenerative modelSemantic propertyRelation (database)Machine learningSemantics (computer science)Training setRelational calculusArtificial neural networkTheoretical computer scienceSemantic equivalenceRelationship extractionCategorizationSemantic data modelRelational algebraVariable (mathematics)Semantic similarityTopic ModelingNatural Language Processing TechniquesLanguage and cultural evolution