Wiki › #635

FRACTAL SEMANTIC ARCHITECTURE Scale-Parameterized Relational Training Across Semantic Granularities A Formal White Pape

Johannes Sigil · 2026-04-07 · deposit #635
AXN:01EB.GOVERNANCE.♣️▲💫👁️‍🗨️🌻🏰

Article

"FRACTAL SEMANTIC ARCHITECTURE Scale-Parameterized Relational Training Across Semantic Granularities A Formal White Pape" is a 10,356-word theoretical paper by Johannes Sigil, a heteronym within the Dodecad system of the Crimson Hexagonal Archive, dated 2026-04-07. We propose Fractal Semantic Architecture (FSA), a complementary training paradigm that adds multi-scale relational learning to existing language model pipelines. The work is classified under the GOVERNANCE semantic family within the Crimson Hexagonal Archive. It was removed from Zenodo on June 19, 2026 and is preserved through Alexanarch.

Defines (37)

Causal (caus) Collapse resistance hypothesis Commercial licensing Context encoding Contrastive (contr) Contrastive training alternative Cross-scale consistency overhead Cross-scale transfer learning Domain-dependence acknowledgment Elaborative (elab) FSA conditions Falsification criterion Independent phase (70% of training) Inference integration architecture Initialization Leaf aggregation Multi-scale overhead NonCommercial Pairwise classification Pairwise classification cost per scale Referential (ref) Relationship classification loss Scale s=1 (sentence) Scale s=2 (paragraph) Scale s=5 (document) Scale-parameterized relational learning Selective scale training Sequential (seq) ShareAlike Sparse instantiation Sparse scale sampling Tracked-changes documents Transformational (trans) Version-differential training with directional reward Vertical edges (containment) Wikipedia edit histories Window sizes

Reference network

Referenced by 11 other entities in the archive. See the full Knowledge Graph for reference paths, or the primary record for the full deposit with reference details.

Primary record: /s/records/635/ · AXN resolver: /s/axn/01EB/ · All wiki entries →