HCMA Paper 3: Toward Automated Expert Intuition — a Drosophila-Inspired Fast Associative Readout for HCMA

HCMA Paper 3: an associative fast-path for the Hippocampal-Cortical Memory Architecture, inspired by the Drosophila mushroom body. Adds a dopaminergic (DAN) outcome layer, a fast MBON readout, and a distributional signifier/signified encoder — all additive, without modifying the HCMA core. Deterministic, cost-free benchmark (FactChain-Synth + hard tier): • synonym accuracy 0.980 (95% Wilson CI [0.950, 0.992]) • known/unknown AUROC 0.931 (95% bootstrap CI [0.917, 0.944]) • calibration ECE 0.867 -> 0.132 (Platt) • DAN 0.985 / 0.000 stale vs frequency 0.000 / 0.998 stale • BM25 baseline: 0.200 synonym (current-only) vs 0.980 distributional • paired bootstrap paper2->paper3 on synonyms: +0.645 [0.575, 0.710] • verdicts: 6 PASS + 1 PARTIAL (H5 efficiency) License: CC-BY 4.0 (paper/docs); Apache-2.0 (code, SPDX headers inline). SHA-256 integrity manifest: manifest.json