AURA: Dual-Memory Resonance Architecture for Bounded Meaning-State Generation — Reproducibility Package

This record contains the reproducibility package for the manuscript “AURA: A Dual-Memory Resonance Architecture for Bounded Meaning-State Generation in Memory-Bearing AI Systems.” The package includes the manuscript, supplementary material, simulation notebook, figures, and internal robustness-audit records for a simulation-stage dual-memory resonance architecture. The core experiments test whether the same present-state representation can produce different bounded meaning-state outputs under different continuity-weighted memory profiles, whether a bounded stabilising term reduces contradiction-shock depth and output jumps, and whether the original Trainable R mechanism improves scalar resonance magnitude without solving categorical anchor selection. The submitted notebook is the frozen reproducibility artifact for the numerical results. Any linked repository should be treated as an optional evolving demonstration space, not as the version-of-record for peer review.