Transparent and reproducible human-centered hybrid phrase logic transformer for musical expressivity

This repository contains a non-archival preprint of my manuscript submitted to the Computer Music Journal Special Issue on Critical Models. The paper proposes a human-centered hybrid system combining phrase-logic structures with a lightweight transformer decoder to model expressive musical phrasing with transparency and reproducibility. To protect the integrity of the peer-review process, supplementary materials(audio stimuli, AB listening test data, symbolic examples, and model components) are withheld from public release at this stage. These materials have been prepared in full and can be provided upon request to editors or reviewers.