Verifier-Guided Twelve-Tone Composition: A Generate-Verify-Repair Harness for Symbolic Music Generation

2026-07-13

AI Summary

This work addresses the frequent degradation in texture and lack of local and structural coherence observed in twelve-tone music generated by large language models. To mitigate these issues, the authors propose the first neuro-symbolic framework specifically designed for twelve-tone composition, embedding a large language model within a generate–verify–repair loop. This approach integrates a symbolic rule verifier and an explicit abstention mechanism to enhance local event consistency without enforcing global legality constraints. Experimental results demonstrate that the method increases the rate of valid outputs from 13.3% to 48.1% and improves sequence consistency pass rates from 33.5% to 58.3%, while maintaining a stable degradation rate of 0.05. Expert blind evaluations show a significant preference for the generated outputs over baseline approaches.

Abstract

Large language models can produce superficially legal twelve-tone scores that collapse into degenerate textures. We introduce a neuro-symbolic harness that wraps a language-model proposer in a generate-verify-repair-trace loop with symbolic verification. The complete pipeline improves event-local consistency without claiming whole-piece legality. Across 40 controlled tasks and four paired models, audited delivery yield rises from 13.3% under raw generation to 48.1% with the harness, which explicitly abstains otherwise. The pass rate of a narrower collision and serialisation-consistency check rises from 33.5% to 58.3%, while degeneracy remains near 0.05, including under exploratory adversarial prompting. A blinded evaluation by five experts also shows a descriptive aggregate preference for harness candidates over raw generation in adherence, perceived legality, coherence, and overall quality.

Problem

• twelve-tone composition
• degenerate textures
• event-local consistency
• serialisation consistency
• symbolic music generation

Innovation

• neuro-symbolic
• twelve-tone composition
• generate-verify-repair
• symbolic music generation
• consistency verification