用验证修复循环提升十二音音乐生成的合法性与一致性
Verifier-Guided Twelve-Tone Composition: A Generate-Verify-Repair Harness for Symbolic Music Generation
- 构建生成-验证-修复-追溯闭环，结合符号验证确保局部合规
- 约束满足率从13.3%升至48.1%，退化率维持在0.05以下
- 适合追求严谨结构的作曲生成与音乐形式化研究者
大型语言模型可生成表面合法的十二音谱例，但常坍缩为退化纹理。我们提出一种神经符号框架，将语言模型生成器嵌入生成-验证-修复-追溯循环，并引入符号验证机制。该流程显著提升事件级一致性，不保证整部作品合法性。在40个控制任务与四组对比模型中，经约束检查的交付率从原始生成的13.3%提升至48.1%，其余51.9%则选择弃权。更严格的冲突与序列一致性检查通过率从33.5%升至58.3%，退化率始终低于0.05，包括对抗性提示下。五位专家盲评显示，该框架生成样本在合规性、感知合法性、连贯性与整体质量上均获更高偏好。
原文摘要 · Abstract (English)
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, constraint-checked delivery rises from 13.3% under raw generation to 48.1% with the harness; it abstains on the remaining 51.9% of runs. 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 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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。