面向AI生成音乐抄袭检测：作为版本识别问题
Towards AI-Generated Music Plagiarism Detection as a Version Identification Problem
- National Technical University of Athens（雅典国立技术大学）
- Athens University of Economics and Business（雅典经济与商业大学）
- Orfium
- Archimedes/Athena R.C.（Archimedes/Athena 研究中心）
机构由 AI 辅助整理，请以论文原文为准。
中文总结 AI 辅助
本研究将AI音乐抄袭检测视为版本识别问题，提出COPYCAT基准并验证坐标级嵌入偏移监督框架，将F0.5从0.612提升至0.803。
中文摘要 AI 辅助
文本到音乐生成模型的快速扩展挑战了音乐创作和知识产权的传统范式。在此背景下，抄袭很少是一个绝对的数学二元问题，而是一个在和声结构、旋律轮廓或整体感知风格特征上协商的模糊阈值。在本工作中，我们测试了最先进的音乐版本识别架构从人-人翻唱领域到人-AI抄袭场景的可迁移性。为评估该任务，我们引入了COPYCAT基准，该基准源自真实抄袭案例，并通过生成式再合成和数字信号处理混淆进行扩展，产生了350,654个评估对。我们表明，标量距离阈值化在生成式再合成下失效，而利用坐标级嵌入偏移的监督框架恢复了分散的抄袭信号，将整体$F_{0.5}$从$0.612$提升至$0.803$。
英文摘要
The rapid expansion of text-to-music generative models challenges traditional paradigms of music creation and intellectual property. Plagiarism in this context is rarely an absolute mathematical binary, but an ambiguous threshold negotiated over harmonic structure, melodic contours, or overall perceived stylistic character. In this work, we test the transferability of state-of-the-art music version identification architectures from the human-to-human cover domain to the human-to-AI plagiarism setting. To evaluate this task, we introduce COPYCAT, a benchmark derived from real-world plagiarism cases and extended through generative re-synthesis and digital signal processing obfuscations, yielding 350,654 evaluation pairs. We show that scalar distance thresholding collapses under generative re-synthesis, while a supervised framework leveraging coordinate-wise embedding shifts recovers the dispersed plagiarism signal, raising overall $F_{0.5}$ from $0.612$ to $0.803$.
