arXiv：cs.LG（机器学习，全量分类）· Jiawei Yu, Jian Liu·· 14 小时前AI 评分35
TrueMuse：文本到音乐模型的数据归因基准
TrueMuse: A Benchmark for Data Attribution in Text-to-Music Models
AI 导读
研究者推出 TrueMuse，一个面向文本到音乐模型数据归因的受控数据集与基准，通过在精心筛选的归因样本上微调三个扩散模型来提供可控的归因目标。该基准覆盖旋律结构、音色特征、艺术家风格与流派模式四类归因设置，包含 133 个属性、648 个微调模型和 95,456 个生成样本。系统评估显示现有黑盒归因方法在不同设置下差异显著，归因仍具挑战性。
正文
Abstract:Text-to-music generation models are trained on massive music collections, creating a growing need for data attribution methods that can quantify the contribution of individual training samples. However, existing attribution methods are difficult to rigorously evaluate due to the lack of reliable ground truth, making it challenging to reliably assess their actual effectiveness. To address this gap, we introduce TrueMuse, a controlled dataset and benchmark for text-to-music data attribution. TrueMuse is constructed by fine-tuning three diffusion-based text-to-music models on carefully curated attribution samples, whose known inclusion in fine-tuning provides controlled attribution targets for evaluation. The benchmark covers four attribution settings, spanning melodic structure, timbral characteristics, artist-level stylistic signatures, and genre-level shared patterns, and includes 133 attributes, 648 fine-tuned models, and 95,456 generated samples across two prompt types. Using TrueMuse, we systematically evaluate existing black-box attribution methods along four dimensions: fine-tuning improvement, prompt-type difficulty, multi-task training, and fine-tuning data size. Our results show that attribution remains challenging, with existing methods exhibiting substantial variation across evaluation settings, highlighting the need for more reliable and generalizable attribution methods for text-to-music generation. Code and Dataset will be released upon acceptance.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.00835 [cs.LG] |
| (or arXiv:2610.00835v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.00835 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jiawei Yu [view email]
[v1]
Wed, 30 Sep 2026 23:48:15 UTC (349 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org