FUTGA

该数据集由FUTGA模型生成，提供了对全长歌曲的细粒度音乐解说，包括结构描述和时间边界。该数据集旨在增强现有音乐解说模型的时间推理能力，同时不影响模型的泛化能力。数据集涵盖了时长在2至5分钟之间的整首歌曲，适用于音乐解说和检索任务。

This dataset is generated by the FUTGA model, providing fine-grained musical commentary for full-length songs, including structural descriptions and temporal boundaries. It aims to enhance the temporal reasoning capabilities of existing musical commentary models without compromising their generalization ability. The dataset covers full songs with durations ranging from 2 to 5 minutes, and is applicable to musical commentary and music retrieval tasks.