Research map / Architectures and efficiency / Attention and transformers
Museformer: Transformer with Fine- and Coarse-Grained Attention for Music Generation
Botao Yu, Peiling Lu, Rui Wang, Wei Hu, Xu Tan, Wei Ye, Shikun Zhang, Tao Qin, Tie-Yan Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use fine- and coarse-grained attention with similarity-selected structure-related bars | Museformer · Lakh MIDI test, first 10240 tokens | PPL: 1.64 -> 1.35 | Improved |
| Use fine- and coarse-grained attention with similarity-selected structure-related bars | Museformer · Lakh MIDI generated music | SE (%): 1.97 -> 0.95 | Improved |
| Use Museformer rather than the strongest comparison model | Lakh MIDI human evaluation · Human evaluation, Lakh MIDI samples | Overall: 6.44 ± 2.01 -> 7.12 ± 1.81 | Improved |
| Add coarse-grained attention to summarize non-structure-related bars (ablation) | Museformer · Lakh MIDI test, first 1024 tokens | PPL: 1.65 -> 1.64 | Improved |
| Add coarse-grained attention to summarize non-structure-related bars (ablation) | Museformer · Lakh MIDI test, first 5120 tokens | PPL: 1.42 -> 1.41 | Improved |
| Add coarse-grained attention to summarize non-structure-related bars (ablation) | Museformer · Lakh MIDI test, first 10240 tokens | PPL: 1.38 -> 1.35 | Improved |
| Add coarse-grained attention to summarize non-structure-related bars (ablation) | Museformer · Lakh MIDI generated music | SE (%): 1.08 -> 0.95 | Improved |
| Select fine-grained bars using similarity statistics rather than the most recent 8 bars (ablation) | Museformer · Lakh MIDI test, first 1024 tokens | PPL: 1.65 -> 1.64 | Improved |
| Select fine-grained bars using similarity statistics rather than the most recent 8 bars (ablation) | Museformer · Lakh MIDI test, first 5120 tokens | PPL: 1.43 -> 1.41 | Improved |
| Select fine-grained bars using similarity statistics rather than the most recent 8 bars (ablation) | Museformer · Lakh MIDI test, first 10240 tokens | PPL: 1.39 -> 1.35 | Improved |
| Select fine-grained bars using similarity statistics rather than the most recent 8 bars (ablation) | Museformer · Lakh MIDI generated music | SE (%): 6.39 -> 0.95 | Improved |
| Disable precise fine-grained attention to previous bars, retaining only summarized previous-bar information (ablation) | Museformer · Lakh MIDI music generation | coherence | Hurt |
Closest papers by meaning
- MuPT: A Generative Symbolic Music Pretrained Transformer (ICLR 2025 · similarity 0.90)
- Efficient Fine-Grained Guidance for Diffusion Model Based Symbolic Music Generation (ICML 2025 · similarity 0.89)
- MusER: Musical Element-Based Regularization for Generating Symbolic Music with Emotion (AAAI 2024 · similarity 0.88)
- MIDI-GPT: A Controllable Generative Model for Computer-Assisted Multitrack Music Composition (AAAI 2025 · similarity 0.88)
- Whole-Song Hierarchical Generation of Symbolic Music Using Cascaded Diffusion Models (ICLR 2024 · similarity 0.88)
- SongCreator: Lyrics-based Universal Song Generation (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
