![[2602.15074] Structure-Aware Piano Accompaniment via Style Planning and Dataset-Aligned Pattern Retrieval](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15074] Structure-Aware Piano Accompaniment via Style Planning and Dataset-Aligned Pattern Retrieval
Summary
This paper presents a structure-aware method for generating piano accompaniments using a transformer model for style planning and dataset-aligned pattern retrieval.
Why It Matters
The research addresses the challenge of creating high-quality piano accompaniments by combining machine learning with music theory. This approach enhances the creativity and efficiency of music composition, making it relevant for musicians, composers, and AI researchers in the field of generative music.
Key Takeaways
- Introduces a transformer-based model for style planning in piano accompaniment.
- Decouples high-level musical structure from note-level realization for better flexibility.
- Utilizes dataset-aligned pattern retrieval to enhance musical diversity and style adherence.
- Demonstrates effective generation of long-form accompaniments with strong style realization.
- Provides insights into planner ablations and quantifies inter-style isolation.
Computer Science > Sound arXiv:2602.15074 (cs) [Submitted on 16 Feb 2026] Title:Structure-Aware Piano Accompaniment via Style Planning and Dataset-Aligned Pattern Retrieval Authors:Wanyu Zang, Yang Yu, Meng Yu View a PDF of the paper titled Structure-Aware Piano Accompaniment via Style Planning and Dataset-Aligned Pattern Retrieval, by Wanyu Zang and 2 other authors View PDF HTML (experimental) Abstract:We introduce a structure-aware approach for symbolic piano accompaniment that decouples high-level planning from note-level realization. A lightweight transformer predicts an interpretable, per-measure style plan conditioned on section/phrase structure and functional harmony, and a retriever then selects and reharmonizes human-performed piano patterns from a corpus. We formulate retrieval as pattern matching under an explicit energy with terms for harmonic feasibility, structural-role compatibility, voice-leading continuity, style preferences, and repetition control. Given a structured lead sheet and optional keyword prompts, the system generates piano-accompaniment MIDI. In our experiments, transformer style-planner-guided retrieval produces diverse long-form accompaniments with strong style realization. We further analyze planner ablations and quantify inter-style isolation. Experimental results demonstrate the effectiveness of our inference-time approach for piano accompaniment generation. Comments: Subjects: Sound (cs.SD); Artificial Intelligence (cs.AI) Cite as: arXiv:...
