Researchers have developed Decomposer, a novel framework designed to translate symbolic music, such as MIDI files, into executable programs within the Strudel music programming language. This system addresses the scarcity of paired data by creating a synthetic corpus called Strudel-Synth for initial training and then employing reinforcement learning with unpaired MIDI data to enhance both code readability and reconstruction accuracy. Decomposer reportedly outperforms closed-source large language models in MIDI reconstruction faithfulness and generates more diverse and readable code than existing heuristic converters. AI
IMPACT Could enable new forms of AI-assisted music creation and analysis by translating musical performances into editable code.
RANK_REASON Academic paper detailing a new AI model and framework for a specific task. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Decomposer
- Gotit.pub
- Hugging Face
- IArxiv
- MIDI
- ScienceCast
- Strudel-Synth
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
