🤖 AI Summary
Existing music representation formats (e.g., MIDI, MusicXML) exhibit structural complexity and poor human readability, hindering effective fine-tuning of large language models (LLMs) for music generation. To address this, we propose YNote—a minimalist, character-based notation system encoding pitch and note duration using only four ASCII characters—designed to balance human interpretability with machine learnability while drastically reducing sequence modeling complexity. We perform supervised fine-tuning of a character-level GPT-2 (124M) on a YNote-encoded dataset and evaluate outputs using BLEU and ROUGE metrics. Experimental results demonstrate that the model generates coherent, stylistically consistent musical phrases from prompts as short as two notes, achieving BLEU-4 and ROUGE-L scores of 0.883 and 0.766, respectively. This validates the efficacy and feasibility of lightweight symbolic representations for LLM-based music generation.
📝 Abstract
The field of music generation using Large Language Models (LLMs) is evolving rapidly, yet existing music notation systems, such as MIDI, ABC Notation, and MusicXML, remain too complex for effective fine-tuning of LLMs. These formats are difficult for both machines and humans to interpret due to their variability and intricate structure. To address these challenges, we introduce YNote, a simplified music notation system that uses only four characters to represent a note and its pitch. YNote's fixed format ensures consistency, making it easy to read and more suitable for fine-tuning LLMs. In our experiments, we fine-tuned GPT-2 (124M) on a YNote-encoded dataset and achieved BLEU and ROUGE scores of 0.883 and 0.766, respectively. With just two notes as prompts, the model was able to generate coherent and stylistically relevant music. We believe YNote offers a practical alternative to existing music notations for machine learning applications and has the potential to significantly enhance the quality of music generation using LLMs.