To open up new opportunities for symbolic music generation using large-scale language models (LLMs), we propose HNote to address the complexity and structural consistency of existing formats. HNote is a hexadecimal-based notation that extends YNote and encodes pitch and duration within a fixed 32-unit measurement framework, providing direct compatibility with the LLM architecture. We converted 12,300 Gangnam Style folk songs into HNote and fine-tuned LLaMA-3.1(8B) using LoRA. Experimental results show that HNote achieves 82.5% syntactic accuracy and generates stylistically coherent compositions, demonstrating strong symbolic and structural similarity in BLEU and ROUGE evaluations.
