用自然语言生成五线谱，不依赖嘈杂的文本配对数据

Developing text-driven symbolic music generation models remains challenging due to the scarcity of aligned text-music datasets and the unreliability of automated captioning pipelines. While most efforts have focused on MIDI, sheet music representations are largely underexplored in text-driven generation. We present \textit{Text2Score}, a two-stage framework comprising a planning stage and an execution stage for generating sheet music from natural language prompts. By deriving supervision signals directly from symbolic XML data, we propose an alternative to caption-based training that bypasses noisy or scarce text-music pairs. In the planning stage, an LLM orchestrator translates a natural language prompt into a structured bar-wise plan defining musical attributes such as instruments, key, time signatures, harmony, etc. This plan guides a generative model in the execution stage to produce interleaved ABC notation conditioned on its structural constraints. To assess output quality, we introduce an evaluation framework covering playability, readability, instrument utilization, structural complexity, and prompt adherence, corroborated by expert musicians. \textit{Text2Score} consistently outperforms both a pure LLM-based agentic framework and three end-to-end baselines across objective and subjective dimensions. We open-source the dataset, code, evaluation set and LLM prompts used in this work; a demo is available on our project page (https://keshavbhandari.github.io/portfolio/text2score).