MusiXQA: Advancing Visual Music Understanding in Multimodal Large Language Models

TL;DR

MusiXQA introduces a new dataset and model to improve multimodal large language models' ability to interpret music sheets, addressing a previously underexplored area in visual music understanding.

Contribution

The paper presents MusiXQA, the first comprehensive dataset for music sheet understanding, and Phi-3-MusiX, a fine-tuned MLLM that significantly improves performance in this domain.

Findings

Current MLLMs have significant limitations in music sheet interpretation.

Phi-3-MusiX outperforms GPT-based methods on the MusiXQA dataset.

The dataset enables diverse visual QA tasks for music sheets.

Abstract

Multimodal Large Language Models (MLLMs) have achieved remarkable visual reasoning abilities in natural images, text-rich documents, and graphic designs. However, their ability to interpret music sheets remains underexplored. To bridge this gap, we introduce MusiXQA, the first comprehensive dataset for evaluating and advancing MLLMs in music sheet understanding. MusiXQA features high-quality synthetic music sheets generated via MusiXTeX, with structured annotations covering note pitch and duration, chords, clefs, key/time signatures, and text, enabling diverse visual QA tasks. Through extensive evaluations, we reveal significant limitations of current state-of-the-art MLLMs in this domain. Beyond benchmarking, we developed Phi-3-MusiX, an MLLM fine-tuned on our dataset, achieving significant performance gains over GPT-based methods. The proposed dataset and model establish a foundation for future advances in MLLMs for music sheet understanding. Code, data, and model will be released upon acceptance.