Abstract:The rapidly evolving multimodal Large Language Models (LLMs) urgently require new benchmarks to uniformly evaluate their performance on understanding and textually describing music. However, due to semantic gaps between Music Information Retrieval (MIR) algorithms and human understanding, discrepancies between professionals and the public, and low precision of annotations, existing music description datasets cannot serve as benchmarks. To this end, we present MuChin, the first open-source music description benchmark in Chinese colloquial language, designed to evaluate the performance of multimodal LLMs in understanding and describing music. We established the Caichong Music Annotation Platform (CaiMAP) that employs an innovative multi-person, multi-stage assurance method, and recruited both amateurs and professionals to ensure the precision of annotations and alignment with popular semantics. Utilizing this method, we built a dataset with multi-dimensional, high-precision music annotations, the Caichong Music Dataset (CaiMD), and carefully selected 1,000 high-quality entries to serve as the test set for MuChin. Based on MuChin, we analyzed the discrepancies between professionals and amateurs in terms of music description, and empirically demonstrated the effectiveness of annotated data for fine-tuning LLMs. Ultimately, we employed MuChin to evaluate existing music understanding models on their ability to provide colloquial descriptions of music. All data related to the benchmark, along with the scoring code and detailed appendices, have been open-sourced (<a class="link-external link-https" href="https://github.com/CarlWangChina/MuChin/" rel="external noopener nofollow">this https URL</a>).

MuSQL: a music structured query language

A Music Data Model and Its Application

PopMash: an Automatic Musical-Mashup System Using Computation of Musical and Lyrical Agreement for Transitions

MQML-Message Queuing Markup Language

Human-centric Music Medical Therapy Exploration System

QueST: querying music databases by acoustic and textual features.

Multimedia object query language and its query processing

The design and implementation of a digital music library

Hit-Dml: A Novel Digital Music Library

Design and implementation of HIT -DML

MSQL: Efficient Similarity Search in Metric Spaces Using SQL.

Multi-SQL: An extensible multi-model data query language

Seed-Music: A Unified Framework for High Quality and Controlled Music Generation

A Comprehensive Trainable Error Model for Sung Music Queries

OpenMU: Your Swiss Army Knife for Music Understanding

MuChin: A Chinese Colloquial Description Benchmark for Evaluating Language Models in the Field of Music

Music Files Search System

Compacting Music Signatures for Efficient Music Retrieval

A multi-paradigm querying approach for a generic multimedia database management system