The MUSE Benchmark: Probing Music Perception and Auditory Relational Reasoning in Audio LLMS

TL;DR

The MUSE Benchmark evaluates music perception and relational reasoning in audio large language models, revealing significant gaps compared to human performance and exposing perceptual deficits in current models.

Contribution

We introduce the MUSE Benchmark with 10 tasks to systematically assess fundamental music perception skills in audio LLMs, highlighting their limitations.

Findings

Wide variance in SOTA model capabilities

Persistent gap between models and human experts

Severe perceptual deficits in some models

Abstract

Multimodal Large Language Models (MLLMs) have demonstrated capabilities in audio understanding, but current evaluations may obscure fundamental weaknesses in relational reasoning. We introduce the Music Understanding and Structural Evaluation (MUSE) Benchmark, an open-source resource with 10 tasks designed to probe fundamental music perception skills. We evaluate four SOTA models (Gemini Pro and Flash, Qwen2.5-Omni, and Audio-Flamingo 3) against a large human baseline (N=200). Our results reveal a wide variance in SOTA capabilities and a persistent gap with human experts. While Gemini Pro succeeds on basic perception, Qwen and Audio Flamingo 3 perform at or near chance, exposing severe perceptual deficits. Furthermore, we find Chain-of-Thought (CoT) prompting provides inconsistent, often detrimental results. Our work provides a critical tool for evaluating invariant musical representations and driving development of more robust AI systems.