---
title: "MMAC: A Massive Multi-dimensional Benchmark for Audio Captioning"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.27109v1"
permalink: "https://feed.astralor.com/items/itm_0be2e1bede7564ba9bac7de5"
published: "2026-07-29T16:38:08.000Z"
score: 75
channel: "tech"
tags: ["arXiv","MMAC","Benchmark","Evaluation Dimensions","AudioLLMs","Audio Captioning"]
---

# MMAC: A Massive Multi-dimensional Benchmark for Audio Captioning

## AI commentary

> What makes this paper stand out is that it pushes audio captioning evaluation from 'how good does it sound' to 'does it actually cover key information and avoid hallucination.' For teams developing or deploying audio LLMs, MMAC can serve as a health check for their models, revealing weak spots across dimensions. As an editor, I recommend everyone in multimodal evaluation keep an eye on the open-source release of this benchmark.

## At a glance

MMAC is a massive multi-dimensional benchmark proposed in an arXiv paper for audio captioning. It contains 5,638 audio clips from over 20 data sources, covering 6 capability categories and 15 evaluation dimensions. The benchmark checks whether model-generated captions mention relevant information in the target dimension and whether the content is consistent with reference labels. Evaluation results show significant differences across models in information coverage and description reliability. The benchmark and evaluation code will be open-sourced.

## Key points

- MMAC includes 5,638 audio clips from over 20 sources, covering 6 capability categories and 15 evaluation dimensions.
- The evaluation checks both information coverage and description reliability, more comprehensive than traditional single metrics.
- Open-source and proprietary audio LLMs show clear differences across multiple dimensions.
- The benchmark and code will be publicly released to standardize audio captioning evaluation.

## Full text

With the development of audio large language models (AudioLLMs), audio captioning needs to move from brief descriptions toward open-ended and fine-grained free-form descriptions. Existing evaluations often focus on generation quality or task performance, making it difficult to diagnose information coverage and description reliability. We propose MMAC, a \textbf{M}assive \textbf{M}ulti-dimensional benchmark for \textbf{A}udio \textbf{C}aptioning. MMAC contains 5,638 audio clips from more than 20 data sources, covering 6 capability categories and 15 evaluation dimensions. Given a model-generated caption, MMAC checks whether it mentions relevant information in the target dimension and whether the mentioned content is consistent with the reference label. We evaluate representative open-source and proprietary AudioLLMs. Results show clear differences across evaluation dimensions, information coverage, and description reliability. We will release the MMAC benchmark and evaluation code.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_0be2e1bede7564ba9bac7de5