---
title: "MMAC：用于音频字幕生成的大规模多维基准"
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","基准测试","评估维度","音频大语言模型","音频字幕生成"]
---

# MMAC：用于音频字幕生成的大规模多维基准

## AI 点评

> 这篇论文最值得关注的地方，是它把音频字幕评估从‘好不好听’推进到了‘到底有没有覆盖关键信息、有没有胡说八道’。对于正在研发或部署音频LLM的团队来说，MMAC可以直接用来给自己的模型做‘体检’，看看哪些维度是短板。作为编辑，我建议所有从事多模态大模型评估的同学都关注一下这个基准的开源进展。

## 速览

MMAC是由arXiv论文提出的用于音频字幕生成的大规模多维基准测试，包含5638个音频片段，覆盖20多个数据源、6大能力类别和15个评估维度。该基准不仅检查模型生成的字幕是否提及目标维度的相关信息，还验证内容是否与参考标签一致。评估结果表明不同音频大语言模型在信息覆盖度和描述可靠性上存在显著差异。该基准和评估代码将开源。

## 关键点

- MMAC包含5638个音频片段，来源超20个，覆盖6大能力类别和15个评估维度。
- 评估方式同时检查信息覆盖度和描述可靠性，比传统单一指标更全面。
- 开源和闭源音频大语言模型在多项维度上表现差异明显。
- 基准和代码将公开发布，促进音频字幕评估标准化。

## 全文

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.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_0be2e1bede7564ba9bac7de5