README.md

---
license: mit
task_categories:
- question-answering
language:
- en
dataset_info:
  features:
  - name: id
    dtype: string
  - name: task_name
    dtype: string
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
  - name: question
    dtype: string
  - name: choice_a
    dtype: string
  - name: choice_b
    dtype: string
  - name: choice_c
    dtype: string
  - name: choice_d
    dtype: string
  - name: answer_gt
    dtype: string
  - name: category
    dtype: string
  - name: sub-category
    dtype: string
  - name: sub-sub-category
    dtype: string
  - name: linguistics_sub_discipline
    dtype: string
  splits:
  - name: train
    num_bytes: 1199569150.0
    num_examples: 5000
  download_size: 1466894219
  dataset_size: 1199569150.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

# [ICLR 2026] MMSU: A Massive Multi-task Spoken Language Understanding and Reasoning Benchmark


[![Paper](https://img.shields.io/badge/arxiv-%20PDF-red)](https://arxiv.org/pdf/2506.04779) [![Project](https://img.shields.io/badge/Project-Page-green)](https://github.com/dingdongwang/MMSU)


![Pipeline](intro.png)

## Overview of MMSU
MMSU (Massive Multi-task Spoken Language Understanding and Reasoning Benchmark) is a comprehensive benchmark for evaluating fine-grained spoken language understanding and reasoning in multimodal models. 

It systematically captures the variance of real-world linguistic phenomena in daily speech through **47 sub-tasks**, including phonetics, prosody, rhetoric, syntactics, semantics, and paralinguistics, spanning both perceptual and higher-level reasoning capabilities. 

The benchmark comprises **5,000 carefully curated audio–question–answer pairs** derived from diverse authentic recordings.

![Pipeline](benchmark.png)

## Usage
You can load the dataset via Hugging Face datasets:

```
from datasets import load_dataset
ds = load_dataset("ddwang2000/MMSU")
```

For evaluation, please refer to [**GitHub Code**](https://github.com/dingdongwang/MMSU)


## Citation
```
@article{wang2025mmsu,
      title={MMSU: A Massive Multi-task Spoken Language Understanding and Reasoning Benchmark}, 
      author={Dingdong Wang and Jincenzi Wu and Junan Li and Dongchao Yang and Xueyuan Chen and Tianhua Zhang and Helen Meng},
      journal={arXiv preprint arXiv:2506.04779},
      year={2025},
}
```