Datasets:

PeacefulData
/

RMS-AQA

---
pretty_name: RMS-AQA
task_categories:
  - question-answering
language:
  - en
tags:
  - audio-question-answering
  - spatial-audio
  - first-order-ambisonics
  - benchmark
  - icassp-grand-challenge
extra_gated_prompt: Please login HuggingFace to register your email and research affiliation to get auto-approval.
extra_gated_fields:
  Full Name (required):
    type: text
  Email (required):
    type: text
  Affiliation (required):
    type: text
  Country:
    type: country
  I want to use this dataset for:
    type: select
    options:
      - Research
      - Education
      - Product Development
      - label: Other
        value: other
---
# RMS-AQA: Real-World Multi-Hop Spatial Audio Question Answering Challenge

[![Code Repository](https://img.shields.io/badge/RMS--AQA-Code%20Repository-444444?logo=github&logoColor=white)](https://github.com/rmsaqachallenge/rmsaqa-code)
[![Challenge Website](https://img.shields.io/badge/RMS--AQA-Challenge%20Website-0b7285?logo=githubpages&logoColor=white)](https://rmsaqachallenge.github.io)
[![arXiv Paper](https://img.shields.io/badge/arXiv-Paper-57A63A?logo=arxiv&logoColor=white)](https://arxiv.org/abs/2610.00935)

**A spatial audio question answering (SAQA) dataset for real-world domestic environments.**

## Overview

**RMS-AQA** is a spatial audio question answering benchmark built from 10-second First-Order Ambisonics (FOA) recordings collected in smart-home environments. The benchmark is organized as a two-stage question-answering task:

1. **Stage 1:** ground the audible sound events present in the scene.
2. **Stage 2:** perform complex spatio-temporal reasoning based on the Stage 1 answer as context.

All audio clips are provided in four-channel, **24 kHz, 10-second FOA format**, and every question answering (QA) pair is a multiple-choice question **(MCQ)** generated by **GPT-5.5** and paraphrased by **Gemini 3.1**. The dataset is split into training, validation, and test sets: the training set is fully synthetic, the validation set combines synthetic and recorded examples, and the hidden evaluation set consists entirely of recorded data. Simulated audio is generated with **SpatialScaper** by convolving dry sound events with measured spatial room impulse responses and the recorded audios are captured with a **Zoom H3-VR** recorder. 

Stage 2 covers six question dimensions:

- **Sound counting (SC)**
- **Spatial location (SL)**
- **Temporal detection (TD)**
- **Temporal relation (TR)**
- **Spatial relation (SR)**
- **Action prediction (AP)**

## Dataset Structure

Each record contains a 10-second audio segment and its two-stage MCQ pair. Stage 1 identifies the sound events present in the scene; Stage 2 asks a complex spatio-temporal question conditioned on the Stage 1 context.The train set contains **170K** clips and **170K** QA pairs and the validation set is made up of **5K** audio clips with **6K** QA pairs. The hidden test set includes **2K** clips with **3K** QA pairs, which will be released on **Nov. 24, 2026**.

### QA example

```json
{
  "records": [
    {
      "segment_id": "train-000000.json",
      "qa_pairs": [
        {
          "stage": 1,
          "question": "What types of sound events can be heard in this audio?",
          "answer": "D",
          "options": {
            "A": "The recording includes key drop.",
            "B": "In the audio, you can hear dishwasher.",
            "C": "The clip contains indoor cricket chorus.",
            "D": "The audible sound types are child shouting."
          }
        },
        {
          "stage": 2,
          "question": "If a domestic robot were monitoring this audio, what should its response plan be?",
          "answer": "D",
          "options": {
            "A": "The best next step is to log and continue passive monitoring for dishwasher near the front-left area.",
            "B": "For this clip, the embodied assistant should trigger a home emergency alert, notify, and monitor safely for child shouting near the front-left area.",
            "C": "The appropriate response plan is to trigger a home emergency alert, notify, and monitor safely for dishwasher near the rear area.",
            "D": "Because of the detected sound, the robot should make a non-intrusive check and notify for child shouting near the rear area."
          },
          "category_id": 6,
          "category_name": "Action Prediction"
        }
      ]
    }
  ]
}
```

## Download

With [huggingface_hub](https://huggingface.co/docs/huggingface_hub) (recommended):

```bash
pip install huggingface_hub
huggingface-cli download PeacefulData/RMS-AQA --repo-type dataset --local-dir ./RMS-AQA
```

With Python:

```python
from huggingface_hub import snapshot_download
snapshot_download(repo_id="PeacefulData/RMS-AQA", repo_type="dataset", local_dir="./RMS-AQA")
```

Or with git (requires [git-lfs](https://git-lfs.com)):

```bash
git clone https://huggingface.co/datasets/PeacefulData/RMS-AQA
```

## Folder Layout

```text
RMS-AQA/
├── train_QA/                  # training set QA pairs (train_QA.tar.zst)
├── train_audio/               # training set audio shards (train-*.tar.zst)
├── dev_QA/                    # development set QA pairs (dev_QA.tar.zst)
└── dev_audio/                 # development set audio shards (dev-*.tar.zst)
```

The `*_QA` archives contain per-sample JSON files in the schema shown above; the `*_audio` archives contain the corresponding four-channel FOA clips.

## License

The simulated audio in RMS-AQA incorporates material from the third-party datasets listed below. License terms may vary between recordings within a dataset. This table summarizes the relevant licenses; users should consult the original sources for attribution and other applicable requirements.

| Dataset | License |
| --- | --- |
| [TAU-SRIR DB](https://zenodo.org/records/6408611) | [TUT Non-Commercial](https://zenodo.org/records/6408611) |
| [METU-SPARG](https://zenodo.org/records/2635758) | [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode) |
| [ARNI-SRIR](https://zenodo.org/records/6985104) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [FOA-MEIR](https://zenodo.org/records/6088574) | [NTT Non-Commercial](https://github.com/nttrd-mdlab/seld-foa-meir/blob/main/LICENSE.pdf) |
| [AudioSet](https://research.google.com/audioset/download.html) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [FSD50K](https://zenodo.org/records/4060432) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [ESC-50](https://github.com/karolpiczak/ESC-50) | [CC-BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/legalcode) |
| [CHiME-Home](https://archive.org/details/chime-home) | [CC-BY-NC-SA 3.0](https://creativecommons.org/licenses/by-nc-sa/3.0/legalcode) |
| [C4DM](http://isophonics.net/content/room-impulse-response-data-set) | [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode) |
| [HOMULA-RIR](https://zenodo.org/records/15053008) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [HiFi-HARP](https://huggingface.co/datasets/whojavumusic/hifi_harp) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [RSoANU](https://zenodo.org/records/10720345) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [cube_soundfield](https://phaidra.kug.ac.at/detail/o:104435) | [Public Domain Mark](https://creativecommons.org/publicdomain/mark/1.0/) |
| [MOTUS](https://zenodo.org/records/4923187) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [DataSEC](https://zenodo.org/records/23097290) | [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode) |
| [TUT_Sound_Events_2016](https://zenodo.org/records/45759) | [TUT Academic EULA (Non-Commercial)](https://raw.githubusercontent.com/TUT-ARG/DCASE2016-baseline-system/master/EULA.pdf) |
| [TUT_SoundEvents_2017](https://zenodo.org/records/400516) | [TUT Academic EULA (Non-Commercial)](https://raw.githubusercontent.com/TUT-ARG/DCASE2017-baseline-system/master/EULA.pdf) |
| [TUT_RareSoundEvents_2017 Target Sound](https://zenodo.org/records/401395) | [CC-BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/legalcode), [CC-BY 3.0](https://creativecommons.org/licenses/by/3.0/legalcode), [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/legalcode), [Sampling+ 1.0](https://creativecommons.org/licenses/sampling+/1.0/) |
| [DCASE 2016 Task 2](https://archive.org/details/dcase2016_task2_train_dev) | [CC-BY 3.0](https://creativecommons.org/licenses/by/3.0/legalcode) |
| [SINS Node 1](https://zenodo.org/records/2546677#.XFR-KlVKhhE) | [CC-BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/legalcode) |
| [UrbanSound8K](https://urbansounddataset.weebly.com/urbansound8k.html) | [CC-BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/legalcode) |
| [GISE-51](https://zenodo.org/records/4593514) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [ReaLISED](https://zenodo.org/records/6488321) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [AID](https://zenodo.org/records/6974033) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [FSDKaggle2018](https://zenodo.org/records/2552860) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [FSDnoisy18k](https://zenodo.org/records/2529934) | [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) |
| [DESED](https://github.com/turpaultn/DESED) | [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/legalcode), [CC-BY 3.0](https://creativecommons.org/licenses/by/3.0/legalcode), [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode), [CC-BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/legalcode), [CC-BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/legalcode), [Sampling+ 1.0](https://creativecommons.org/licenses/sampling+/1.0/) |

In addition, the RMS-AQA dataset as a whole is the result of a curation process and it has an additional license. The RMS-AQA dataset is released under [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode).

## Citation

If you use RMS-AQA in your research, please cite:

```bibtex
@misc{chen2026rmsaqatwostagespatialaudio,
      title={RMS-AQA: A Two-Stage Spatial Audio Question Answering Benchmark for Real-World Domestic Environments},
      author={Peihao Chen and Qing Wang and Lichun Fan and Yufeng Hao and Zhifeng Kong and Mengyao Zhu and Hengyi Hong and Hang Chen and Hang Su and Yujie Jian and Chao-Han Huck Yang and Shichao Hu and Jun Du and Jian Luan and Ke Li},
      year={2026},
      eprint={2610.00935},
      archivePrefix={arXiv},
      primaryClass={cs.SD},
      url={https://arxiv.org/abs/2610.00935}
}
```

	---
	pretty_name: RMS-AQA
	task_categories:
	- question-answering
	language:
	- en
	tags:
	- audio-question-answering
	- spatial-audio
	- first-order-ambisonics
	- benchmark
	- icassp-grand-challenge
	extra_gated_prompt: Please login HuggingFace to register your email and research affiliation to get auto-approval.
	extra_gated_fields:
	Full Name (required):
	type: text
	Email (required):
	type: text
	Affiliation (required):
	type: text
	Country:
	type: country
	I want to use this dataset for:
	type: select
	options:
	- Research
	- Education
	- Product Development
	- label: Other
	value: other
	---

	# RMS-AQA: Real-World Multi-Hop Spatial Audio Question Answering Challenge

	[![Code Repository](https://img.shields.io/badge/RMS--AQA-Code%20Repository-444444?logo=github&logoColor=white)](https://github.com/rmsaqachallenge/rmsaqa-code)
	[![Challenge Website](https://img.shields.io/badge/RMS--AQA-Challenge%20Website-0b7285?logo=githubpages&logoColor=white)](https://rmsaqachallenge.github.io)
	[![arXiv Paper](https://img.shields.io/badge/arXiv-Paper-57A63A?logo=arxiv&logoColor=white)](https://arxiv.org/abs/2610.00935)

	A spatial audio question answering (SAQA) dataset for real-world domestic environments.

	## Overview

	RMS-AQA is a spatial audio question answering benchmark built from 10-second First-Order Ambisonics (FOA) recordings collected in smart-home environments. The benchmark is organized as a two-stage question-answering task:

	1. Stage 1: ground the audible sound events present in the scene.
	2. Stage 2: perform complex spatio-temporal reasoning based on the Stage 1 answer as context.

	All audio clips are provided in four-channel, 24 kHz, 10-second FOA format, and every question answering (QA) pair is a multiple-choice question (MCQ) generated by GPT-5.5 and paraphrased by Gemini 3.1. The dataset is split into training, validation, and test sets: the training set is fully synthetic, the validation set combines synthetic and recorded examples, and the hidden evaluation set consists entirely of recorded data. Simulated audio is generated with SpatialScaper by convolving dry sound events with measured spatial room impulse responses and the recorded audios are captured with a Zoom H3-VR recorder.

	Stage 2 covers six question dimensions:

	- Sound counting (SC)
	- Spatial location (SL)
	- Temporal detection (TD)
	- Temporal relation (TR)
	- Spatial relation (SR)
	- Action prediction (AP)

	## Dataset Structure

	Each record contains a 10-second audio segment and its two-stage MCQ pair. Stage 1 identifies the sound events present in the scene; Stage 2 asks a complex spatio-temporal question conditioned on the Stage 1 context.The train set contains 170K clips and 170K QA pairs and the validation set is made up of 5K audio clips with 6K QA pairs. The hidden test set includes 2K clips with 3K QA pairs, which will be released on Nov. 24, 2026.

	### QA example

	```json
	{
	"records": [
	{
	"segment_id": "train-000000.json",
	"qa_pairs": [
	{
	"stage": 1,
	"question": "What types of sound events can be heard in this audio?",
	"answer": "D",
	"options": {
	"A": "The recording includes key drop.",
	"B": "In the audio, you can hear dishwasher.",
	"C": "The clip contains indoor cricket chorus.",
	"D": "The audible sound types are child shouting."
	}
	},
	{
	"stage": 2,
	"question": "If a domestic robot were monitoring this audio, what should its response plan be?",
	"answer": "D",
	"options": {
	"A": "The best next step is to log and continue passive monitoring for dishwasher near the front-left area.",
	"B": "For this clip, the embodied assistant should trigger a home emergency alert, notify, and monitor safely for child shouting near the front-left area.",
	"C": "The appropriate response plan is to trigger a home emergency alert, notify, and monitor safely for dishwasher near the rear area.",
	"D": "Because of the detected sound, the robot should make a non-intrusive check and notify for child shouting near the rear area."
	},
	"category_id": 6,
	"category_name": "Action Prediction"
	}
	]
	}
	]
	}
	```

	## Download

	With [huggingface_hub](https://huggingface.co/docs/huggingface_hub) (recommended):

	```bash
	pip install huggingface_hub
	huggingface-cli download PeacefulData/RMS-AQA --repo-type dataset --local-dir ./RMS-AQA
	```

	With Python:

	```python
	from huggingface_hub import snapshot_download

	snapshot_download(repo_id="PeacefulData/RMS-AQA", repo_type="dataset", local_dir="./RMS-AQA")
	```

	Or with git (requires [git-lfs](https://git-lfs.com)):

	```bash
	git clone https://huggingface.co/datasets/PeacefulData/RMS-AQA
	```

	## Folder Layout

	```text
	RMS-AQA/
	├── train_QA/ # training set QA pairs (train_QA.tar.zst)
	├── train_audio/ # training set audio shards (train-*.tar.zst)
	├── dev_QA/ # development set QA pairs (dev_QA.tar.zst)
	└── dev_audio/ # development set audio shards (dev-*.tar.zst)
	```

	The `_QA` archives contain per-sample JSON files in the schema shown above; the `_audio` archives contain the corresponding four-channel FOA clips.

	## License

	The simulated audio in RMS-AQA incorporates material from the third-party datasets listed below. License terms may vary between recordings within a dataset. This table summarizes the relevant licenses; users should consult the original sources for attribution and other applicable requirements.

	\| Dataset \| License \|
	\| --- \| --- \|
	\| [TAU-SRIR DB](https://zenodo.org/records/6408611) \| [TUT Non-Commercial](https://zenodo.org/records/6408611) \|
	\| [METU-SPARG](https://zenodo.org/records/2635758) \| [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode) \|
	\| [ARNI-SRIR](https://zenodo.org/records/6985104) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [FOA-MEIR](https://zenodo.org/records/6088574) \| [NTT Non-Commercial](https://github.com/nttrd-mdlab/seld-foa-meir/blob/main/LICENSE.pdf) \|
	\| [AudioSet](https://research.google.com/audioset/download.html) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [FSD50K](https://zenodo.org/records/4060432) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [ESC-50](https://github.com/karolpiczak/ESC-50) \| [CC-BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/legalcode) \|
	\| [CHiME-Home](https://archive.org/details/chime-home) \| [CC-BY-NC-SA 3.0](https://creativecommons.org/licenses/by-nc-sa/3.0/legalcode) \|
	\| [C4DM](http://isophonics.net/content/room-impulse-response-data-set) \| [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode) \|
	\| [HOMULA-RIR](https://zenodo.org/records/15053008) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [HiFi-HARP](https://huggingface.co/datasets/whojavumusic/hifi_harp) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [RSoANU](https://zenodo.org/records/10720345) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [cube_soundfield](https://phaidra.kug.ac.at/detail/o:104435) \| [Public Domain Mark](https://creativecommons.org/publicdomain/mark/1.0/) \|
	\| [MOTUS](https://zenodo.org/records/4923187) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [DataSEC](https://zenodo.org/records/23097290) \| [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode) \|
	\| [TUT_Sound_Events_2016](https://zenodo.org/records/45759) \| [TUT Academic EULA (Non-Commercial)](https://raw.githubusercontent.com/TUT-ARG/DCASE2016-baseline-system/master/EULA.pdf) \|
	\| [TUT_SoundEvents_2017](https://zenodo.org/records/400516) \| [TUT Academic EULA (Non-Commercial)](https://raw.githubusercontent.com/TUT-ARG/DCASE2017-baseline-system/master/EULA.pdf) \|
	\| [TUT_RareSoundEvents_2017 Target Sound](https://zenodo.org/records/401395) \| [CC-BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/legalcode), [CC-BY 3.0](https://creativecommons.org/licenses/by/3.0/legalcode), [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/legalcode), [Sampling+ 1.0](https://creativecommons.org/licenses/sampling+/1.0/) \|
	\| [DCASE 2016 Task 2](https://archive.org/details/dcase2016_task2_train_dev) \| [CC-BY 3.0](https://creativecommons.org/licenses/by/3.0/legalcode) \|
	\| [SINS Node 1](https://zenodo.org/records/2546677#.XFR-KlVKhhE) \| [CC-BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/legalcode) \|
	\| [UrbanSound8K](https://urbansounddataset.weebly.com/urbansound8k.html) \| [CC-BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/legalcode) \|
	\| [GISE-51](https://zenodo.org/records/4593514) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [ReaLISED](https://zenodo.org/records/6488321) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [AID](https://zenodo.org/records/6974033) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [FSDKaggle2018](https://zenodo.org/records/2552860) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [FSDnoisy18k](https://zenodo.org/records/2529934) \| [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) \|
	\| [DESED](https://github.com/turpaultn/DESED) \| [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/legalcode), [CC-BY 3.0](https://creativecommons.org/licenses/by/3.0/legalcode), [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode), [CC-BY-NC 3.0](https://creativecommons.org/licenses/by-nc/3.0/legalcode), [CC-BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/legalcode), [Sampling+ 1.0](https://creativecommons.org/licenses/sampling+/1.0/) \|

	In addition, the RMS-AQA dataset as a whole is the result of a curation process and it has an additional license. The RMS-AQA dataset is released under [CC-BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode).

	## Citation

	If you use RMS-AQA in your research, please cite:

	```bibtex
	@misc{chen2026rmsaqatwostagespatialaudio,
	title={RMS-AQA: A Two-Stage Spatial Audio Question Answering Benchmark for Real-World Domestic Environments},
	author={Peihao Chen and Qing Wang and Lichun Fan and Yufeng Hao and Zhifeng Kong and Mengyao Zhu and Hengyi Hong and Hang Chen and Hang Su and Yujie Jian and Chao-Han Huck Yang and Shichao Hu and Jun Du and Jian Luan and Ke Li},
	year={2026},
	eprint={2610.00935},
	archivePrefix={arXiv},
	primaryClass={cs.SD},
	url={https://arxiv.org/abs/2610.00935}
	}
	```