Dataset · PeacefulData
Spatial audio and household scene questions
PeacefulData/RMS-AQA
English benchmark for evaluating spatial sound understanding and multi-hop reasoning.
What's inside
The dataset is a benchmark for spatial audio question answering (SAQA) in real-world home environments. It features 10-second First-Order Ambisonics (FOA) audio recordings collected in a smart home environment along with English questions. The task is split into two stages: first, localizing audible sound events in the scene, and then answering the question.
What you can do with it
- Evaluating audio-language models on spatial sound understanding
- Training smart home systems to recognize and associate household sounds
- Testing multi-hop QA reasoning on audio data
- Tasks
- Question answering
- Languages
- English
- Downloads this month
- 10.1K
- Likes
- 15
- Access
- on request on Hugging Face
How to load it
from datasets import load_dataset
ds = load_dataset("PeacefulData/RMS-AQA")