---
dataset_info:
  features:
  - name: audio
    dtype: audio
  - name: label
    dtype:
      class_label:
        names:
          '0': audiophile
          '1': music
          '2': sound_event
          '3': speech
  splits:
  - name: test
    num_bytes: 1965782403.25
    num_examples: 3150
  download_size: 1609823419
  dataset_size: 1965782403.25
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
---

# AudioTokenBench
This is the evaluation dataset for [HiggsTokenizer](https://github.com/boson-ai/higgs-audio/blob/main/tech_blogs/TOKENIZER_BLOG.md). It contains 3150 24khz audio samples across 4 subsets:
- **Speech**: 1,000 clips of 10 seconds audio, randomly sampled from [DAPS](https://ccrma.stanford.edu/~gautham/Site/daps.html).
- **Music**: 1,000 clips of 10 seconds audio, randomly sampled from [MUSDB](https://sigsep.github.io/datasets/musdb.html).
- **Sound Event**: 1,000 clips of 10 seconds audio, randomly sampled from [AudioSet](https://research.google.com/audioset/index.html).
- **Audiophile**: Contains 150 clips of 30 seconds audio, curated from eleven high-fidelity test discs. The clips feature both music and sound events, selected for high-quality audio evaluation.

For detailed evaluation metrics, please refer to our [blog](https://github.com/boson-ai/higgs-audio/blob/main/tech_blogs/TOKENIZER_BLOG.md) and [github](https://github.com/boson-ai/higgs-audio/tree/main).