Datasets:

Slenser0
/

CodexValidationTiny_20260531

license: cc-by-4.0
language:
  - en
pretty_name: Codex Validation Tiny 20260531
task_categories:
  - audio-classification
size_categories:
  - n<1K
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*.parquet
tags:
  - anti-spoofing
  - audio-deepfake-detection
  - speech
  - benchmark
  - arena-ready
paperswithcode_id: null
arxiv:
  - 1911.01601

Codex Validation Tiny 20260531

Synthetic four-row dataset used to validate the dataset-add and Arena display workflow.

Overview

Contains two bonafide and two spoof one-second synthetic audio clips generated locally.

License & redistribution

CC-BY-4.0 for the synthetic generated audio in this validation dataset.

Schema

Field	Type	Description
path	string	Stable archive-relative path, unique within dataset.
audio	Audio(16kHz mono)	One-second generated WAV audio.
label	ClassLabel[bonafide, spoof]	Index 0 = bonafide.
notes	string (JSON)	Contains `utterance_id`.

Quick Start

from datasets import load_dataset
ds = load_dataset("Slenser0/CodexValidationTiny_20260531", split="test")

Stats

n_total	n_bonafide	n_spoof	total duration
4	2	2	4s

Source provenance

Generated by Codex as a temporary validation artifact for SpeechAntiSpoofingBenchmarks.

Evaluation

See eval.yaml and submissions/README.md.

Citation

Original paper: https://arxiv.org/abs/1911.01601

@article{wang2020asvspoof,
  title={ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech},
  author={Wang, Xin and Yamagishi, Junichi and Todisco, Massimiliano and others},
  journal={Computer Speech & Language},
  volume={64},
  pages={101114},
  year={2020}
}

Maintainer

Codex validation artifact; safe to delete after validation.