HIR-SDD-raw

HIR-SDD-raw is the unprocessed annotation release accompanying HIR-SDD.

It publishes the original human annotations before filtering, translation, and merging used to build the clean release. Audio is not bundled here; use HIR-SDD for bundled clips or obtain audio from the upstream corpora cited below.

• Paper: Towards Robust Speech Deepfake Detection via Human-Inspired Reasoning
• Dataset: marsianin500/HIR-SDD-raw
• Clean release: marsianin500/HIR-SDD

Repository layout

HIR-SDD-raw/
├── README.md
└── annotations/
    ├── modern-generators.parquet
    └── speechllm.parquet

No audio files are included in this repository.

Configurations

modern-generators

Crowd annotations collected in the modern-generators labeling pipeline, exported from the original Russian annotation pool before filtering and English translation.

Item	Value
Rows	140,483
Unique audio_id	45,118
Excluded	all rows whose audio comes from data_modern-generators (1,254 rows)

Fields:

Field	Description
audio_id	Clip identifier
source_corpus	Upstream audio corpus inferred from the source path
reasons	Structured spoofing cue tags
comment	Annotator free-text comment (Russian)
marker_id	Annotator identifier
consistency	Annotator consistency score
is_bonafide_labelled	Annotator label
is_bonafide_gt	Clip ground-truth label
language	Clip language metadata
complexity	Clip complexity score
annotator_quality	Annotator quality tier

Multiple rows per audio_id are preserved.

speechllm

Annotations derived from the SpeechLLM / SpeechEval deepfake-detection subset.

Item	Value
Rows	3,952

Fields:

Field	Description
audio_id	Clip identifier
source_corpus	SpeechLLM
is_bonafide	Bonafide/spoof label
reasoning	English natural-language explanation
reasons	Structured spoofing cue tags
text_suggest	Suggested speech improvements
text_improve	Detailed quality assessment text

Relationship to HIR-SDD

Aspect	HIR-SDD-raw	HIR-SDD
Annotations	original, unmerged	filtered, merged, English reasoning only
Audio	not bundled	bundled under license folders

The clean release is built from filtered modern-generators annotations plus SpeechLLM annotations. HIR-SDD-raw is provided for reproducibility and for researchers who want the pre-processing annotation pool.

Usage

from datasets import load_dataset

modern_generators = load_dataset("marsianin500/HIR-SDD-raw", "modern-generators", split="train")
speechllm = load_dataset("marsianin500/HIR-SDD-raw", "speechllm", split="train")

To obtain audio for a row, either:

1. load the matching clip from HIR-SDD by audio_id, or
2. download the clip from the upstream corpus listed under source_corpus.

License

The annotations in HIR-SDD-raw are released under CC BY-NC-SA 4.0.

This repository does not redistribute audio. If you retrieve audio elsewhere, you must comply with the license of the corresponding upstream corpus.

Upstream audio sources

The annotation rows refer to clips from third-party corpora. If you obtain and use those clips, cite the corresponding source.

ASVspoof 5

@article{wang2025asvspoof5,
  title   = {{ASVspoof} 5: Design, Collection and Validation of Resources for Spoofing, Deepfake, and Adversarial Attack Detection Using Crowdsourced Speech},
  author  = {Wang, Xin and Delgado, H{\'e}ctor and Tak, Hemlata and Jung, Jee-weon and Shim, Hye-jin and Todisco, Massimiliano and Kukanov, Ivan and Liu, Xuechen and Sahidullah, Md and Kinnunen, Tomi and Evans, Nicholas and Lee, Kong Aik and Yamagishi, Junichi and others},
  journal = {Computer Speech \& Language},
  year    = {2025},
  doi     = {10.1016/j.csl.2025.101825}
}

MLAAD

@inproceedings{muller2024mlaad,
  title     = {{MLAAD}: The Multi-Language Audio Anti-Spoofing Dataset},
  author    = {M{\"u}ller, Nicolas M. and Kawa, Piotr and Choong, Wei Herng and Casanova, Edresson and G{\"o}lge, Eren and M{\"u}ller, Thorsten and Syga, Piotr and Sperl, Philip and B{\"o}ttinger, Konstantin},
  booktitle = {International Joint Conference on Neural Networks (IJCNN)},
  year      = {2024}
}

LibriSeVoc

@inproceedings{sun2023librisevoc,
  title     = {AI-Synthesized Voice Detection Using Neural Vocoder Artifacts},
  author    = {Sun, Chengzhe and Jia, Shan and Hou, Shuwei and Lyu, Siwei},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2023}
}

Golos

@inproceedings{karpov2021golos,
  title     = {{Golos}: Russian Dataset for Speech Research},
  author    = {Karpov, Nikolay and Denisenko, Alexander and Minkin, Fedor},
  booktitle = {Interspeech 2021},
  pages     = {1419--1423},
  year      = {2021},
  doi       = {10.21437/Interspeech.2021-462}
}

M-AILABS Speech Dataset

@misc{mailabs2017,
  author       = {Solak, I. Celeste Aurora and Naumov, Dima},
  title        = {The {M-AILABS} Speech Dataset},
  year         = {2017},
  howpublished = {\url{https://github.com/i-celeste-aurora/m-ailabs-dataset}}
}

DFADD

@inproceedings{du2024dfadd,
  title     = {{DFADD}: The Diffusion and Flow-Matching Based Audio Deepfake Dataset},
  author    = {Du, Jiawei and Lin, I-Ming and Chiu, I-Hsiang and Chen, Xuanjun and Wu, Haibin and Ren, Wenze and Tsao, Yu and Lee, Hung-yi and Jang, Jyh-Shing Roger},
  booktitle = {2024 IEEE Spoken Language Technology Workshop (SLT)},
  pages     = {921--928},
  year      = {2024},
  organization = {IEEE}
}

SOVA

@misc{sova2021dataset,
  author       = {Zubarev, Egor and Moskalets, Timofey and {SOVA.ai}},
  title        = {{SOVA} Dataset: Free Public {STT}/{ASR} Dataset},
  year         = {2021},
  howpublished = {\url{https://github.com/sovaai/sova-dataset}}
}

Russian LibriSpeech (ruLS)

@techreport{bakhturina2021toolbox,
  title       = {A Toolbox for Construction and Analysis of Speech Datasets},
  author      = {Bakhturina, Evelina and Lavrukhin, Vitaly and Ginsburg, Boris},
  year        = {2021},
  institution = {NVIDIA}
}

ESpeech podcasts / spoof audio

@dataset{espeech_podcasts2025,
  title        = {{ESpeech}: Large-Scale Russian Podcast Speech Dataset},
  author       = {{ESpeech Team}},
  year         = {2025},
  howpublished = {\url{https://huggingface.co/datasets/ESpeech/ESpeech-podcasts}}
}

XTTS-generated audio

@inproceedings{casanova24_interspeech,
  title     = {{XTTS}: a Massively Multilingual Zero-Shot Text-to-Speech Model},
  author    = {Casanova, Edresson and Davis, Kelly and G{\"o}lge, Eren and G{\"o}knar, G{\"o}rkem and Gulea, Iulian and Hart, Logan and Aljafari, Aya and Meyer, Joshua and Morais, Reuben and Olayemi, Samuel and Weber, Julian},
  booktitle = {Interspeech 2024},
  pages     = {4978--4982},
  year      = {2024},
  doi       = {10.21437/Interspeech.2024-2016}
}

SpeechLLM / SpeechEval

@article{wang2025speechllm,
  title   = {{SpeechLLM}-as-Judges: Towards General and Interpretable Speech Quality Evaluation},
  author  = {Wang, Hui and Zhao, Jinghua and Yang, Yifan and Liu, Shujie and Chen, Junyang and Zhang, Yanzhe and Zhao, Shiwan and Li, Jinyu and Zhou, Jiaming and Sun, Haoqin and others},
  journal = {arXiv preprint arXiv:2510.14664},
  year    = {2025}
}

PyAra

PyAra does not have a dedicated dataset paper. If you use PyAra audio, cite the dataset page:

• https://www.kaggle.com/datasets/alep079/pyara

Related work:

@article{efanov2023pyara,
  title   = {Comparison of the effectiveness of cepstral coefficients for Russian speech synthesis detection},
  author  = {Efanov, Dmitry and Aleksandrov, Pavel and Mironov, Ilia},
  journal = {Journal of Computer Virology and Hacking Techniques},
  year    = {2023},
  doi     = {10.1007/s11416-023-00491-0}
}

Citing HIR-SDD-raw

If you use these annotations or the HIR-SDD framework, please cite:

@misc{dvirniak2026robustspeechdeepfakedetection,
  title         = {Towards Robust Speech Deepfake Detection via Human-Inspired Reasoning},
  author        = {Dvirniak, Artem and Kushnir, Evgeny and Tarasov, Dmitrii and Iudin, Artem and Kiriukhin, Oleg and Pautov, Mikhail and Korzh, Dmitrii and Rogov, Oleg Y.},
  year          = {2026},
  eprint        = {2603.10725},
  archivePrefix = {arXiv},
  primaryClass  = {cs.SD},
  url           = {https://arxiv.org/abs/2603.10725}
}

You may also reference this Hugging Face release:

@misc{hir_sdd_raw_hf_2026,
  title        = {{HIR-SDD-raw}: Raw Human Annotations for Spoofing and Deepfake Detection Reasoning},
  author       = {Dvirniak, Artem and Kushnir, Evgeny and Tarasov, Dmitrii and Iudin, Artem and Kiriukhin, Oleg and Pautov, Mikhail and Korzh, Dmitrii and Rogov, Oleg Y.},
  year         = {2026},
  howpublished = {Hugging Face dataset},
  url          = {https://huggingface.co/datasets/marsianin500/HIR-SDD-raw}
}

Please also cite HIR-SDD and any upstream audio corpora you use.