ModelListedExternal

XLS R 2B AntiDeepfake NDA

Bit-exact copy of nii-yamagishilab/xls-r-2b-anti-deepfake-nda (the AntiDeepfake collection, nii-yamagishilab), hosted for the Speech Anti-Spoofing Arena. Trained without data augmentation (NDA).

Sourcehuggingface.co/SpeechAntiSpoofingBenchmarks/XLS-R-2B-AntiDeepfake-NDA/tree/1360a3e4e1f0e803b2edb6763be1a67615bd681d

KindOtherBase modelnii-yamagishilab/xls-r-2b-anti-deepfake-ndaVersionRevision 1360a3eBgrade

Model source

Kind: Other
File size: 5.64 MB–8.06 GB
Base model: nii-yamagishilab/xls-r-2b-anti-deepfake-nda
Version: vRevision 1360a3e
Updated: Oct 25, 2026
Downloads: 34
Source: Hugging Face

Source description

XLS-R-2B-AntiDeepfake-NDA

Bit-exact copy of nii-yamagishilab/xls-r-2b-anti-deepfake-nda (the AntiDeepfake collection, nii-yamagishilab), hosted for the Speech Anti-Spoofing Arena. Trained without data augmentation (NDA).

Paper: Post-training for Deepfake Speech Detection, arXiv:2506.21090
Original code: https://github.com/nii-yamagishilab/AntiDeepfake
Backbone: XLS-R 2B; head = AdaptiveAvgPool1d + Linear(D, 2). Score = real logit.
License: CC BY-NC-SA 4.0 (non-commercial) — inherited from the source.

⚠️ Evaluation restriction

It is STRICTLY PROHIBITED to evaluate this model on its training / in-domain corpora: ASVspoof2019 LA, ASVspoof2021 LA/DF, ASVspoof5, CFAD, DECRO, DFADD, "Diffuse or Confuse", DiffSSD, DSD, HABLA, MLAAD, SpoofCeleb, CVoiceFake, any LibriTTS-based set (incl. SONAR, CD-ADD, LibriSeVoc), any VoxCeleb-based set, WaveFake, any FLEURS-based set, CodecFake (any variant), any AIShell3-based set, and any MLS-based set. On the Arena it is evaluated only on ArAD and In-the-Wild.

All credit to the original authors (Ge, Wang, et al., NII).

Arena results

EER% 0.25 on XMAD EER% 0.29 on PyAra EER% 1.47 on ODSS EER% 0.00 on J-SPAW_LA EER% 5.25 on DeepVoice EER% 2.99 on InTheWild EER% 1.86 on ArAD 1-SRR% 0.32 on LRLspoof EER% 1.83 on EmoFake_test 1-SRR% 4.6 on EmoSpoofTTS arena tier arena rank

Maintainer

Maintained by Kirill Borodin (SpeechAntiSpoofingBenchmarks).

Email: [email protected]
Telegram: @korallll_ai

Sources

1 source

Hugging FacePrimaryVersion 1360a3e4e1f0e803b2edb6763be1a67615bd681d

huggingface.co/SpeechAntiSpoofingBenchmarks/XLS-R-2B-AntiDeepfake-NDA/tree/1360a3e4e1f0e803b2edb6763be1a67615bd681d

Verified Oct 25

Download files

Model artifacts

2 artifacts

model.safetensors

safetensors · 8.06 GB · SHA-256 ce3d3fe3d1c9…6f86 · Hugging Face

Download file

xls-r-2b-antideepfake-nda.onnx

onnx · 5.64 MB · SHA-256 1ae5c633d15f…18f2 · Hugging Face

Download file

Source excerpts

3 excerpts

ReadmeAug 7huggingface.co/SpeechAntiSpoofingBenchmarks/XLS-R-2B-AntiDeepfake-NDA

--- license: cc-by-nc-sa-4.0 base_model: - nii-yamagishilab/xls-r-2b-anti-deepfake-nda tags: - anti-spoofing - audio-deepfake-detection - speech pipeline_tag: audio-classification --- # XLS-R-2B-AntiDeepfake-NDA Bit-exact copy of [`nii-yamagishilab/xls-r-2b-anti-deepfake-nda`](https://huggingface.co/nii-yamagishilab/xls-r-2b-anti-deepfake-nda) (the AntiDeepfake collection, nii-yamagishilab), hosted for the Speech Anti-Spoofing **Arena**. Trained **without** data augmentation (NDA). - **Paper:** Post-training for Deepfake Speech Detection, [arXiv:2506.21090](https://arxiv.org/abs/2506.21090) - **Original code:** https://github.com/nii-yamagishilab/AntiDeepfake - **Backbone:** XLS-R 2B; head = AdaptiveAvgPool1d + Linear(D, 2). Score = real logit. - **License:** CC BY-NC-SA 4.0 (non-commercial) — inherited from the source. ## ⚠️ Evaluation restriction It is **STRICTLY PROHIBITED** to evaluate this model on its training / in-domain corpora: ASVspoof2019 LA, ASVspoof2021 LA/DF, ASVspoof5, CFAD, DECRO, DFADD, "Diffuse or Confuse", DiffSSD, DSD, HABLA, MLAAD, SpoofCeleb, CVoiceFake, any LibriTTS-based set (incl. SONAR, CD-ADD, LibriSeVoc), any VoxCeleb-based set, WaveFake, any FLEURS-based set, CodecFake (any variant), any AIShell3-based set, and any MLS-based set. On the Arena it is evaluated **only** on ArAD and In-the-Wild. All credit to the original authors (Ge, Wang, et al., NII). ## Arena results  [![EER% 0.25 on XMAD](https://img.shields.io/badge/EER%25%20on%20XMAD-0.25%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=xls-r-2b-antideepfake-nda) [![EER% 0.29 on PyAra](https://img.shields.io/badge/EER%25%20on%20PyAra-0.29%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=xls-r-2b-antideepfake-nda) [![EER% 1.47 on ODSS](https://img.shields.io/badge/EER%25%20on%20ODSS-1.47%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=xls-r-2b-antideepfake-nda) [![EER% 0.00 on J-SPAW_LA](https://img.shields.io/badge/EER%25%20on%20J--SPAW_LA-0.00%25-lightgrey)](https://huggingface.co/spaces/SpeechAntiSpoofingBenchmarks/SpeechAntiSpoofingArena?system=xls-r-2b-antideepfake-nda) [![EER% 5.25 on DeepVoice](https://img.shields.io/badge/EER%25%20on%20DeepVoice-5.25%25-lightgrey)](https:/...

NoteOct 25huggingface.co/SpeechAntiSpoofingBenchmarks/XLS-R-2B-AntiDeepfake-NDA

Source context: 34 downloads · 0 likes · Pipeline audio-classification · Repo SpeechAntiSpoofingBenchmarks/XLS-R-2B-AntiDeepfake-NDA

NoteAug 7huggingface.co/SpeechAntiSpoofingBenchmarks/XLS-R-2B-AntiDeepfake-NDA

Source context: 16 downloads · 0 likes · Pipeline audio-classification · Repo SpeechAntiSpoofingBenchmarks/XLS-R-2B-AntiDeepfake-NDA