Forensics-XLSR-Wild: Speech Anti-Spoofing Arena wrapper

Packaged for the Speech Anti-Spoofing Arena from the authors' code.

No paper: the model sits in the Arena's unpublished tier (shown, not ranked).

Forensics 0.3B XLS-R "wild" speech deepfake detector (Eliya Segev, 2026): a fully fine-tuned wav2vec 2.0 XLS-R 300M front-end whose 24 transformer hidden states are pooled by an AASIST-style spectro-temporal graph-attention pool (layer x time graphs, GAT dims 64/32), followed by a 3-block residual MLP classifier (160-d) and a single logit (label 1 = real). Authors' checkpoint_epoch_4 as released. Input as the authors' inference.py: peak-normalised waveform, 5 s (80000 samples) centre crop, shorter clips repeat-padded. score = the logit (sigmoid = P(real)). Unpublished model (HF model card only, no paper).

• Checkpoint (commit-pinned): checkpoint_epoch_4.safetensors
• Params: 315.926375 M. Trained on the author's mix (SpeechFake, MD-CommonVoice, DFADD, CodecFake, ASVspoof 2021 DF, ASVspoof 2019 LA, Common Voice, AISHELL, proprietary; splits unstated). The Arena marks ASVspoof2019_LA, ASVspoof2021_DF, DFADD, InTheWild and DECRO as n/a* (training overlap).
• Score: higher = bona fide. J-SPAW_LA was scored with the PyTorch wrapper; the other datasets with the TensorRT drop-in class (FP32 on the RTX 4070 Ti SUPER; FP32 on the RTX 5060 Ti), which matched PyTorch on the full J-SPAW_LA set.

Arena results

EER% (lower is better); EmoSpoofTTS and LRLspoof are spoof-only and use 1-SRR% at the model's DeepVoice EER threshold. Per-dataset scores are under .eval_results/.