Add W2V2-AASIST submission

wav2vec2 XLS-R 300M + AASIST. EER from local run, scores pinned to model repo @75ec0a4. reproduce --scoring --no-local verified locally.

submissions/w2v2-aasist.yaml

@@ -0,0 +1,43 @@
+schema_version: 4
+system:
+  name: W2V2-AASIST
+  slug: w2v2-aasist
+  description: 'wav2vec 2.0 (XLS-R 300M) self-supervised front-end fine-tuned end-to-end
+    with an AASIST spectro-temporal graph-attention back-end for speech anti-spoofing.
+    The XLS-R features are projected to 128-d, max-pooled, and fed through a RawNet2-style
+    residual encoder and heterogeneous stacking graph-attention layers. Official TakHemlata/SSL_Anti-spoofing
+    LA checkpoint (LA_model.pth), trained on ASVspoof2019 LA with RawBoost augmentation,
+    FP32, deterministic first-64600-sample window (no random crop).
+
+    '
+  code: https://github.com/TakHemlata/SSL_Anti-spoofing
+  checkpoint: https://huggingface.co/SpeechAntiSpoofingBenchmarks/W2V2-AASIST
+  params_millions: 317.8378
+  paper:
+    arxiv_id: '2202.12233'
+    url: https://arxiv.org/abs/2202.12233
+    bibtex: "@inproceedings{tak2022automatic,\n  title={Automatic speaker verification\
+      \ spoofing and deepfake detection using wav2vec 2.0 and data augmentation},\n\
+      \  author={Tak, Hemlata and Todisco, Massimiliano and Wang, Xin and Jung, Jee-weon\
+      \ and Yamagishi, Junichi and Evans, Nicholas},\n  booktitle={The Speaker and\
+      \ Language Recognition Workshop (Odyssey 2022)},\n  pages={112--119},\n  year={2022}\n\
+      }\n"
+dataset:
+  id: SpeechAntiSpoofingBenchmarks/ASVspoof2021_LA
+  revision: dc119733697c946fcd17fe7c1541d7f26b4bbe07
+  split: test
+scores:
+  eer_percent: 8.112941686538043
+  n_trials: 181566
+  n_skipped: 0
+artifact:
+  scores_url: https://huggingface.co/SpeechAntiSpoofingBenchmarks/W2V2-AASIST/resolve/75ec0a4aa6491cee7748c12996a87561d1b02fb7/.eval_results/SpeechAntiSpoofingBenchmarks/ASVspoof2021_LA/scores.txt
+  scores_sha256: 29a22a3d568225ff6fa31d2637de0c68a72bd8cac2105f9049a7f3b60b470096
+  bench_version: speech-spoof-bench==0.3.3
+reproduction: {}
+submitter:
+  hf_username: korallll
+  contact: k.n.borodin@mtuci.ru
+submitted_at: '2026-06-04'
+notes: XLS-R 300M (wav2vec 2.0) front-end + AASIST back-end, LA_model.pth. Deterministic
+  first-64600-sample window (no random crop).