Upload README.md with huggingface_hub

README.md

@@ -31,6 +31,29 @@ Code, feature extraction, and the inference CLI: **https://github.com/vlaght/dsl
 Per-video feature expansion: delta (×2) + temporal statistics (std for DSLP, mean+std for
 TAFreq/TANoise). Output: P(fake) ∈ [0,1], decision at 0.5.
 
+## Results
+
+10-fold stratified cross-validation.
+
+**Ensemble**
+
+| Dataset | AUC | F1 | Accuracy | Cohen's κ |
+|---|---|---|---|---|
+| FakeAVCeleb (21,544 videos)  | 0.9593 | 0.9599 | 0.9248 | 0.3352 |
+| DeepSpeak v2 (16,465 videos) | 0.9984 | 0.9783 | 0.9810 | 0.9615 |
+
+**Per-component AUC**
+
+| Model | FakeAVCeleb | DeepSpeak v2 |
+|---|---|---|
+| DSLP    | 0.8606 | 0.9579 |
+| TAFreq  | 0.9674 | 0.9954 |
+| TANoise | 0.8100 | 0.9356 |
+| **Ensemble** | **0.9593** | **0.9984** |
+
+On FakeAVCeleb the threshold-free AUC stays high while Cohen's κ drops, because many
+FakeAVCeleb forgeries leave mouth motion largely intact.
+
 ## Files
 
 ```