Watermark Radioactivity Detection on LibriSpeech
100Detection Accuracy (Accdet)LambdaMark
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LambdaMarkMadv (Generation Model)=SemanticVocoder, FT (Fine-tuning Strategy)=Full2026.06 | 100 | 96.47 | 100 | 100 | 100 | 1 | |
| LambdaMarkMadv (Generation Model)=SemanticVocoder, FT (Fine-tuning Strategy)=LoRA2026.06 | 100 | 96.31 | 100 | 100 | 100 | 1 | |
| LambdaMarkMadv (Generation Model)=YourTTS, FT (Fine-tuning Strategy)=Full2026.06 | 99.5 | 86.56 | 100 | 99 | 99.5 | 0.995 | |
| LambdaMarkMadv (Generation Model)=AudioLDM2, FT (Fine-tuning Strategy)=Full2026.06 | 94.5 | 82.38 | 100 | 89 | 94.17 | 0.999 | |
| LambdaMarkMadv (Generation Model)=AudioLDM2, FT (Fine-tuning Strategy)=LoRA2026.06 | 84.5 | 73.66 | 100 | 69 | 81.66 | 0.985 |