Watermark Detectability on C4 RealNewsLike (Del-0.2) (test)
99.3AUCGumbel-max
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gumbel-maxBackbone=Qwen3-4B, Temperature=0.72025.12 | 99.3 | 94.8 | |
| Gumbel-maxBackbone=Llama2-7B, Temperature=1.02025.12 | 99.3 | 92.4 | |
| Gumbel-maxBackbone=Qwen3-4B, Temperature=0.72025.12 | 99.3 | 92.2 | |
| Gumbel-maxBackbone=Llama2-7B, Temperature=1.02025.12 | 99.3 | 95 | |
| SynthIDBackbone=Qwen3-4B, Temperature=0.72025.12 | 98.7 | 89 | |
| SynthIDBackbone=Llama2-7B, Temperature=1.02025.12 | 98.7 | 86.2 | |
| KGW-2Backbone=Qwen3-4B, Temperature=0.72025.12 | 98.6 | 82.6 | |
| SynthIDBackbone=Qwen3-4B, Temperature=0.72025.12 | 98.6 | 84.6 | |
| KGW-2Backbone=Llama2-7B, Temperature=1.02025.12 | 98 | 75.6 | |
| SynthIDBackbone=Llama2-7B, Temperature=1.02025.12 | 97.9 | 77.4 | |
| KGW-2Backbone=Qwen3-4B, Temperature=0.72025.12 | 97.5 | 74.6 | |
| KGW-2Backbone=Llama2-7B, Temperature=1.02025.12 | 97.5 | 71.2 | |
| MarkTuneBackbone=Qwen3-4B, Temperature=0.72025.12 | 94.8 | 55.2 | |
| KGW-DBackbone=Qwen3-4B, Temperature=0.72025.12 | 94.7 | 58.2 | |
| MarkTuneBackbone=Llama2-7B, Temperature=1.02025.12 | 94.4 | 57.8 | |
| KGW-DBackbone=Qwen3-4B, Temperature=0.72025.12 | 93.9 | 57.8 | |
| MarkTuneBackbone=Llama2-7B, Temperature=1.02025.12 | 92.8 | 50.6 | |
| MarkTuneBackbone=Qwen3-4B, Temperature=0.72025.12 | 92.1 | 48.4 | |
| GaussMarkBackbone=Qwen3-4B, Temperature=0.72025.12 | 91.9 | 38.4 | |
| KGW-DBackbone=Llama2-7B, Temperature=1.02025.12 | 91.4 | 41 | |
| GaussMarkBackbone=Qwen3-4B, Temperature=0.72025.12 | 90.8 | 36.6 | |
| KGW-DBackbone=Llama2-7B, Temperature=1.02025.12 | 88.7 | 40.6 | |
| KGW-1Backbone=Qwen3-4B, Temperature=0.72025.12 | 88.1 | 33 | |
| KGW-1Backbone=Llama2-7B, Temperature=1.02025.12 | 87.8 | 30 | |
| KGW-1Backbone=Qwen3-4B, Temperature=0.72025.12 | 87.3 | 32.4 | |
| KGW-1Backbone=Llama2-7B, Temperature=1.02025.12 | 85.3 | 23.2 | |
| GaussMarkBackbone=Llama2-7B, Temperature=1.02025.12 | 85.3 | 22.8 | |
| GaussMarkBackbone=Llama2-7B, Temperature=1.02025.12 | 83.5 | 24.2 |