Text Generation Quality and Watermark Detectability on ELI5
100AUCKGW-2
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| KGW-2Model=Llama2-7B, Temperature (T)=12025.12 | 100 | 100 | 9.61 | 3 | 91 | |
| Gumbel-maxModel=Llama2-7B, Temperature (T)=12025.12 | 99.8 | 99.4 | 7.98 | 18 | 81 | |
| SynthIDModel=Llama2-7B, Temperature (T)=12025.12 | 99.7 | 98.8 | 8.06 | 9 | 89 | |
| KGW-2Model=Qwen3-4B, Temperature (T)=0.72025.12 | 99.6 | 98.8 | 3.69 | 3 | 87 | |
| MarkTuneModel=Qwen3-4B, Temperature (T)=0.72025.12 | 99.6 | 94.2 | 2.78 | 3 | 93 | |
| SynthIDModel=Qwen3-4B, Temperature (T)=0.72025.12 | 99.3 | 94.6 | 2.76 | 6 | 91 | |
| KGW-1Model=Llama2-7B, Temperature (T)=12025.12 | 99.3 | 94.6 | 8.47 | 4 | 94 | |
| MarkTuneModel=Llama2-7B, Temperature (T)=12025.12 | 99.3 | 93 | 8.19 | 1 | 96 | |
| Gumbel-maxModel=Qwen3-4B, Temperature (T)=0.72025.12 | 99.1 | 93.8 | 2.71 | 13 | 77 | |
| GaussMarkModel=Qwen3-4B, Temperature (T)=0.72025.12 | 97.9 | 73.4 | 2.83 | 3 | 93 | |
| KGW-1Model=Qwen3-4B, Temperature (T)=0.72025.12 | 97.1 | 71 | 3.04 | 4 | 92 | |
| KGW-DModel=Llama2-7B, Temperature (T)=12025.12 | 96.9 | 75.4 | 8.53 | 5 | 96 | |
| GaussMarkModel=Llama2-7B, Temperature (T)=12025.12 | 95.8 | 62.2 | 8.13 | 2 | 94 | |
| KGW-DModel=Qwen3-4B, Temperature (T)=0.72025.12 | 94.1 | 54.2 | 3.09 | 4 | 92 | |
| No WatermarkModel=Qwen3-4B, Temperature (T)=0.72025.12 | 50 | 1 | 2.67 | 3 | 94 | |
| No WatermarkModel=Llama2-7B, Temperature (T)=12025.12 | 50 | 1 | 7.86 | 2 | 97 |