Text Generation Quality Evaluation on WaterBench 1000 prompts
9.878PPLGreen-List
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Green-Listbeta=0.05, gamma*=0.1, delta*=8, t_start=1, t*_end=202025.11 | 9.878 | 7.399 | 7.484 | 6.214 | 7.782 | 7.22 | |
| Green-Listbeta=0.01, gamma*=0.1, delta*=8, t_start=1, t*_end=202025.11 | 9.878 | 7.399 | 7.484 | 6.214 | 7.782 | 7.22 | |
| No WatermarkCondition=Baseline2025.11 | 10.391 | 8.508 | 8.954 | 7.364 | 8.474 | 8.325 | |
| Green-Listbeta=0.1, gamma*=0.25, delta*=8, t_start=1, t*_end=202025.11 | 10.485 | 8.302 | 8.574 | 6.978 | 8.29 | 8.036 | |
| Gumbel-maxtau*=1.192025.11 | 10.652 | 7.922 | 7.849 | 6.712 | 7.904 | 7.597 | |
| Green-Listbeta=0.15, gamma*=0.25, delta*=4, t_start=1, t*_end=802025.11 | 12.195 | 8.457 | 8.782 | 7.044 | 8.32 | 8.151 |