Watermarking on Natural Questions (NQ) (test)
100AUROCKGW
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| KGWBackbone=OPT-1.3B, Attack=No Attack2026.05 | 100 | 100 | 100 | |
| SynthIDBackbone=OPT-1.3B, Attack=No Attack2026.05 | 100 | 100 | 100 | |
| k-SemStampBackbone=OPT-1.3B, Attack=No Attack2026.05 | 100 | 96.3 | 98.4 | |
| SimMarkBackbone=OPT-1.3B, Attack=No Attack2026.05 | 100 | 99.2 | 99.4 | |
| AliMarkBackbone=OPT-1.3B, Attack=No Attack2026.05 | 100 | 100 | 100 | |
| AliMarkBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 99.9 | 93.8 | 97.2 | |
| AliMarkBackbone=OPT-1.3B, Attack=Parrot2026.05 | 99.9 | 93.7 | 96.9 | |
| SimMarkBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 99.8 | 93.6 | 95.8 | |
| SemStampBackbone=OPT-1.3B, Attack=No Attack2026.05 | 99.7 | 94.1 | 97 | |
| SIRBackbone=OPT-1.3B, Attack=No Attack2026.05 | 99.5 | 0 | 91 | |
| SimMarkBackbone=OPT-1.3B, Attack=Parrot2026.05 | 98.7 | 90.6 | 93.2 | |
| PMarkBackbone=OPT-1.3B, Attack=No Attack2026.05 | 98.6 | 92.2 | 94.6 | |
| EXPBackbone=OPT-1.3B, Attack=No Attack2026.05 | 98 | 96.4 | 97 | |
| SemStampBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 97.6 | 58 | 79.4 | |
| PMarkBackbone=OPT-1.3B, Attack=Parrot2026.05 | 97.1 | 47 | 81.8 | |
| SemStampBackbone=OPT-1.3B, Attack=Parrot2026.05 | 97 | 41.4 | 70.8 | |
| k-SemStampBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 97 | 45.3 | 54.5 | |
| SIRBackbone=OPT-1.3B, Attack=Parrot2026.05 | 96.6 | 0 | 33.6 | |
| k-SemStampBackbone=OPT-1.3B, Attack=Parrot2026.05 | 96 | 37.3 | 47.1 | |
| PMarkBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 95.8 | 48 | 79 | |
| SIRBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 95.7 | 0 | 33.8 | |
| AliMarkBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 95.1 | 54.7 | 59.8 | |
| SimMarkBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 93.6 | 40.6 | 53.6 | |
| AliMarkBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 91.7 | 49.5 | 55.2 | |
| SIRBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 90.7 | 0 | 4 | |
| SimMarkBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 90.1 | 42.2 | 50.3 | |
| SynthIDBackbone=OPT-1.3B, Attack=Parrot2026.05 | 86.9 | 8.2 | 18.2 | |
| SynthIDBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 86.4 | 6.2 | 14 | |
| SIRBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 84.5 | 0 | 5.4 | |
| SemStampBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 84.5 | 10.3 | 24.7 | |
| SemStampBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 79.5 | 5.2 | 17.6 | |
| k-SemStampBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 79.5 | 2.6 | 5.8 | |
| PMarkBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 77.1 | 5.6 | 17.6 | |
| k-SemStampBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 76.4 | 1.8 | 3.8 | |
| PMarkBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 75.9 | 5.4 | 16.2 | |
| KGWBackbone=OPT-1.3B, Attack=Parrot2026.05 | 61.6 | 0.6 | 1.8 | |
| KGWBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 61.2 | 0.6 | 2 | |
| EXPBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 59.2 | 2.4 | 4.6 | |
| KGWBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 59.1 | 2.2 | 3.2 | |
| SynthIDBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 57.9 | 0 | 0 | |
| KGWBackbone=OPT-1.3B, Attack=DIPPER2026.05 | 57.6 | 0.4 | 1.4 | |
| EXPBackbone=OPT-1.3B, Attack=Parrot2026.05 | 57.2 | 1.2 | 3 | |
| EXPBackbone=OPT-1.3B, Attack=Pegasus2026.05 | 55.6 | 0.8 | 2.6 | |
| EXPBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 55.6 | 0.2 | 0.8 | |
| SynthIDBackbone=OPT-1.3B, Attack=GPT-3.52026.05 | 55 | 0 | 0 |