Spoofing Attack Robustness on C4 RealNewsLike
0.9284AUCDualGuard
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DualGuardBase LLM=OPT-1.3B2025.12 | 0.9284 | 0.3505 | 0.8247 | 0.9482 | |
| DualGuardBase LLM=Llama3.1-8B-Instruct2025.12 | 0.9159 | 0.2552 | 0.6562 | 0.9201 | |
| SynthIDBase LLM=OPT-1.3B2025.12 | 0.5786 | 0.0622 | 0.2021 | 0.6447 | |
| EWDBase LLM=OPT-1.3B2025.12 | 0.578 | 0.099 | 0.1927 | 0.7769 | |
| SWEETBase LLM=OPT-1.3B2025.12 | 0.573 | 0.1031 | 0.2113 | 0.773 | |
| SynthIDBase LLM=Llama3.1-8B-Instruct2025.12 | 0.5686 | 0.0838 | 0.1571 | 0.5912 | |
| KGWBase LLM=Llama3.1-8B-Instruct2025.12 | 0.5573 | 0.1451 | 0.2124 | 0.7153 | |
| DIPmarkBase LLM=OPT-1.3B2025.12 | 0.5569 | 0.0729 | 0.1354 | 0.5365 | |
| SWEETBase LLM=Llama3.1-8B-Instruct2025.12 | 0.5564 | 0.0208 | 0.1354 | 0.7374 | |
| KGWBase LLM=OPT-1.3B2025.12 | 0.5141 | 0.0667 | 0.0923 | 0.7506 | |
| DIPmarkBase LLM=Llama3.1-8B-Instruct2025.12 | 0.5123 | 0.0263 | 0.0632 | 0.523 | |
| XSIRBase LLM=Llama3.1-8B-Instruct2025.12 | 0.5069 | 0.0695 | 0.0909 | 0.6518 | |
| EWDBase LLM=Llama3.1-8B-Instruct2025.12 | 0.5031 | 0.1077 | 0.159 | 0.722 | |
| UnbiasedBase LLM=Llama3.1-8B-Instruct2025.12 | 0.5021 | 0.05 | 0.1 | 0.5012 | |
| UnbiasedBase LLM=OPT-1.3B2025.12 | 0.4956 | 0.0492 | 0.0985 | 0.4983 | |
| SIRBase LLM=Llama3.1-8B-Instruct2025.12 | 0.4466 | 0.0052 | 0.0309 | 0.687 | |
| AARBase LLM=Llama3.1-8B-Instruct2025.12 | 0.4332 | 0.041 | 0.0769 | 0.5724 | |
| XSIRBase LLM=OPT-1.3B2025.12 | 0.43 | 0.0306 | 0.0561 | 0.6762 | |
| SIRBase LLM=OPT-1.3B2025.12 | 0.419 | 0.0308 | 0.0667 | 0.6713 | |
| AARBase LLM=OPT-1.3B2025.12 | 0.3785 | 0.016 | 0.0588 | 0.5649 |