Detection of LLM-Generated Text on XSum GPT-J-6B
88.35AUROCSurpMark ref-attack
Evaluation results
| Method | Links | |
|---|---|---|
| SurpMark ref-attackAdversarial Attack Type=Original2025.10 | 88.35 | |
| SurpMark test-attackAdversarial Attack Type=Original2025.10 | 88.35 | |
| SurpMark both-attackAdversarial Attack Type=Original2025.10 | 88.35 | |
| SurpMark test-attackAdversarial Attack Type=Attack 12025.10 | 86.37 | |
| Lastde++Adversarial Attack Type=Original2025.10 | 85.38 | |
| SurpMark both-attackAdversarial Attack Type=Attack 22025.10 | 85.23 | |
| SurpMark test-attackAdversarial Attack Type=Attack 22025.10 | 84.86 | |
| SurpMark both-attackAdversarial Attack Type=Attack 12025.10 | 84.44 | |
| SurpMark ref-attackAdversarial Attack Type=Attack 22025.10 | 83.85 | |
| SurpMark ref-attackAdversarial Attack Type=Attack 12025.10 | 83.84 | |
| Fast-DetectGPTAdversarial Attack Type=Original2025.10 | 78.6 | |
| Fast-DetectGPTAdversarial Attack Type=Attack 12025.10 | 75.44 | |
| Fast-DetectGPTAdversarial Attack Type=Attack 22025.10 | 74.09 | |
| Lastde++Adversarial Attack Type=Attack 22025.10 | 73.55 | |
| Lastde++Adversarial Attack Type=Attack 12025.10 | 69.79 |