LLM-generated text detection on RAID Wikipedia-related samples
99.35GPT-4 Performance ScoreSearchLLM
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SearchLLMBase Detector=Binoculars2026.01 | 99.35 | 99.97 | 98.93 | 99.97 | 99.85 | |
| Binoculars2026.01 | 99.12 | 99.97 | 98.47 | 99.97 | 99.85 | |
| SearchLLMBase Detector=Longformer2026.01 | 98.02 | 99.43 | 91.95 | 97.68 | 97.31 | |
| Longformer2026.01 | 97.05 | 98.79 | 88.27 | 96.09 | 95.92 | |
| SearchLLMBase Detector=RADAR2026.01 | 97.02 | 98.66 | 92.49 | 98.32 | 97.96 | |
| RADAR2026.01 | 96.05 | 97.94 | 89.58 | 97.59 | 97.12 | |
| SearchLLMBase Detector=DetectGPT2026.01 | 88.67 | 91.43 | 84.18 | 87.81 | 88.56 | |
| DetectGPT2026.01 | 82.98 | 86.83 | 73.83 | 80.31 | 83.05 |