LLM-generated text detection on RAID (train)
99.69ChatGPT ScoreReMoDetect
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ReMoDetectBackbone=LLaMA-3.2-1B, Sample size=512, Metric=AUROC2026.04 | 99.69 | 97.82 | 94.57 | 57.36 | 99.33 | 59.18 | 97.7 | 43.06 | 89.63 | 85.29 | 93.02 | |
| BinocularsBackbone=LLaMA-3.2-1B, Sample size=512, Metric=AUROC2026.04 | 99.66 | 96.38 | 99.12 | 79.3 | 99.05 | 72.45 | 98.22 | 51.99 | 83.41 | 97.06 | 98 | |
| IRMBackbone=LLaMA-3.2-1B, Sample size=512, Metric=AUROC2026.04 | 99.58 | 99.27 | 87.43 | 86.31 | 99.99 | 62.25 | 99.35 | 57.57 | 92.42 | 87.14 | 92.1 | |
| Log-RankBackbone=LLaMA-3.2-1B, Sample size=512, Metric=AUROC2026.04 | 98.71 | 92.95 | 96.94 | 71.13 | 99.12 | 71.34 | 98.03 | 51.83 | 81.27 | 87.43 | 91.89 | |
| Log-LikelihoodBackbone=LLaMA-3.2-1B, Sample size=512, Metric=AUROC2026.04 | 98.48 | 93.51 | 97.35 | 68.77 | 98.93 | 69.82 | 97.78 | 48.84 | 80.08 | 89.26 | 92.42 |