LLM-generated content detection on HumanEval (AUROC)
0.8173AUROCRAIDAR
Evaluation results
| Method | Links | |
|---|---|---|
| RAIDAR2026.05 | 0.8173 | |
| LiSCPparaphrase model=GPT-3.5-Turbo, encoder=SBERT2026.05 | 0.8108 | |
| Fast-DetectGPT2026.05 | 0.6679 | |
| R-Detect2026.05 | 0.649 | |
| Ghostbustersubstitute LLM=GPT-3.5-Turbo2026.05 | 0.5378 | |
| LogRank2026.05 | 0.535 | |
| Rank2026.05 | 0.5348 | |
| DetectGPTperturbation model=T5-3B2026.05 | 0.5267 | |
| RoBERTa-largeBackbone=large2026.05 | 0.4692 | |
| Entropy2026.05 | 0.4306 | |
| RoBERTa-baseBackbone=base2026.05 | 0.4217 |