Detection of LLM generated text on MAGE QA
65.33ROC AUC (FPR=1%)Binoculars
Evaluation Results
| Method | Links | |
|---|---|---|
| BinocularsGeneration Strategy=Topic-based 3.5-turbo, Regeneration Strategy=Paraphrase 4o-mini2026.01 | 65.33 | |
| SearchLLMBase Detector=Binoculars, Generation Strategy=Topic-based 3.5-turbo, Regeneration Strategy=Paraphrase 4o-mini2026.01 | 65.33 | |
| SearchLLMBase Detector=Longformer, Generation Strategy=Topic-based 3.5-turbo, Regeneration Strategy=Paraphrase 4o-mini2026.01 | 38.22 | |
| DetectGPTGeneration Strategy=Topic-based 3.5-turbo, Regeneration Strategy=Paraphrase 4o-mini2026.01 | 20.88 | |
| SearchLLMBase Detector=DetectGPT, Generation Strategy=Topic-based 3.5-turbo, Regeneration Strategy=Paraphrase 4o-mini2026.01 | 20.88 | |
| LongformerGeneration Strategy=Topic-based 3.5-turbo, Regeneration Strategy=Paraphrase 4o-mini2026.01 | 4.66 | |
| RADARGeneration Strategy=Topic-based 3.5-turbo, Regeneration Strategy=Paraphrase 4o-mini2026.01 | 0 | |
| SearchLLMBase Detector=RADAR, Generation Strategy=Topic-based 3.5-turbo, Regeneration Strategy=Paraphrase 4o-mini2026.01 | 0 |