Hallucination Detection on RAGTruth Summarization (Llama-2-13b)
72.9AUCROCPersImg
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| PersImgEvaluation Setting=Black-box, Generator=Llama-2-13b2026.01 | 72.9 | 61.63 | 18.39 | 68.63 | |
| Betti CurveEvaluation Setting=Black-box, Generator=Llama-2-13b2026.01 | 67.93 | 60.05 | 14.8 | 67.99 | |
| Attention ScoreEvaluation Setting=Black-box, Generator=Llama-2-13b2026.01 | 60.05 | 59.66 | 14.48 | 55.97 | |
| PersEntropyEvaluation Setting=Black-box, Generator=Llama-2-13b2026.01 | 52.78 | 53.95 | 2.69 | 61.22 |