Hallucination Detection on CSL-Daily
0.963AUCMETA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| METAModel Type=Gloss-free (GF)2025.10 | 0.963 | 0.999 | 88.3 | |
| Token entropyModel Type=Gloss-free (GF)2025.10 | 0.958 | 0.998 | 86.6 | |
| ConfidenceModel Type=Gloss-free (GF)2025.10 | 0.953 | 0.998 | 86.8 | |
| PerplexityModel Type=Gloss-free (GF)2025.10 | 0.952 | 0.998 | 84.2 | |
| GroundingModel Type=Gloss-free (GF)2025.10 | 0.951 | 0.998 | 97 | |
| METAModel Type=Gloss-based (GB)2025.10 | 0.865 | 0.994 | 75.3 | |
| GroundingModel Type=Gloss-based (GB)2025.10 | 0.803 | 0.991 | 96.3 | |
| Token entropyModel Type=Gloss-based (GB)2025.10 | 0.729 | 0.985 | 66.5 | |
| ConfidenceModel Type=Gloss-based (GB)2025.10 | 0.647 | 0.976 | 50.8 | |
| PerplexityModel Type=Gloss-based (GB)2025.10 | 0.609 | 0.98 | 61 |