Hallucination Detection on FOIL (Reference-Conditioned Accuracy)
98.4Accuracy (4 Refs)Ref-FLEUR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Ref-FLEURReference-based=true2025.12 | 98.4 | 97.3 | |
| Ref-G-VEvalReference-based=true2025.12 | 98.4 | 97.8 | |
| PearlReference-based=true, Backbone=ViT-B/322025.12 | 97.2 | 96.5 | |
| Ref-HICEScoreReference-based=true2025.12 | 97 | 96.4 | |
| FLEURReference-based=false2025.12 | 96.8 | 96.8 | |
| PearlReference-based=false, Backbone=ViT-B/322025.12 | 96.7 | 96.7 | |
| DENEBReference-based=true, Backbone=ViT-B/322025.12 | 96.1 | 95.1 | |
| PolosReference-based=true2025.12 | 95.4 | 93.3 | |
| RefPAC-Sreference-based=true2023.03 | 94.9 | 93.7 | |
| RefPAC-SReference-based=true, Backbone=ViT-B/322025.12 | 94.9 | 93.7 | |
| BLIP2ScoreReference-based=false2025.12 | 94.3 | 94.3 | |
| CIDErreference-based=true2023.03 | 94.1 | 85.7 | |
| RefPAC-S++Reference-based=true, Backbone=ViT-B/322025.12 | 94.1 | 93.5 | |
| CLAIRReference-based=true2025.12 | 93.6 | — | |
| MIDreference-based=true2023.03 | 93.5 | 90.5 | |
| HICEScoreReference-based=false2025.12 | 93.1 | 93.1 | |
| RefCLIP-Sreference-based=true2023.03 | 92.6 | 91 | |
| RefCLIP-SReference-based=true, Backbone=ViT-B/322025.12 | 92.6 | 91 | |
| CIDErReference-based=true2025.12 | 90.6 | 82.5 | |
| PAC-S++Reference-based=false, Backbone=ViT-B/322025.12 | 90.2 | 90.2 | |
| PAC-Sreference-based=false2023.03 | 89.9 | 89.9 | |
| PAC-SReference-based=false, Backbone=ViT-B/322025.12 | 89.9 | 89.9 | |
| CLIP-Sreference-based=false2023.03 | 87.2 | 87.2 | |
| CLIP-SReference-based=false2025.12 | 87.2 | 87.2 | |
| BLEU-4reference-based=true2023.03 | 87 | 66.2 | |
| SPICEReference-based=true2025.12 | 86.1 | 75.5 | |
| BLEU-1reference-based=true2023.03 | 85.4 | 65.7 | |
| BLEUReference-based=true2025.12 | 82.6 | 66.5 | |
| METEORReference-based=true2025.12 | 82.6 | 78.8 | |
| METEORreference-based=true2023.03 | 82 | 70.1 | |
| ROUGEReference-based=true2025.12 | 79.3 | 71.7 | |
| ROUGEreference-based=true2023.03 | 70.4 | 54.6 |