Hallucination detection on FinHalu
94.1PrecisionGPT-4o (teacher)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4o (teacher)role=teacher, retrieval-equalized=true2026.04 | 94.1 | 95.9 | 95 | |
| FINGROUND (8B distilled)model_size=8B, type=distilled, retrieval-equalized=true2026.04 | 92.7 | 90.2 | 91.4 | |
| Self-RAGretrieval-equalized=true2026.04 | 81.2 | 77.1 | 79.1 | |
| CRAGretrieval-equalized=true2026.04 | 80.6 | 74.9 | 77.6 | |
| HHEMretrieval-equalized=true2026.04 | 78.9 | 73.8 | 76.3 | |
| FActScoreretrieval-equalized=true2026.04 | 74.2 | 79.3 | 76.7 | |
| SelfCheckGPTretrieval-equalized=true2026.04 | 69.4 | 76.5 | 72.8 |