Hallucination Detection on DiaHalu sampled (test)
245Support (Total Samples)GPT-4
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-4evaluation_category=Overall accuracy2026.05 | 245 | — | — | — | 54.3 | |
| GPT-4evaluation_category=macro avg2026.05 | 245 | 59.5 | 54.4 | 47.3 | — | |
| GPT-4evaluation_category=weighted avg2026.05 | 245 | 59.5 | 54.3 | 47.3 | — | |
| GPT-4evaluation_category=Class 1 (hallucination)2026.05 | 123 | 66.7 | 17.9 | 28.2 | — | |
| GPT-4evaluation_category=Class 0 (factual)2026.05 | 122 | 52.4 | 91 | 66.5 | — |