Hallucination Classification on DiaHalu sampled
245SupportGPT-5.2
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT-5.2Classification category=accuracy2026.05 | 245 | — | — | — | 69.8 | |
| GPT-5.2Classification category=macro avg2026.05 | 245 | 74.6 | 69.7 | 68.2 | — | |
| GPT-5.2Classification category=weighted avg2026.05 | 245 | 74.5 | 69.8 | 68.2 | — | |
| GPT-5.2Classification category=Class 1 (hallucination)2026.05 | 123 | 63.8 | 91.9 | 75.3 | — | |
| GPT-5.2Classification category=Class 0 (factual)2026.05 | 122 | 85.3 | 47.5 | 61.1 | — |