Hallucination Detection on PUGG
84.3F1 ScoreKG-Guard
Evaluation Results
| Method | Links | |
|---|---|---|
| KG-GuardGraph encoder=GraphTransformer2026.05 | 84.3 | |
| Most Frequent2026.05 | 83.5 | |
| KG-GuardGraph encoder=GAT2026.05 | 82.7 | |
| GPT 5 MiniInput modalities=G2026.05 | 82.6 | |
| GPT 5 MiniInput modalities=G+RS+RT2026.05 | 81.2 | |
| SelfCheckNumber of samples=102026.05 | 73.8 | |
| Llama 4 ScoutInput modalities=G2026.05 | 68.8 | |
| Llama 4 ScoutInput modalities=G+RS2026.05 | 61.3 | |
| Llama 4 ScoutInput modalities=G+RS+RT2026.05 | 60.8 | |
| Random2026.05 | 58.8 | |
| SelfCheckNumber of samples=52026.05 | 48.3 |