Hallucination Detection on CWQ
87.4F1 ScoreKG-Guard
Evaluation Results
| Method | Links | |
|---|---|---|
| KG-GuardGraph encoder=GraphTransformer2026.05 | 87.4 | |
| KG-GuardGraph encoder=GAT2026.05 | 86.3 | |
| GPT 5 MiniInput modalities=G+RS+RT2026.05 | 84.2 | |
| Most Frequent2026.05 | 83.5 | |
| GPT 5 MiniInput modalities=G2026.05 | 82.6 | |
| SelfCheckNumber of samples=102026.05 | 72.2 | |
| SelfCheckNumber of samples=52026.05 | 67.1 | |
| Llama 4 ScoutInput modalities=G+RS2026.05 | 61.3 | |
| Llama 4 ScoutInput modalities=G+RS+RT2026.05 | 61.2 | |
| Llama 4 ScoutInput modalities=G2026.05 | 60.2 | |
| Random2026.05 | 58.3 |