Answer-level hallucination detection on RAGTruth Enhance
100PrecisionSelfCheckGPT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SelfCheckGPTcontext-only setting=true2026.03 | 100 | 16.7 | 28.6 | |
| Lettuce2026.03 | 90.7 | 54.5 | 68.1 | |
| Vectaracontext-only setting=true2026.03 | 89.4 | 60.8 | 72.4 | |
| MetaQAcontext-only setting=true2026.03 | 85.3 | 10.2 | 18.2 | |
| RT4CHART2026.03 | 78.1 | 92 | 84.5 |