Answer-level hallucination detection on RAGTruth ++
100PrecisionVectara
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Vectaracontext-only setting=true2026.03 | 100 | 26.9 | 42.4 | |
| MetaQAcontext-only setting=true2026.03 | 94.1 | 21.6 | 35.2 | |
| Lettuce2026.03 | 92.1 | 23.3 | 37.1 | |
| RT4CHART2026.03 | 84.5 | 71.8 | 77.6 | |
| SelfCheckGPTcontext-only setting=true2026.03 | 84.3 | 14.3 | 24.4 |