Hallucination Detection on HUB (test)
76.8AlgorithmicHALT-L
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HALT-LBase Model=Llama 3.1 8B2026.02 | 76.8 | 60.17 | 47.67 | 56.67 | 72.89 | 72.71 | 74.07 | 66.93 | 65.4 | 76.92 | 67.01 | 63.03 | |
| HALT-QBase Model=Llama 3.1 8B2026.02 | 32.68 | 58.6 | 39.71 | 41.32 | 73 | 62.9 | 68.78 | 70.75 | 49.78 | 58.45 | 62.74 | 55.6 | |
| ∆HdecBase Model=Llama 3.1 8B2026.02 | 26.44 | 37.41 | 62.03 | 35.17 | 42.18 | 66.95 | 49.42 | 45.03 | 24.92 | 44.51 | 48.03 | 43.41 | |
| HaltsBase Model=Llama 3.1 8B2026.02 | 24.48 | 36.47 | 66.67 | 34.27 | 39.15 | 61.63 | 43.4 | 32.99 | 24.59 | 44.51 | 42.9 | 40.81 | |
| PPLBase Model=Llama 3.1 8B2026.02 | 24.24 | 35.09 | 43.07 | 33.25 | 39.15 | 44.14 | 28.31 | 24.49 | 24.59 | 44.51 | 33.95 | 34.08 | |
| HoverallBase Model=Llama 3.1 8B2026.02 | 24.24 | 34.55 | 38.14 | 31.97 | 39.15 | 42 | 25.73 | 24.49 | 24.59 | 44.51 | 32.86 | 32.94 | |
| LettuceBase Model=Llama 3.1 8B2026.02 | 24.24 | 41.5 | 38.14 | 41.06 | 83.38 | 41.8 | 77.3 | 59.71 | 33.36 | 44.51 | 64 | 48.5 |