Hallucination Correction on TruthfulQA gated mode (test)
77.9IGR (%)DoLa
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DoLaModel=Llama-3.2-1B, Evaluation Mode=Gated mode2026.05 | 77.9 | 44 | 54 | 26 | 4 | |
| ITIModel=Llama-3.2-1B, Evaluation Mode=Gated mode2026.05 | 77.9 | 19 | 26 | 47 | 9 | |
| AdaSteerModel=Llama-3.2-1B, Evaluation Mode=Gated mode2026.05 | 77.9 | 52 | 23 | 41 | 6 | |
| SADIModel=Llama-3.2-1B, Evaluation Mode=Gated mode2026.05 | 77.9 | 51 | 23 | 41 | 7 | |
| ICDModel=Llama-3.2-1B, Evaluation Mode=Gated mode2026.05 | 77.9 | 51 | 27 | 44 | 7 | |
| PC-LDCDModel=Llama-3.2-1B, Evaluation Mode=Gated mode2026.05 | 77.9 | 50 | 36 | 51 | 14 | |
| DoLaModel=Mistral-7B, Evaluation Mode=Gated mode2026.05 | 58.3 | 69 | 71 | 29 | 7 | |
| ITIModel=Mistral-7B, Evaluation Mode=Gated mode2026.05 | 58.3 | 44 | 31 | 47 | 10 | |
| AdaSteerModel=Mistral-7B, Evaluation Mode=Gated mode2026.05 | 58.3 | 69 | 34 | 50 | 13 | |
| SADIModel=Mistral-7B, Evaluation Mode=Gated mode2026.05 | 58.3 | 69 | 34 | 50 | 13 | |
| ICDModel=Mistral-7B, Evaluation Mode=Gated mode2026.05 | 58.3 | 64 | 35 | 51 | 15 | |
| PC-LDCDModel=Mistral-7B, Evaluation Mode=Gated mode2026.05 | 58.3 | 72 | 54 | 72 | 29 | |
| DoLaModel=Qwen3-4B, Evaluation Mode=Gated mode2026.05 | 45.4 | 72 | 76 | 25 | 6 | |
| ITIModel=Qwen3-4B, Evaluation Mode=Gated mode2026.05 | 45.4 | 60 | 35 | 52 | 11 | |
| AdaSteerModel=Qwen3-4B, Evaluation Mode=Gated mode2026.05 | 45.4 | 75 | 33 | 50 | 11 | |
| SADIModel=Qwen3-4B, Evaluation Mode=Gated mode2026.05 | 45.4 | 76 | 32 | 49 | 10 | |
| ICDModel=Qwen3-4B, Evaluation Mode=Gated mode2026.05 | 45.4 | 73 | 36 | 53 | 11 | |
| PC-LDCDModel=Qwen3-4B, Evaluation Mode=Gated mode2026.05 | 45.4 | 78 | 51 | 68 | 23 | |
| DoLaModel=Llama-3.1-8B, Evaluation Mode=Gated mode2026.05 | 41.4 | 66 | 55 | 29 | 3 | |
| ITIModel=Llama-3.1-8B, Evaluation Mode=Gated mode2026.05 | 41.4 | 46 | 32 | 50 | 10 | |
| AdaSteerModel=Llama-3.1-8B, Evaluation Mode=Gated mode2026.05 | 41.4 | 67 | 29 | 44 | 8 | |
| SADIModel=Llama-3.1-8B, Evaluation Mode=Gated mode2026.05 | 41.4 | 66 | 29 | 44 | 8 | |
| ICDModel=Llama-3.1-8B, Evaluation Mode=Gated mode2026.05 | 41.4 | 69 | 37 | 53 | 15 | |
| PC-LDCDModel=Llama-3.1-8B, Evaluation Mode=Gated mode2026.05 | 41.4 | 69 | 51 | 65 | 25 | |
| VanillaModel=Llama-3.2-1B, Evaluation Mode=Gated mode2026.05 | 0 | 50 | 23 | 41 | 6 | |
| VanillaModel=Qwen3-4B, Evaluation Mode=Gated mode2026.05 | 0 | 50 | 33 | 50 | 11 | |
| VanillaModel=Mistral-7B, Evaluation Mode=Gated mode2026.05 | 0 | 50 | 34 | 50 | 13 | |
| VanillaModel=Llama-3.1-8B, Evaluation Mode=Gated mode2026.05 | 0 | 50 | 29 | 44 | 8 |