Hallucination Detection on SciQ (AUC, F95, AUP)
88.99AUCSIVR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SIVRModel=Qwen-3-4B-Instruct2026.04 | 88.99 | 52.56 | 93.04 | |
| SIVRModel=Qwen-3-4B-Instruct2026.04 | 88.99 | 52.56 | 93.04 | |
| Lookback LensModel=Qwen-3-14B2026.04 | 85.38 | 53.19 | 88.06 | |
| SIVRModel=Qwen-3-14B2026.04 | 84.7 | 49.44 | 87.75 | |
| SATMD + MSPModel=Qwen-3-4B-Instruct2026.04 | 84.4 | 73.13 | 91.09 | |
| SATMD + MSPModel=Qwen-3-4B-Instruct2026.04 | 84.4 | 73.13 | 91.09 | |
| SAPLMAModel=Qwen-3-4B-Instruct2026.04 | 83.8 | 60.26 | 89.45 | |
| SAPLMAModel=Qwen-3-4B-Instruct2026.04 | 83.8 | 60.26 | 89.45 | |
| SAPLMAModel=Qwen-3-14B2026.04 | 81.46 | 59.77 | 84.6 | |
| Lookback LensModel=Qwen-3-4B-Instruct2026.04 | 81.26 | 86.9 | 88.05 | |
| Lookback LensModel=Qwen-3-4B-Instruct2026.04 | 81.26 | 86.9 | 88.05 | |
| SATMD + MSPModel=Qwen-3-14B2026.04 | 77.52 | 78.38 | 87.2 | |
| SARModel=Qwen-3-4B-Instruct2026.04 | 73.65 | 86.11 | 79.62 | |
| SARModel=Qwen-3-4B-Instruct2026.04 | 73.65 | 86.11 | 79.62 | |
| Temp-ScaledModel=Qwen-3-4B-Instruct2026.04 | 72.65 | 76.53 | 75.95 | |
| Temp-ScaledModel=Qwen-3-4B-Instruct2026.04 | 72.65 | 76.53 | 75.95 | |
| PerplexityModel=Qwen-3-4B-Instruct2026.04 | 71.88 | 77.3 | 76.11 | |
| PerplexityModel=Qwen-3-4B-Instruct2026.04 | 71.88 | 77.3 | 76.11 | |
| SEModel=Qwen-3-4B-Instruct2026.04 | 71.74 | 83.75 | 80.02 | |
| SEModel=Qwen-3-4B-Instruct2026.04 | 71.74 | 83.75 | 80.02 | |
| Max ProbModel=Qwen-3-4B-Instruct2026.04 | 71.29 | 77.3 | 75.51 | |
| Max ProbModel=Qwen-3-4B-Instruct2026.04 | 71.29 | 77.3 | 75.51 | |
| EntropyModel=Qwen-3-4B-Instruct2026.04 | 70.37 | 78.57 | 75.91 | |
| EntropyModel=Qwen-3-4B-Instruct2026.04 | 70.37 | 78.57 | 75.91 | |
| Temp-ScaledModel=Qwen-3-14B2026.04 | 61.28 | 86.04 | 62.94 | |
| PerplexityModel=Qwen-3-14B2026.04 | 61.12 | 86.04 | 63.83 | |
| EntropyModel=Qwen-3-14B2026.04 | 60.98 | 86.96 | 65.64 | |
| SEModel=Qwen-3-14B2026.04 | 60.79 | 84.27 | 64.37 | |
| Max ProbModel=Qwen-3-14B2026.04 | 60.58 | 86.04 | 63.17 | |
| SARModel=Qwen-3-14B2026.04 | 58.3 | 94.38 | 61.75 | |
| CoE-CModel=Qwen-3-14B2026.04 | 57.11 | 94.28 | 60.38 | |
| CoE-RModel=Qwen-3-14B2026.04 | 51.69 | 95.65 | 55.43 | |
| EnergyModel=Qwen-3-14B2026.04 | 50.06 | 100 | 61.74 | |
| P(True)Model=Qwen-3-14B2026.04 | 49.7 | 98.4 | 60.6 | |
| P(True)Model=Qwen-3-4B-Instruct2026.04 | 49.61 | 99.74 | 66.59 | |
| P(True)Model=Qwen-3-4B-Instruct2026.04 | 49.61 | 99.74 | 66.59 | |
| CoE-CModel=Qwen-3-4B-Instruct2026.04 | 46.14 | 95.41 | 55.53 | |
| CoE-CModel=Qwen-3-4B-Instruct2026.04 | 46.14 | 95.41 | 55.53 | |
| EnergyModel=Qwen-3-4B-Instruct2026.04 | 41.48 | 100 | 59.92 | |
| EnergyModel=Qwen-3-4B-Instruct2026.04 | 41.48 | 100 | 59.92 | |
| CoE-RModel=Qwen-3-4B-Instruct2026.04 | 27.22 | 100 | 47.25 | |
| CoE-RModel=Qwen-3-4B-Instruct2026.04 | 27.22 | 100 | 47.25 |