Question Answering on SciQ (AUC, F95, AUP)
87.79AUCSIVR (Ours)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SIVR (Ours)Backbone=Llama-3.1-8B-Instruct2026.04 | 87.79 | 54.55 | 94.14 | |
| SIVRBackbone=Llama-3.1-8B-Instruct2026.04 | 87.79 | 54.55 | 94.14 | |
| SAPLMABackbone=Llama-3.1-8B-Instruct2026.04 | 87.41 | 57.35 | 92.91 | |
| SIVRmodel=Llama-3.2-3B-Instruct2026.04 | 85.7 | 52.54 | 93.68 | |
| SIVRBackbone=Llama-3.2-3B-Instruct2026.04 | 85.7 | 52.54 | 93.68 | |
| TADmodel=Llama-3.2-3B-Instruct2026.04 | 85.16 | 77.59 | 94.07 | |
| TADBackbone=Llama-3.2-3B-Instruct2026.04 | 85.16 | 77.59 | 94.07 | |
| SATMD + MSPBackbone=Llama-3.1-8B-Instruct2026.04 | 84.83 | 51.67 | 92.69 | |
| Lookback LensBackbone=Llama-3.1-8B-Instruct2026.04 | 83.78 | 80 | 90.63 | |
| SAPLMAmodel=Llama-3.2-3B-Instruct2026.04 | 83.62 | 60.34 | 93 | |
| TADBackbone=Llama-3.1-8B-Instruct2026.04 | 83.6 | 65.67 | 91.69 | |
| TADBackbone=Llama-3.1-8B-Instruct2026.04 | 83.6 | 65.67 | 91.69 | |
| SIVRModel=Ministral-8B-Instruct2026.04 | 80.81 | 63.95 | 86.03 | |
| SIVRBackbone=Ministral-8B-Instruct2026.04 | 80.81 | 63.95 | 86.03 | |
| Lookback Lensmodel=Llama-3.2-3B-Instruct2026.04 | 80.54 | 77.59 | 92.09 | |
| SEmodel=Llama-3.2-3B-Instruct2026.04 | 79.44 | 77.08 | 92.03 | |
| SEBackbone=Llama-3.1-8B-Instruct2026.04 | 79.01 | 74.63 | 87.85 | |
| SEModel=Ministral-8B-Instruct2026.04 | 78.98 | 81.08 | 87.83 | |
| SATMD + MSPmodel=Llama-3.2-3B-Instruct2026.04 | 78.75 | 80.7 | 91.07 | |
| PerplexityBackbone=Llama-3.1-8B-Instruct2026.04 | 78.59 | 62.35 | 85.17 | |
| SARModel=Ministral-8B-Instruct2026.04 | 78.51 | 67.57 | 83.35 | |
| Temp-ScaledBackbone=Llama-3.1-8B-Instruct2026.04 | 77.87 | 61.76 | 83.55 | |
| Max ProbBackbone=Llama-3.1-8B-Instruct2026.04 | 77.63 | 62.35 | 83.97 | |
| EntropyBackbone=Llama-3.1-8B-Instruct2026.04 | 77.49 | 66.47 | 85.13 | |
| SATMD + MSPModel=Ministral-8B-Instruct2026.04 | 77.46 | 62.2 | 83.34 | |
| TADModel=Ministral-8B-Instruct2026.04 | 75.54 | 84.44 | 80.9 | |
| TADBackbone=Ministral-8B-Instruct2026.04 | 75.54 | 84.44 | 80.9 | |
| CoE-Cmodel=Llama-3.2-3B-Instruct2026.04 | 75.06 | 83.85 | 89.14 | |
| SARmodel=Llama-3.2-3B-Instruct2026.04 | 73.98 | 72.92 | 89.17 | |
| PerplexityModel=Ministral-8B-Instruct2026.04 | 73.78 | 72.83 | 76.2 | |
| SAPLMAModel=Ministral-8B-Instruct2026.04 | 73.55 | 82.95 | 79.42 | |
| EntropyModel=Ministral-8B-Instruct2026.04 | 72.98 | 77.85 | 75.79 | |
| Lookback LensModel=Ministral-8B-Instruct2026.04 | 72.93 | 88.16 | 83.28 | |
| Entropymodel=Llama-3.2-3B-Instruct2026.04 | 72.85 | 80.41 | 85.7 | |
| SARBackbone=Llama-3.1-8B-Instruct2026.04 | 72.78 | 82.09 | 82.69 | |
| Energymodel=Llama-3.2-3B-Instruct2026.04 | 72.73 | 79.73 | 86.21 | |
| Max ProbModel=Ministral-8B-Instruct2026.04 | 72.56 | 74.2 | 74.93 | |
| Perplexitymodel=Llama-3.2-3B-Instruct2026.04 | 72.42 | 77.66 | 84.63 | |
| Temp-ScaledModel=Ministral-8B-Instruct2026.04 | 72.1 | 73.06 | 74.59 | |
| EnergyModel=Ministral-8B-Instruct2026.04 | 71.65 | 81.51 | 75.28 | |
| Max Probmodel=Llama-3.2-3B-Instruct2026.04 | 71.54 | 78.01 | 84.05 | |
| Temp-Scaledmodel=Llama-3.2-3B-Instruct2026.04 | 70.71 | 78.35 | 82.71 | |
| CoE-CBackbone=Llama-3.1-8B-Instruct2026.04 | 69.56 | 90.88 | 82.59 | |
| P(True)Backbone=Llama-3.1-8B-Instruct2026.04 | 68.76 | 86.18 | 80.67 | |
| CoE-Rmodel=Llama-3.2-3B-Instruct2026.04 | 65.41 | 94.16 | 84.73 | |
| EnergyBackbone=Llama-3.1-8B-Instruct2026.04 | 61.56 | 91.47 | 75.54 | |
| P(True)model=Llama-3.2-3B-Instruct2026.04 | 58.66 | 93.13 | 79.04 | |
| CoE-CModel=Ministral-8B-Instruct2026.04 | 58.55 | 93.61 | 68.29 | |
| P(True)Model=Ministral-8B-Instruct2026.04 | 49.92 | 94.52 | 56.88 | |
| CoE-RModel=Ministral-8B-Instruct2026.04 | 41.16 | 100 | 53.06 | |
| CoE-RBackbone=Llama-3.1-8B-Instruct2026.04 | 32.31 | 100 | 57.84 |