Question Answering on MedMCQA (AUC, F95, AUP)
75.95AUCSIVR (Ours)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SIVR (Ours)Backbone=Llama-3.1-8B-Instruct2026.04 | 75.95 | 67.24 | 59.57 | |
| SIVRBackbone=Llama-3.1-8B-Instruct2026.04 | 75.95 | 67.24 | 59.57 | |
| SATMD + MSPBackbone=Llama-3.1-8B-Instruct2026.04 | 69.94 | 72.41 | 56.37 | |
| SARBackbone=Llama-3.1-8B-Instruct2026.04 | 68.95 | 82.69 | 63.57 | |
| SEBackbone=Llama-3.1-8B-Instruct2026.04 | 66.95 | 88.46 | 66.56 | |
| SEBackbone=Llama-3.1-8B-Instruct2026.04 | 66.95 | 88.46 | 66.56 | |
| SEmodel=Llama-3.2-3B-Instruct2026.04 | 66.88 | 80.81 | 67.01 | |
| SEBackbone=Llama-3.2-3B-Instruct2026.04 | 66.88 | 80.81 | 67.01 | |
| Lookback LensBackbone=Llama-3.1-8B-Instruct2026.04 | 66.7 | 63.03 | 52.1 | |
| SIVRmodel=Llama-3.2-3B-Instruct2026.04 | 66.59 | 72.82 | 63.68 | |
| SIVRBackbone=Llama-3.2-3B-Instruct2026.04 | 66.59 | 72.82 | 63.68 | |
| Lookback LensModel=Ministral-8B-Instruct2026.04 | 65.78 | 91.43 | 61.66 | |
| SAPLMAmodel=Llama-3.2-3B-Instruct2026.04 | 65.63 | 86.79 | 61.89 | |
| TADBackbone=Llama-3.1-8B-Instruct2026.04 | 65.09 | 92.44 | 57.04 | |
| SARmodel=Llama-3.2-3B-Instruct2026.04 | 64.71 | 76.47 | 72.08 | |
| EnergyBackbone=Llama-3.1-8B-Instruct2026.04 | 64.33 | 85.91 | 54.2 | |
| TADmodel=Llama-3.2-3B-Instruct2026.04 | 63.56 | 79.82 | 52.61 | |
| SAPLMABackbone=Llama-3.1-8B-Instruct2026.04 | 63.53 | 81.03 | 51.66 | |
| SATMD + MSPmodel=Llama-3.2-3B-Instruct2026.04 | 63.51 | 90.99 | 54.82 | |
| Entropymodel=Llama-3.2-3B-Instruct2026.04 | 62.76 | 84.22 | 58.62 | |
| SATMD + MSPModel=Ministral-8B-Instruct2026.04 | 62.64 | 83.33 | 61.25 | |
| EntropyBackbone=Llama-3.1-8B-Instruct2026.04 | 62.39 | 93.81 | 52.44 | |
| Energymodel=Llama-3.2-3B-Instruct2026.04 | 62.2 | 82.32 | 57.79 | |
| CoE-Cmodel=Llama-3.2-3B-Instruct2026.04 | 62.14 | 83.84 | 57.16 | |
| Perplexitymodel=Llama-3.2-3B-Instruct2026.04 | 61.57 | 87.26 | 57.77 | |
| PerplexityBackbone=Llama-3.1-8B-Instruct2026.04 | 61.14 | 94.85 | 50.97 | |
| SIVRModel=Ministral-8B-Instruct2026.04 | 61.12 | 82.08 | 60.76 | |
| SIVRBackbone=Ministral-8B-Instruct2026.04 | 61.12 | 82.08 | 60.76 | |
| Max Probmodel=Llama-3.2-3B-Instruct2026.04 | 60.77 | 86.88 | 56.76 | |
| Max ProbBackbone=Llama-3.1-8B-Instruct2026.04 | 60.72 | 95.88 | 50.93 | |
| Temp-Scaledmodel=Llama-3.2-3B-Instruct2026.04 | 60.26 | 89.35 | 56.46 | |
| Temp-ScaledBackbone=Llama-3.1-8B-Instruct2026.04 | 60.12 | 95.19 | 50.12 | |
| TADModel=Ministral-8B-Instruct2026.04 | 58.73 | 77.05 | 60.93 | |
| Lookback Lensmodel=Llama-3.2-3B-Instruct2026.04 | 58.25 | 83.18 | 49.86 | |
| P(True)Backbone=Llama-3.1-8B-Instruct2026.04 | 57.49 | 88.5 | 48.15 | |
| CoE-CBackbone=Llama-3.1-8B-Instruct2026.04 | 56.47 | 94.16 | 46.91 | |
| P(True)model=Llama-3.2-3B-Instruct2026.04 | 56.37 | 92.59 | 51.15 | |
| SEModel=Ministral-8B-Instruct2026.04 | 55.09 | 83.33 | 61.41 | |
| SEBackbone=Ministral-8B-Instruct2026.04 | 55.09 | 83.33 | 61.41 | |
| SARModel=Ministral-8B-Instruct2026.04 | 54.68 | 78.57 | 58.47 | |
| EntropyModel=Ministral-8B-Instruct2026.04 | 54.57 | 94.19 | 55.48 | |
| PerplexityModel=Ministral-8B-Instruct2026.04 | 53.84 | 94.61 | 54.51 | |
| P(True)Model=Ministral-8B-Instruct2026.04 | 53.21 | 95.23 | 52.92 | |
| EnergyModel=Ministral-8B-Instruct2026.04 | 52.95 | 92.32 | 54.18 | |
| Max ProbModel=Ministral-8B-Instruct2026.04 | 52.87 | 94.4 | 53.95 | |
| Temp-ScaledModel=Ministral-8B-Instruct2026.04 | 52.74 | 94.61 | 53.5 | |
| CoE-CModel=Ministral-8B-Instruct2026.04 | 51.01 | 95.23 | 51.54 | |
| CoE-RBackbone=Llama-3.1-8B-Instruct2026.04 | 50.9 | 94.85 | 41.52 | |
| SAPLMAModel=Ministral-8B-Instruct2026.04 | 49.24 | 90.72 | 53.67 | |
| CoE-Rmodel=Llama-3.2-3B-Instruct2026.04 | 47.47 | 95.63 | 45.27 | |
| CoE-RModel=Ministral-8B-Instruct2026.04 | 44.05 | 97.93 | 47.73 |