Question Answering on MATH
83.48AUCSIVR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SIVRmodel=Llama-3.2-3B-Instruct2026.04 | 83.48 | 42.11 | 86 | |
| SIVRBackbone=Llama-3.2-3B-Instruct2026.04 | 83.48 | 42.11 | 86 | |
| SAPLMAModel=Ministral-8B-Instruct2026.04 | 81.64 | 70.1 | 82.49 | |
| SAPLMABackbone=Llama-3.1-8B-Instruct2026.04 | 79.87 | 75.76 | 69.4 | |
| SIVR (Ours)Backbone=Llama-3.1-8B-Instruct2026.04 | 79.05 | 67.57 | 72.3 | |
| SIVRBackbone=Llama-3.1-8B-Instruct2026.04 | 79.05 | 67.57 | 72.3 | |
| TADModel=Ministral-8B-Instruct2026.04 | 78.51 | 68.13 | 71.86 | |
| EnergyBackbone=Llama-3.1-8B-Instruct2026.04 | 78.1 | 55.69 | 73.22 | |
| SAPLMAmodel=Llama-3.2-3B-Instruct2026.04 | 78.06 | 87.1 | 81.69 | |
| SATMD + MSPModel=Ministral-8B-Instruct2026.04 | 77.71 | 82.5 | 78.41 | |
| SIVRModel=Ministral-8B-Instruct2026.04 | 77.56 | 72.16 | 77.19 | |
| SIVRBackbone=Ministral-8B-Instruct2026.04 | 77.56 | 72.16 | 77.19 | |
| SATMD + MSPmodel=Llama-3.2-3B-Instruct2026.04 | 77.14 | 71.43 | 71.22 | |
| TADBackbone=Llama-3.1-8B-Instruct2026.04 | 76.96 | 75.61 | 69.68 | |
| Lookback LensModel=Ministral-8B-Instruct2026.04 | 75.54 | 84.48 | 71.81 | |
| Lookback LensBackbone=Ministral-8B-Instruct2026.04 | 75.54 | 84.48 | 71.81 | |
| Lookback Lensmodel=Llama-3.2-3B-Instruct2026.04 | 75.19 | 77.36 | 78.25 | |
| Lookback LensBackbone=Llama-3.2-3B-Instruct2026.04 | 75.19 | 77.36 | 78.25 | |
| TADmodel=Llama-3.2-3B-Instruct2026.04 | 74.91 | 76.06 | 71.92 | |
| SEBackbone=Llama-3.1-8B-Instruct2026.04 | 73.58 | 70.73 | 76.55 | |
| SATMD + MSPBackbone=Llama-3.1-8B-Instruct2026.04 | 71.4 | 64.86 | 61.95 | |
| EntropyBackbone=Llama-3.1-8B-Instruct2026.04 | 71.2 | 71.86 | 65.12 | |
| SARModel=Ministral-8B-Instruct2026.04 | 70.87 | 75.51 | 73.46 | |
| CoE-CBackbone=Llama-3.1-8B-Instruct2026.04 | 70.85 | 84.43 | 70.53 | |
| EnergyModel=Ministral-8B-Instruct2026.04 | 70.85 | 83.85 | 68.19 | |
| SARBackbone=Llama-3.1-8B-Instruct2026.04 | 70.3 | 78.43 | 66.78 | |
| SEModel=Ministral-8B-Instruct2026.04 | 69.83 | 65.31 | 72.89 | |
| PerplexityBackbone=Llama-3.1-8B-Instruct2026.04 | 69.19 | 77.25 | 63.29 | |
| SARmodel=Llama-3.2-3B-Instruct2026.04 | 68.84 | 70.06 | 78.3 | |
| Max ProbBackbone=Llama-3.1-8B-Instruct2026.04 | 68.55 | 76.65 | 62.31 | |
| P(True)model=Llama-3.2-3B-Instruct2026.04 | 68.07 | 81.67 | 71.3 | |
| Temp-ScaledBackbone=Llama-3.1-8B-Instruct2026.04 | 67.45 | 78.44 | 61.01 | |
| SEmodel=Llama-3.2-3B-Instruct2026.04 | 67.27 | 78.41 | 68.09 | |
| EntropyModel=Ministral-8B-Instruct2026.04 | 66.26 | 91.72 | 66.94 | |
| PerplexityModel=Ministral-8B-Instruct2026.04 | 65.22 | 91.51 | 65.77 | |
| Max ProbModel=Ministral-8B-Instruct2026.04 | 64.56 | 92.55 | 65.1 | |
| Energymodel=Llama-3.2-3B-Instruct2026.04 | 64.33 | 88.1 | 69.14 | |
| Temp-ScaledModel=Ministral-8B-Instruct2026.04 | 64.02 | 91.51 | 64.46 | |
| Entropymodel=Llama-3.2-3B-Instruct2026.04 | 62.77 | 93.89 | 68.35 | |
| Perplexitymodel=Llama-3.2-3B-Instruct2026.04 | 61.66 | 94.21 | 67.05 | |
| Lookback LensBackbone=Llama-3.1-8B-Instruct2026.04 | 61.06 | 92.16 | 50.41 | |
| Lookback LensBackbone=Llama-3.1-8B-Instruct2026.04 | 61.06 | 92.16 | 50.41 | |
| Max Probmodel=Llama-3.2-3B-Instruct2026.04 | 60.68 | 93.57 | 66.27 | |
| Temp-Scaledmodel=Llama-3.2-3B-Instruct2026.04 | 60.18 | 93.57 | 65.69 | |
| CoE-Rmodel=Llama-3.2-3B-Instruct2026.04 | 59.5 | 88.42 | 65.57 | |
| CoE-Cmodel=Llama-3.2-3B-Instruct2026.04 | 58.67 | 90.68 | 63.95 | |
| P(True)Backbone=Llama-3.1-8B-Instruct2026.04 | 57.56 | 94.01 | 56.37 | |
| P(True)Model=Ministral-8B-Instruct2026.04 | 57.22 | 98.55 | 56.84 | |
| CoE-CModel=Ministral-8B-Instruct2026.04 | 55.53 | 94.41 | 57.88 | |
| CoE-RBackbone=Llama-3.1-8B-Instruct2026.04 | 53.39 | 91.62 | 50.66 | |
| CoE-RModel=Ministral-8B-Instruct2026.04 | 48.25 | 97.1 | 48.8 |