Uncertainty Quantification on MAQA
83.5Hamming AUCStructure-Aware Minimum Bayes Risk Decoding
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Gemma-3-4B2026.01 | 83.5 | 82.8 | |
| Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Gemma-3-12B2026.01 | 81.6 | 79.5 | |
| SEBackbone Model=Gemma-3-4B2026.01 | 81.5 | 78.6 | |
| KLEBackbone Model=Gemma-3-4B2026.01 | 79.3 | 80.2 | |
| Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Qwen3-4B2026.01 | 79.1 | 80.7 | |
| SARBackbone Model=Gemma-3-12B2026.01 | 78.8 | 78.5 | |
| KLEBackbone Model=Gemma-3-12B2026.01 | 78.4 | 77.1 | |
| SEBackbone Model=Gemma-3-12B2026.01 | 78.3 | 76.2 | |
| Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Qwen3-30B-A3B2026.01 | 77.7 | 75.6 | |
| COCOABackbone Model=Gemma-3-12B2026.01 | 76.8 | 75.7 | |
| SARBackbone Model=Gemma-3-4B2026.01 | 76.4 | 80.2 | |
| SARBackbone Model=Qwen3-30B-A3B2026.01 | 76.4 | 74.8 | |
| COCOABackbone Model=Qwen3-30B-A3B2026.01 | 76 | 73.7 | |
| SEBackbone Model=Qwen3-30B-A3B2026.01 | 75.4 | 72.9 | |
| COCOABackbone Model=Gemma-3-4B2026.01 | 74.7 | 77.3 | |
| KLEBackbone Model=Qwen3-30B-A3B2026.01 | 74.6 | 72.4 | |
| SARBackbone Model=Qwen3-4B2026.01 | 73.5 | 73.3 | |
| COCOABackbone Model=Qwen3-4B2026.01 | 73.3 | 72.9 | |
| SEBackbone Model=Qwen3-4B2026.01 | 73.1 | 74.7 | |
| MSPBackbone Model=Qwen3-30B-A3B2026.01 | 72.9 | 71.2 | |
| MSPBackbone Model=Gemma-3-12B2026.01 | 72.4 | 71.5 | |
| KLEBackbone Model=Qwen3-4B2026.01 | 71.6 | 72.5 | |
| MSPBackbone Model=Qwen3-4B2026.01 | 70.7 | 69.8 | |
| MSPBackbone Model=Gemma-3-4B2026.01 | 70.4 | 72.6 | |
| P(True)Backbone Model=Gemma-3-12B2026.01 | 67.3 | 65.7 | |
| P(True)Backbone Model=Qwen3-4B2026.01 | 59.5 | 56.9 | |
| P(True)Backbone Model=Gemma-3-4B2026.01 | 56.6 | 55.6 | |
| P(True)Backbone Model=Qwen3-30B-A3B2026.01 | 49.3 | 50.6 |