Uncertainty Quantification on MAQA ∆K−1
0.757KL Divergence AUCStructure-Aware Minimum Bayes Risk Decoding
Evaluation Results
| Method | Links | |
|---|---|---|
| Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Gemma-3-4B2026.01 | 0.757 | |
| Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Gemma-3-12B2026.01 | 0.756 | |
| Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Qwen3-30B-A3B2026.01 | 0.743 | |
| SEBackbone Model=Gemma-3-4B2026.01 | 0.723 | |
| COCOABackbone Model=Qwen3-30B-A3B2026.01 | 0.721 | |
| SEBackbone Model=Qwen3-30B-A3B2026.01 | 0.72 | |
| SARBackbone Model=Gemma-3-4B2026.01 | 0.718 | |
| SEBackbone Model=Gemma-3-12B2026.01 | 0.718 | |
| SARBackbone Model=Qwen3-30B-A3B2026.01 | 0.716 | |
| SARBackbone Model=Gemma-3-12B2026.01 | 0.715 | |
| KLEBackbone Model=Gemma-3-12B2026.01 | 0.708 | |
| KLEBackbone Model=Qwen3-30B-A3B2026.01 | 0.708 | |
| KLEBackbone Model=Gemma-3-4B2026.01 | 0.705 | |
| COCOABackbone Model=Gemma-3-4B2026.01 | 0.704 | |
| COCOABackbone Model=Gemma-3-12B2026.01 | 0.699 | |
| MSPBackbone Model=Qwen3-30B-A3B2026.01 | 0.69 | |
| MSPBackbone Model=Gemma-3-4B2026.01 | 0.689 | |
| MSPBackbone Model=Gemma-3-12B2026.01 | 0.678 | |
| Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Qwen3-4B2026.01 | 0.666 | |
| SARBackbone Model=Qwen3-4B2026.01 | 0.662 | |
| COCOABackbone Model=Qwen3-4B2026.01 | 0.658 | |
| MSPBackbone Model=Qwen3-4B2026.01 | 0.655 | |
| KLEBackbone Model=Qwen3-4B2026.01 | 0.639 | |
| SEBackbone Model=Qwen3-4B2026.01 | 0.639 | |
| P(True)Backbone Model=Qwen3-30B-A3B2026.01 | 0.541 | |
| P(True)Backbone Model=Qwen3-4B2026.01 | 0.54 | |
| P(True)Backbone Model=Gemma-3-12B2026.01 | 0.534 | |
| P(True)Backbone Model=Gemma-3-4B2026.01 | 0.474 |