List your company

Uncertainty Quantification on MAQA ∆K−1

0.757KL Divergence AUCStructure-Aware Minimum Bayes Risk Decoding
0.462680.539090.61550.69191Jan 29, 2026

Evaluation results

MethodLinks
Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Gemma-3-4B2026.010.757
Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Gemma-3-12B2026.010.756
Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Qwen3-30B-A3B2026.010.743
SEBackbone Model=Gemma-3-4B2026.010.723
COCOABackbone Model=Qwen3-30B-A3B2026.010.721
SEBackbone Model=Qwen3-30B-A3B2026.010.72
SARBackbone Model=Gemma-3-4B2026.010.718
SEBackbone Model=Gemma-3-12B2026.010.718
SARBackbone Model=Qwen3-30B-A3B2026.010.716
SARBackbone Model=Gemma-3-12B2026.010.715
KLEBackbone Model=Gemma-3-12B2026.010.708
KLEBackbone Model=Qwen3-30B-A3B2026.010.708
KLEBackbone Model=Gemma-3-4B2026.010.705
COCOABackbone Model=Gemma-3-4B2026.010.704
COCOABackbone Model=Gemma-3-12B2026.010.699
MSPBackbone Model=Qwen3-30B-A3B2026.010.69
MSPBackbone Model=Gemma-3-4B2026.010.689
MSPBackbone Model=Gemma-3-12B2026.010.678
Structure-Aware Minimum Bayes Risk DecodingBackbone Model=Qwen3-4B2026.010.666
SARBackbone Model=Qwen3-4B2026.010.662
COCOABackbone Model=Qwen3-4B2026.010.658
MSPBackbone Model=Qwen3-4B2026.010.655
KLEBackbone Model=Qwen3-4B2026.010.639
SEBackbone Model=Qwen3-4B2026.010.639
P(True)Backbone Model=Qwen3-30B-A3B2026.010.541
P(True)Backbone Model=Qwen3-4B2026.010.54
P(True)Backbone Model=Gemma-3-12B2026.010.534
P(True)Backbone Model=Gemma-3-4B2026.010.474