Commonsense Reasoning on CommonSenseQA (ACC, BS, ECE, CL)
0.1054BSBase Model (Q0)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Base Model (Q0)Model Scale=14B2026.01 | 0.1054 | 82.56 | 0.0173 | -0.7201 | |
| Isotonic RegressionBackbone=Qwen3-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1119 | — | 0.0219 | -0.717 | |
| Isotonic RegressionModel Scale=14B, Category=Supervised Baseline2026.01 | 0.1135 | — | 0.019 | -0.7145 | |
| Base Model (Q0)Backbone=Qwen3-8B2026.01 | 0.1166 | 82.88 | 0.0601 | -0.7122 | |
| TS (Supervised)Backbone=Qwen3-8B, Protocol=Supervised Post-hoc2026.01 | 0.1166 | — | 0.0524 | -0.7122 | |
| Base Model (Q0)Backbone=Qwen3-8B, Method Category=Standard Model2026.01 | 0.1166 | 82.88 | 0.0601 | -0.7122 | |
| Temperature ScalingBackbone=Qwen3-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1166 | — | 0.0524 | -0.7122 | |
| Temperature ScalingModel Scale=14B, Category=Supervised Baseline2026.01 | 0.1193 | — | 0.064 | -0.7087 | |
| Histogram BinningBackbone=Qwen3-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1249 | — | 0.0254 | -0.7039 | |
| Base Model (Q0)Model Scale=4B2026.01 | 0.1252 | 78.46 | 0.0524 | -0.6594 | |
| Isotonic RegressionModel Scale=4B, Category=Supervised Baseline2026.01 | 0.1259 | — | 0.0286 | -0.6767 | |
| Ours (Label-Free Framework)Model Scale=14B2026.01 | 0.1262 | 83.37 | 0.0232 | -0.7075 | |
| Ours (Label-Free)Backbone=Qwen3-8B, Protocol=Label-Free Post-hoc2026.01 | 0.1291 | 83.29 | 0.0389 | -0.7038 | |
| label-free post-processing frameworkBackbone=Qwen3-8B, Method Category=Label Free2026.01 | 0.1291 | 83.29 | 0.0389 | -0.7038 | |
| Histogram BinningModel Scale=14B, Category=Supervised Baseline2026.01 | 0.1295 | — | 0.022 | -0.6985 | |
| Temperature ScalingModel Scale=4B, Category=Supervised Baseline2026.01 | 0.1322 | — | 0.046 | -0.6704 | |
| Instruct Model (Q1)Backbone=Qwen3-8B2026.01 | 0.143 | 82.88 | 0.1385 | -0.6859 | |
| Instruct Model (Q1)Backbone=Qwen3-8B, Method Category=Standard Model2026.01 | 0.143 | 82.88 | 0.1385 | -0.6859 | |
| Ours (Label-Free Framework)Model Scale=4B2026.01 | 0.1438 | 80.67 | 0.0297 | -0.6629 | |
| Histogram BinningModel Scale=4B, Category=Supervised Baseline2026.01 | 0.1449 | — | 0.0283 | -0.6577 | |
| Instruct Model (Q1)Model Scale=14B2026.01 | 0.1512 | 82.8 | 0.1483 | -0.6768 | |
| Isotonic RegressionBackbone=LLaMA-3.1-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1616 | — | 0.0427 | -0.6247 | |
| Base Model (Q0)Model Scale=1.7B2026.01 | 0.1636 | 70.84 | 0.0517 | -0.5449 | |
| Histogram BinningBackbone=LLaMA-3.1-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1655 | — | 0.0445 | -0.6207 | |
| Ours (Label-Free)Backbone=LLaMA-3.1-8B, Protocol=Label-Free Post-hoc2026.01 | 0.1657 | 78.71 | 0.029 | -0.6214 | |
| label-free post-processing frameworkBackbone=LLaMA-3.1-8B, Method Category=Label Free2026.01 | 0.1657 | 78.71 | 0.029 | -0.6214 | |
| Isotonic RegressionModel Scale=1.7B, Category=Supervised Baseline2026.01 | 0.169 | — | 0.0443 | -0.5468 | |
| Base Model (Q0)Backbone=LLaMA-3.1-8B2026.01 | 0.1698 | 70.93 | 0.0342 | -0.5395 | |
| Base Model (Q0)Backbone=LLaMA-3.1-8B, Method Category=Standard Model2026.01 | 0.1698 | 70.93 | 0.0342 | -0.5395 | |
| Instruct Model (Q1)Model Scale=4B2026.01 | 0.1703 | 80.26 | 0.166 | -0.6323 | |
| Temperature ScalingModel Scale=1.7B, Category=Supervised Baseline2026.01 | 0.1727 | — | 0.0645 | -0.5431 | |
| TS (Supervised)Backbone=LLaMA-3.1-8B, Protocol=Supervised Post-hoc2026.01 | 0.1749 | — | 0.0935 | -0.6114 | |
| Temperature ScalingBackbone=LLaMA-3.1-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1749 | — | 0.0935 | -0.6114 | |
| Histogram BinningBackbone=Ministral-3-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1771 | — | 0.0354 | -0.5633 | |
| Isotonic RegressionBackbone=Ministral-3-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1773 | — | 0.0345 | -0.5631 | |
| Ours (Label-Free)Backbone=Ministral-3-8B, Protocol=Label-Free Post-hoc2026.01 | 0.1777 | 74.3 | 0.0295 | -0.5653 | |
| label-free post-processing frameworkBackbone=Ministral-3-8B, Method Category=Label Free2026.01 | 0.1777 | 74.3 | 0.0295 | -0.5653 | |
| TS (Supervised)Backbone=Ministral-3-8B, Protocol=Supervised Post-hoc2026.01 | 0.1899 | — | 0.074 | -0.5505 | |
| Temperature ScalingBackbone=Ministral-3-8B, Method Category=Supervised Baseline, Calibration Protocol=Supervised2026.01 | 0.1899 | — | 0.074 | -0.5505 | |
| Histogram BinningModel Scale=1.7B, Category=Supervised Baseline2026.01 | 0.1934 | — | 0.0257 | -0.5224 | |
| Base Model (Q0)Backbone=Ministral-3-8B2026.01 | 0.196 | 70.27 | 0.1388 | -0.5067 | |
| Base Model (Q0)Backbone=Ministral-3-8B, Method Category=Standard Model2026.01 | 0.196 | 70.27 | 0.1388 | -0.5067 | |
| Base Model (Q0)Model Scale=0.6B2026.01 | 0.1966 | 58.56 | 0.0211 | -0.389 | |
| Ours (Label-Free Framework)Model Scale=1.7B2026.01 | 0.2002 | 71.66 | 0.0885 | -0.5165 | |
| Instruct Model (Q1)Backbone=LLaMA-3.1-8B2026.01 | 0.203 | 78.62 | 0.1999 | -0.5832 | |
| Instruct Model (Q1)Backbone=LLaMA-3.1-8B, Method Category=Standard Model2026.01 | 0.203 | 78.62 | 0.1999 | -0.5832 | |
| Instruct Model (Q1)Backbone=Ministral-3-8B2026.01 | 0.2041 | 74.04 | 0.1682 | -0.5363 | |
| Instruct Model (Q1)Backbone=Ministral-3-8B, Method Category=Standard Model2026.01 | 0.2041 | 74.04 | 0.1682 | -0.5363 | |
| Isotonic RegressionModel Scale=0.6B, Category=Supervised Baseline2026.01 | 0.2309 | — | 0.0155 | -0.3359 | |
| Temperature ScalingModel Scale=0.6B, Category=Supervised Baseline2026.01 | 0.235 | — | 0.0315 | -0.3318 | |
| Histogram BinningModel Scale=0.6B, Category=Supervised Baseline2026.01 | 0.237 | — | 0.0236 | -0.3297 | |
| Instruct Model (Q1)Model Scale=1.7B2026.01 | 0.2579 | 71.58 | 0.2574 | -0.4579 | |
| Ours (Label-Free Framework)Model Scale=0.6B2026.01 | 0.2847 | 56.67 | 0.2182 | -0.2821 | |
| Instruct Model (Q1)Model Scale=0.6B2026.01 | 0.3775 | 56.67 | 0.3767 | -0.1893 |