Common-sense Reasoning on EWOK 2-Choice
6.6ΔAccContextual Calibration (CC)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Contextual Calibration (CC)Model=Llama3.1-8B2026.02 | 6.6 | — | — | -0.28 | |
| RBCorrModel=Llama3.1-8B2026.02 | 6.6 | — | — | -0.287 | |
| Batch Calibration (BC)Model=Llama3.1-8B2026.02 | 6 | — | — | -0.221 | |
| Contextual Calibration (CC)Model=Llama3.1-8B-Instruct2026.02 | 0.7 | — | — | -0.087 | |
| RBCorrModel=Llama3.1-8B-Instruct2026.02 | 0.5 | — | — | -0.104 | |
| Batch Calibration (BC)Model=Llama3.1-8B-Instruct2026.02 | -0.7 | — | — | -0.037 | |
| BaselineModel=Llama3.1-8B2026.02 | — | 65.9 | 0.305 | — | |
| BaselineModel=Llama3.1-8B-Instruct2026.02 | — | 75 | 0.122 | — |